[RFC v2 0/3] DRM Anonymous Inodes

[RFC v2 0/3] DRM Anonymous Inodes

[David Herrmann]

Hi

With 3.13-rc1 the required VFS core changes for anonymous backing inodes in DRM
got merged. This series reworks the previous patches (I think from early
August '13) and finally replaces the ugly drm_device->dev_mapping hack.

The patches should be fairly obvious. As I currently only have nouveau here, I'd
be happy to get some more testers with other drivers.

Patch #1 adds an internal private mount point to DRM core.
Patch #2 replaces ->dev_mapping with a private anonymous inode for each device.
Patch #3 removes the delayed dev_mapping initialization in TTM.

The internal mnt-point is based on fs/aio.c and fs/anon_inode.c which do the
same. It's not clear to me why we cannot share the mnt-point between all these,
but Al Viro made clear that he will reject any patches on anon_inode.c which do
that. So lets follow his recommendation and use our own internal mnt-point.

Thanks
David

David Herrmann (3):
  drm: add pseudo filesystem for shared inodes
  drm: use anon-inode instead of relying on cdevs
  drm: init TTM dev_mapping in ttm_bo_device_init()

 drivers/gpu/drm/ast/ast_ttm.c          |  5 +++--
 drivers/gpu/drm/cirrus/cirrus_ttm.c    |  5 +++--
 drivers/gpu/drm/drm_drv.c              | 37 ++++++++++++++++++++++++++++++++++
 drivers/gpu/drm/drm_fops.c             | 25 +++--------------------
 drivers/gpu/drm/drm_stub.c             | 12 ++++++++++-
 drivers/gpu/drm/i915/i915_gem.c        |  3 ++-
 drivers/gpu/drm/mgag200/mgag200_ttm.c  |  5 +++--
 drivers/gpu/drm/nouveau/nouveau_gem.c  |  2 --
 drivers/gpu/drm/nouveau/nouveau_ttm.c  |  4 +++-
 drivers/gpu/drm/omapdrm/omap_gem.c     | 34 +++++++++++++++----------------
 drivers/gpu/drm/qxl/qxl_object.c       |  2 --
 drivers/gpu/drm/qxl/qxl_ttm.c          |  6 +++---
 drivers/gpu/drm/radeon/radeon_object.c |  1 -
 drivers/gpu/drm/radeon/radeon_ttm.c    |  5 +++--
 drivers/gpu/drm/ttm/ttm_bo.c           |  3 ++-
 drivers/gpu/drm/vmwgfx/vmwgfx_drv.c    |  5 +++--
 include/drm/drmP.h                     |  3 ++-
 include/drm/drm_vma_manager.h          |  6 +++---
 include/drm/ttm/ttm_bo_driver.h        |  2 ++
 19 files changed, 100 insertions(+), 65 deletions(-)

-- 
1.8.5.2

[RFC v2 1/3] drm: add pseudo filesystem for shared inodes

[David Herrmann]

Our current DRM design uses a single address_space for all users of the
same DRM device. However, there is no way to create an anonymous
address_space without an underlying inode. Therefore, we wait for the
first ->open() callback on a registered char-dev and take-over the inode
of the char-dev. This worked well so far, but has several drawbacks:
 - We screw with FS internals and rely on some non-obvious invariants like
   inode->i_mapping being the same as inode->i_data for char-devs.
 - We don't have any address_space prior to the first ->open() from
   user-space. This leads to ugly fallback code and we cannot allocate
   global objects early.

As pointed out by Al-Viro, fs/anon_inode.c is *not* supposed to be used by
drivers for anonymous inode-allocation. Therefore, this patch follows the
proposed alternative solution and adds a pseudo filesystem mount-point to
DRM. We can then allocate private inodes including a private address_space
for each DRM device at initialization time.

Note that we could use:
  sysfs_get_inode(sysfs_mnt->mnt_sb, drm_device->dev->kobj.sd);
to get access to the underlying sysfs-inode of a "struct device" object.
However, most of this information is currently hidden and it's not clear
whether this address_space is suitable for driver access. Thus, unless
linux allows anonymous address_space objects or driver-core provides a
public inode per device, we're left with our own private internal mount
point.

Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
---
 drivers/gpu/drm/drm_drv.c | 37 +++++++++++++++++++++++++++++++++++++
 include/drm/drmP.h        |  1 +
 2 files changed, 38 insertions(+)

diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c
index 345be03..48ee03f 100644
--- a/drivers/gpu/drm/drm_drv.c
+++ b/drivers/gpu/drm/drm_drv.c
@@ -49,6 +49,8 @@
 #include <linux/debugfs.h>
 #include <linux/slab.h>
 #include <linux/export.h>
+#include <linux/fs.h>
+#include <linux/mount.h>
 #include <drm/drmP.h>
 #include <drm/drm_core.h>
 
@@ -171,6 +173,30 @@ static const struct drm_ioctl_desc drm_ioctls[] = {
 
 #define DRM_CORE_IOCTL_COUNT	ARRAY_SIZE( drm_ioctls )
 
+static struct vfsmount *drm_mnt;
+
+struct inode *drm_alloc_inode(void)
+{
+	return alloc_anon_inode(drm_mnt->mnt_sb);
+}
+
+static const struct dentry_operations drm_dops = {
+	.d_dname	= simple_dname,
+};
+
+static struct dentry *drm_mount(struct file_system_type *fs_type, int flags,
+				const char *dev_name, void *data)
+{
+	return mount_pseudo(fs_type, "drm:", NULL, &drm_dops, 0x010203ff);
+}
+
+static struct file_system_type drm_fs = {
+	.name		= "drm",
+	.owner		= THIS_MODULE,
+	.mount		= drm_mount,
+	.kill_sb	= kill_anon_super,
+};
+
 /** File operations structure */
 static const struct file_operations drm_stub_fops = {
 	.owner = THIS_MODULE,
@@ -203,9 +229,19 @@ static int __init drm_core_init(void)
 		goto err_p3;
 	}
 
+	drm_mnt = kern_mount(&drm_fs);
+	if (IS_ERR(drm_mnt)) {
+		ret = PTR_ERR(drm_mnt);
+		DRM_ERROR("Cannot mount pseudo fs: %d\n", ret);
+		goto err_p4;
+	}
+
 	DRM_INFO("Initialized %s %d.%d.%d %s\n",
 		 CORE_NAME, CORE_MAJOR, CORE_MINOR, CORE_PATCHLEVEL, CORE_DATE);
 	return 0;
+
+err_p4:
+	debugfs_remove(drm_debugfs_root);
 err_p3:
 	drm_sysfs_destroy();
 err_p2:
@@ -218,6 +254,7 @@ err_p1:
 
 static void __exit drm_core_exit(void)
 {
+	kern_unmount(drm_mnt);
 	debugfs_remove(drm_debugfs_root);
 	drm_sysfs_destroy();
 
diff --git a/include/drm/drmP.h b/include/drm/drmP.h
index 2fe9b5d..02637a5 100644
--- a/include/drm/drmP.h
+++ b/include/drm/drmP.h
@@ -1258,6 +1258,7 @@ extern long drm_ioctl(struct file *filp,
 extern long drm_compat_ioctl(struct file *filp,
 			     unsigned int cmd, unsigned long arg);
 extern int drm_lastclose(struct drm_device *dev);
+extern struct inode *drm_alloc_inode(void);
 
 				/* Device support (drm_fops.h) */
 extern struct mutex drm_global_mutex;
-- 
1.8.5.2

[RFC v2 2/3] drm: use anon-inode instead of relying on cdevs

[David Herrmann]

DRM drivers share a common address_space across all character-devices of a
single DRM device. This allows simple buffer eviction and mapping-control.
However, DRM core currently waits for the first ->open() on any char-dev
to mark the underlying inode as backing inode of the device. This delayed
initialization causes ugly conditions all over the place:
  if (dev->dev_mapping)
    do_sth();

To avoid delayed initialization and to stop reusing the inode of the
char-dev, we allocate an anonymous inode for each DRM device and reset
filp->f_mapping to it on ->open().

Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
---
 drivers/gpu/drm/ast/ast_ttm.c          |  2 +-
 drivers/gpu/drm/cirrus/cirrus_ttm.c    |  2 +-
 drivers/gpu/drm/drm_fops.c             | 25 +++----------------------
 drivers/gpu/drm/drm_stub.c             | 12 +++++++++++-
 drivers/gpu/drm/i915/i915_gem.c        |  3 ++-
 drivers/gpu/drm/mgag200/mgag200_ttm.c  |  2 +-
 drivers/gpu/drm/nouveau/nouveau_gem.c  |  2 +-
 drivers/gpu/drm/omapdrm/omap_gem.c     | 34 +++++++++++++++++-----------------
 drivers/gpu/drm/qxl/qxl_object.c       |  3 +--
 drivers/gpu/drm/qxl/qxl_ttm.c          |  3 +--
 drivers/gpu/drm/radeon/radeon_object.c |  2 +-
 drivers/gpu/drm/radeon/radeon_ttm.c    |  2 +-
 drivers/gpu/drm/vmwgfx/vmwgfx_drv.c    |  2 +-
 include/drm/drmP.h                     |  2 +-
 14 files changed, 43 insertions(+), 53 deletions(-)

diff --git a/drivers/gpu/drm/ast/ast_ttm.c b/drivers/gpu/drm/ast/ast_ttm.c
index 32aecb3..74eed74 100644
--- a/drivers/gpu/drm/ast/ast_ttm.c
+++ b/drivers/gpu/drm/ast/ast_ttm.c
@@ -324,7 +324,7 @@ int ast_bo_create(struct drm_device *dev, int size, int align,
 	}
 
 	astbo->bo.bdev = &ast->ttm.bdev;
-	astbo->bo.bdev->dev_mapping = dev->dev_mapping;
+	astbo->bo.bdev->dev_mapping = dev->anon_inode->i_mapping;
 
 	ast_ttm_placement(astbo, TTM_PL_FLAG_VRAM | TTM_PL_FLAG_SYSTEM);
 
diff --git a/drivers/gpu/drm/cirrus/cirrus_ttm.c b/drivers/gpu/drm/cirrus/cirrus_ttm.c
index 75becde..abd401d 100644
--- a/drivers/gpu/drm/cirrus/cirrus_ttm.c
+++ b/drivers/gpu/drm/cirrus/cirrus_ttm.c
@@ -329,7 +329,7 @@ int cirrus_bo_create(struct drm_device *dev, int size, int align,
 	}
 
 	cirrusbo->bo.bdev = &cirrus->ttm.bdev;
-	cirrusbo->bo.bdev->dev_mapping = dev->dev_mapping;
+	cirrusbo->bo.bdev->dev_mapping = dev->anon_inode->i_mapping;
 
 	cirrus_ttm_placement(cirrusbo, TTM_PL_FLAG_VRAM | TTM_PL_FLAG_SYSTEM);
 
diff --git a/drivers/gpu/drm/drm_fops.c b/drivers/gpu/drm/drm_fops.c
index 7f2af9a..147a84d 100644
--- a/drivers/gpu/drm/drm_fops.c
+++ b/drivers/gpu/drm/drm_fops.c
@@ -84,8 +84,6 @@ int drm_open(struct inode *inode, struct file *filp)
 	struct drm_minor *minor;
 	int retcode = 0;
 	int need_setup = 0;
-	struct address_space *old_mapping;
-	struct address_space *old_imapping;
 
 	minor = idr_find(&drm_minors_idr, minor_id);
 	if (!minor)
@@ -99,16 +97,9 @@ int drm_open(struct inode *inode, struct file *filp)
 
 	if (!dev->open_count++)
 		need_setup = 1;
-	mutex_lock(&dev->struct_mutex);
-	old_imapping = inode->i_mapping;
-	old_mapping = dev->dev_mapping;
-	if (old_mapping == NULL)
-		dev->dev_mapping = &inode->i_data;
-	/* ihold ensures nobody can remove inode with our i_data */
-	ihold(container_of(dev->dev_mapping, struct inode, i_data));
-	inode->i_mapping = dev->dev_mapping;
-	filp->f_mapping = dev->dev_mapping;
-	mutex_unlock(&dev->struct_mutex);
+
+	/* share address_space across all char-devs of a single device */
+	filp->f_mapping = dev->anon_inode->i_mapping;
 
 	retcode = drm_open_helper(inode, filp, dev);
 	if (retcode)
@@ -121,12 +112,6 @@ int drm_open(struct inode *inode, struct file *filp)
 	return 0;
 
 err_undo:
-	mutex_lock(&dev->struct_mutex);
-	filp->f_mapping = old_imapping;
-	inode->i_mapping = old_imapping;
-	iput(container_of(dev->dev_mapping, struct inode, i_data));
-	dev->dev_mapping = old_mapping;
-	mutex_unlock(&dev->struct_mutex);
 	dev->open_count--;
 	return retcode;
 }
@@ -434,7 +419,6 @@ int drm_lastclose(struct drm_device * dev)
 
 	drm_legacy_dma_takedown(dev);
 
-	dev->dev_mapping = NULL;
 	mutex_unlock(&dev->struct_mutex);
 
 	drm_legacy_dev_reinit(dev);
@@ -549,9 +533,6 @@ int drm_release(struct inode *inode, struct file *filp)
 		}
 	}
 
-	BUG_ON(dev->dev_mapping == NULL);
-	iput(container_of(dev->dev_mapping, struct inode, i_data));
-
 	/* drop the reference held my the file priv */
 	if (file_priv->master)
 		drm_master_put(&file_priv->master);
diff --git a/drivers/gpu/drm/drm_stub.c b/drivers/gpu/drm/drm_stub.c
index a4a5c6a..0caf35f 100644
--- a/drivers/gpu/drm/drm_stub.c
+++ b/drivers/gpu/drm/drm_stub.c
@@ -452,8 +452,15 @@ struct drm_device *drm_dev_alloc(struct drm_driver *driver,
 	mutex_init(&dev->struct_mutex);
 	mutex_init(&dev->ctxlist_mutex);
 
-	if (drm_ht_create(&dev->map_hash, 12))
+	dev->anon_inode = drm_alloc_inode();
+	if (IS_ERR(dev->anon_inode)) {
+		ret = PTR_ERR(dev->anon_inode);
+		DRM_ERROR("Cannot allocate anonymous inode: %d\n", ret);
 		goto err_free;
+	}
+
+	if (drm_ht_create(&dev->map_hash, 12))
+		goto err_inode;
 
 	ret = drm_ctxbitmap_init(dev);
 	if (ret) {
@@ -475,6 +482,8 @@ err_ctxbitmap:
 	drm_ctxbitmap_cleanup(dev);
 err_ht:
 	drm_ht_remove(&dev->map_hash);
+err_inode:
+	iput(dev->anon_inode);
 err_free:
 	kfree(dev);
 	return NULL;
@@ -502,6 +511,7 @@ void drm_dev_free(struct drm_device *dev)
 
 	drm_ctxbitmap_cleanup(dev);
 	drm_ht_remove(&dev->map_hash);
+	iput(dev->anon_inode);
 
 	kfree(dev->devname);
 	kfree(dev);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 182c521..8cc47da 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1474,7 +1474,8 @@ i915_gem_release_mmap(struct drm_i915_gem_object *obj)
 	if (!obj->fault_mappable)
 		return;
 
-	drm_vma_node_unmap(&obj->base.vma_node, obj->base.dev->dev_mapping);
+	drm_vma_node_unmap(&obj->base.vma_node,
+			   obj->base.dev->anon_inode->i_mapping);
 	obj->fault_mappable = false;
 }
 
diff --git a/drivers/gpu/drm/mgag200/mgag200_ttm.c b/drivers/gpu/drm/mgag200/mgag200_ttm.c
index 07b192f..bb2a1cb 100644
--- a/drivers/gpu/drm/mgag200/mgag200_ttm.c
+++ b/drivers/gpu/drm/mgag200/mgag200_ttm.c
@@ -324,7 +324,7 @@ int mgag200_bo_create(struct drm_device *dev, int size, int align,
 	}
 
 	mgabo->bo.bdev = &mdev->ttm.bdev;
-	mgabo->bo.bdev->dev_mapping = dev->dev_mapping;
+	mgabo->bo.bdev->dev_mapping = dev->anon_inode->i_mapping;
 
 	mgag200_ttm_placement(mgabo, TTM_PL_FLAG_VRAM | TTM_PL_FLAG_SYSTEM);
 
diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c
index 0447163..aafb8d6 100644
--- a/drivers/gpu/drm/nouveau/nouveau_gem.c
+++ b/drivers/gpu/drm/nouveau/nouveau_gem.c
@@ -228,7 +228,7 @@ nouveau_gem_ioctl_new(struct drm_device *dev, void *data,
 	struct nouveau_bo *nvbo = NULL;
 	int ret = 0;
 
-	drm->ttm.bdev.dev_mapping = drm->dev->dev_mapping;
+	drm->ttm.bdev.dev_mapping = drm->dev->anon_inode->i_mapping;
 
 	if (!pfb->memtype_valid(pfb, req->info.tile_flags)) {
 		NV_ERROR(cli, "bad page flags: 0x%08x\n", req->info.tile_flags);
diff --git a/drivers/gpu/drm/omapdrm/omap_gem.c b/drivers/gpu/drm/omapdrm/omap_gem.c
index 5aec3e8..c8d9727 100644
--- a/drivers/gpu/drm/omapdrm/omap_gem.c
+++ b/drivers/gpu/drm/omapdrm/omap_gem.c
@@ -153,24 +153,24 @@ static struct {
 static void evict_entry(struct drm_gem_object *obj,
 		enum tiler_fmt fmt, struct usergart_entry *entry)
 {
-	if (obj->dev->dev_mapping) {
-		struct omap_gem_object *omap_obj = to_omap_bo(obj);
-		int n = usergart[fmt].height;
-		size_t size = PAGE_SIZE * n;
-		loff_t off = mmap_offset(obj) +
-				(entry->obj_pgoff << PAGE_SHIFT);
-		const int m = 1 + ((omap_obj->width << fmt) / PAGE_SIZE);
-		if (m > 1) {
-			int i;
-			/* if stride > than PAGE_SIZE then sparse mapping: */
-			for (i = n; i > 0; i--) {
-				unmap_mapping_range(obj->dev->dev_mapping,
-						off, PAGE_SIZE, 1);
-				off += PAGE_SIZE * m;
-			}
-		} else {
-			unmap_mapping_range(obj->dev->dev_mapping, off, size, 1);
+	struct omap_gem_object *omap_obj = to_omap_bo(obj);
+	int n = usergart[fmt].height;
+	size_t size = PAGE_SIZE * n;
+	loff_t off = mmap_offset(obj) +
+			(entry->obj_pgoff << PAGE_SHIFT);
+	const int m = 1 + ((omap_obj->width << fmt) / PAGE_SIZE);
+
+	if (m > 1) {
+		int i;
+		/* if stride > than PAGE_SIZE then sparse mapping: */
+		for (i = n; i > 0; i--) {
+			unmap_mapping_range(obj->dev->anon_inode->i_mapping,
+					    off, PAGE_SIZE, 1);
+			off += PAGE_SIZE * m;
 		}
+	} else {
+		unmap_mapping_range(obj->dev->anon_inode->i_mapping,
+				    off, size, 1);
 	}
 
 	entry->obj = NULL;
diff --git a/drivers/gpu/drm/qxl/qxl_object.c b/drivers/gpu/drm/qxl/qxl_object.c
index 8691c76..7e20c21 100644
--- a/drivers/gpu/drm/qxl/qxl_object.c
+++ b/drivers/gpu/drm/qxl/qxl_object.c
@@ -82,8 +82,7 @@ int qxl_bo_create(struct qxl_device *qdev,
 	enum ttm_bo_type type;
 	int r;
 
-	if (unlikely(qdev->mman.bdev.dev_mapping == NULL))
-		qdev->mman.bdev.dev_mapping = qdev->ddev->dev_mapping;
+	qdev->mman.bdev.dev_mapping = qdev->ddev->anon_inode->i_mapping;
 	if (kernel)
 		type = ttm_bo_type_kernel;
 	else
diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
index c7e7e65..78cbc40 100644
--- a/drivers/gpu/drm/qxl/qxl_ttm.c
+++ b/drivers/gpu/drm/qxl/qxl_ttm.c
@@ -518,8 +518,7 @@ int qxl_ttm_init(struct qxl_device *qdev)
 		 ((unsigned)num_io_pages * PAGE_SIZE) / (1024 * 1024));
 	DRM_INFO("qxl: %uM of Surface memory size\n",
 		 (unsigned)qdev->surfaceram_size / (1024 * 1024));
-	if (unlikely(qdev->mman.bdev.dev_mapping == NULL))
-		qdev->mman.bdev.dev_mapping = qdev->ddev->dev_mapping;
+	qdev->mman.bdev.dev_mapping = qdev->ddev->anon_inode->i_mapping;
 	r = qxl_ttm_debugfs_init(qdev);
 	if (r) {
 		DRM_ERROR("Failed to init debugfs\n");
diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
index c0fa4aa..0900b36 100644
--- a/drivers/gpu/drm/radeon/radeon_object.c
+++ b/drivers/gpu/drm/radeon/radeon_object.c
@@ -120,7 +120,7 @@ int radeon_bo_create(struct radeon_device *rdev,
 
 	size = ALIGN(size, PAGE_SIZE);
 
-	rdev->mman.bdev.dev_mapping = rdev->ddev->dev_mapping;
+	rdev->mman.bdev.dev_mapping = rdev->ddev->anon_inode->i_mapping;
 	if (kernel) {
 		type = ttm_bo_type_kernel;
 	} else if (sg) {
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index 051fa87..669c448 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -737,7 +737,7 @@ int radeon_ttm_init(struct radeon_device *rdev)
 	}
 	DRM_INFO("radeon: %uM of GTT memory ready.\n",
 		 (unsigned)(rdev->mc.gtt_size / (1024 * 1024)));
-	rdev->mman.bdev.dev_mapping = rdev->ddev->dev_mapping;
+	rdev->mman.bdev.dev_mapping = rdev->ddev->anon_inode->i_mapping;
 
 	r = radeon_ttm_debugfs_init(rdev);
 	if (r) {
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
index c7a5496..218b711 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
@@ -860,7 +860,7 @@ static int vmw_driver_open(struct drm_device *dev, struct drm_file *file_priv)
 		goto out_no_tfile;
 
 	file_priv->driver_priv = vmw_fp;
-	dev_priv->bdev.dev_mapping = dev->dev_mapping;
+	dev_priv->bdev.dev_mapping = dev->anon_inode->i_mapping;
 
 	return 0;
 
diff --git a/include/drm/drmP.h b/include/drm/drmP.h
index 02637a5..48d3720 100644
--- a/include/drm/drmP.h
+++ b/include/drm/drmP.h
@@ -1186,7 +1186,7 @@ struct drm_device {
 	unsigned int num_crtcs;                  /**< Number of CRTCs on this device */
 	void *dev_private;		/**< device private data */
 	void *mm_private;
-	struct address_space *dev_mapping;
+	struct inode *anon_inode;
 	struct drm_sigdata sigdata;	   /**< For block_all_signals */
 	sigset_t sigmask;
 
-- 
1.8.5.2

[RFC v2 3/3] drm: init TTM dev_mapping in ttm_bo_device_init()

[David Herrmann]

With dev->anon_inode we have a global address_space ready for operation
right from the beginning. Therefore, there is no need to do a delayed
setup with TTM. Instead, set dev_mapping during initialization in
ttm_bo_device_init() and remove any "if (dev_mapping)" conditions.

Cc: Dave Airlie <airlied@redhat.com>
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Cc: Alex Deucher <alexdeucher@gmail.com>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
---
 drivers/gpu/drm/ast/ast_ttm.c          | 5 +++--
 drivers/gpu/drm/cirrus/cirrus_ttm.c    | 5 +++--
 drivers/gpu/drm/mgag200/mgag200_ttm.c  | 5 +++--
 drivers/gpu/drm/nouveau/nouveau_gem.c  | 2 --
 drivers/gpu/drm/nouveau/nouveau_ttm.c  | 4 +++-
 drivers/gpu/drm/qxl/qxl_object.c       | 1 -
 drivers/gpu/drm/qxl/qxl_ttm.c          | 5 +++--
 drivers/gpu/drm/radeon/radeon_object.c | 1 -
 drivers/gpu/drm/radeon/radeon_ttm.c    | 5 +++--
 drivers/gpu/drm/ttm/ttm_bo.c           | 3 ++-
 drivers/gpu/drm/vmwgfx/vmwgfx_drv.c    | 5 +++--
 include/drm/drm_vma_manager.h          | 6 +++---
 include/drm/ttm/ttm_bo_driver.h        | 2 ++
 13 files changed, 28 insertions(+), 21 deletions(-)

diff --git a/drivers/gpu/drm/ast/ast_ttm.c b/drivers/gpu/drm/ast/ast_ttm.c
index 74eed74..5465013 100644
--- a/drivers/gpu/drm/ast/ast_ttm.c
+++ b/drivers/gpu/drm/ast/ast_ttm.c
@@ -259,7 +259,9 @@ int ast_mm_init(struct ast_private *ast)
 
 	ret = ttm_bo_device_init(&ast->ttm.bdev,
 				 ast->ttm.bo_global_ref.ref.object,
-				 &ast_bo_driver, DRM_FILE_PAGE_OFFSET,
+				 &ast_bo_driver,
+				 dev->anon_inode->i_mapping,
+				 DRM_FILE_PAGE_OFFSET,
 				 true);
 	if (ret) {
 		DRM_ERROR("Error initialising bo driver; %d\n", ret);
@@ -324,7 +326,6 @@ int ast_bo_create(struct drm_device *dev, int size, int align,
 	}
 
 	astbo->bo.bdev = &ast->ttm.bdev;
-	astbo->bo.bdev->dev_mapping = dev->anon_inode->i_mapping;
 
 	ast_ttm_placement(astbo, TTM_PL_FLAG_VRAM | TTM_PL_FLAG_SYSTEM);
 
diff --git a/drivers/gpu/drm/cirrus/cirrus_ttm.c b/drivers/gpu/drm/cirrus/cirrus_ttm.c
index abd401d..ee7703e 100644
--- a/drivers/gpu/drm/cirrus/cirrus_ttm.c
+++ b/drivers/gpu/drm/cirrus/cirrus_ttm.c
@@ -259,7 +259,9 @@ int cirrus_mm_init(struct cirrus_device *cirrus)
 
 	ret = ttm_bo_device_init(&cirrus->ttm.bdev,
 				 cirrus->ttm.bo_global_ref.ref.object,
-				 &cirrus_bo_driver, DRM_FILE_PAGE_OFFSET,
+				 &cirrus_bo_driver,
+				 dev->anon_inode->i_mapping,
+				 DRM_FILE_PAGE_OFFSET,
 				 true);
 	if (ret) {
 		DRM_ERROR("Error initialising bo driver; %d\n", ret);
@@ -329,7 +331,6 @@ int cirrus_bo_create(struct drm_device *dev, int size, int align,
 	}
 
 	cirrusbo->bo.bdev = &cirrus->ttm.bdev;
-	cirrusbo->bo.bdev->dev_mapping = dev->anon_inode->i_mapping;
 
 	cirrus_ttm_placement(cirrusbo, TTM_PL_FLAG_VRAM | TTM_PL_FLAG_SYSTEM);
 
diff --git a/drivers/gpu/drm/mgag200/mgag200_ttm.c b/drivers/gpu/drm/mgag200/mgag200_ttm.c
index bb2a1cb..0893f7c 100644
--- a/drivers/gpu/drm/mgag200/mgag200_ttm.c
+++ b/drivers/gpu/drm/mgag200/mgag200_ttm.c
@@ -259,7 +259,9 @@ int mgag200_mm_init(struct mga_device *mdev)
 
 	ret = ttm_bo_device_init(&mdev->ttm.bdev,
 				 mdev->ttm.bo_global_ref.ref.object,
-				 &mgag200_bo_driver, DRM_FILE_PAGE_OFFSET,
+				 &mgag200_bo_driver,
+				 dev->anon_inode->i_mapping,
+				 DRM_FILE_PAGE_OFFSET,
 				 true);
 	if (ret) {
 		DRM_ERROR("Error initialising bo driver; %d\n", ret);
@@ -324,7 +326,6 @@ int mgag200_bo_create(struct drm_device *dev, int size, int align,
 	}
 
 	mgabo->bo.bdev = &mdev->ttm.bdev;
-	mgabo->bo.bdev->dev_mapping = dev->anon_inode->i_mapping;
 
 	mgag200_ttm_placement(mgabo, TTM_PL_FLAG_VRAM | TTM_PL_FLAG_SYSTEM);
 
diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c
index aafb8d6..1eeb393 100644
--- a/drivers/gpu/drm/nouveau/nouveau_gem.c
+++ b/drivers/gpu/drm/nouveau/nouveau_gem.c
@@ -228,8 +228,6 @@ nouveau_gem_ioctl_new(struct drm_device *dev, void *data,
 	struct nouveau_bo *nvbo = NULL;
 	int ret = 0;
 
-	drm->ttm.bdev.dev_mapping = drm->dev->anon_inode->i_mapping;
-
 	if (!pfb->memtype_valid(pfb, req->info.tile_flags)) {
 		NV_ERROR(cli, "bad page flags: 0x%08x\n", req->info.tile_flags);
 		return -EINVAL;
diff --git a/drivers/gpu/drm/nouveau/nouveau_ttm.c b/drivers/gpu/drm/nouveau/nouveau_ttm.c
index 19e3757..d486f25 100644
--- a/drivers/gpu/drm/nouveau/nouveau_ttm.c
+++ b/drivers/gpu/drm/nouveau/nouveau_ttm.c
@@ -375,7 +375,9 @@ nouveau_ttm_init(struct nouveau_drm *drm)
 
 	ret = ttm_bo_device_init(&drm->ttm.bdev,
 				  drm->ttm.bo_global_ref.ref.object,
-				  &nouveau_bo_driver, DRM_FILE_PAGE_OFFSET,
+				  &nouveau_bo_driver,
+				  dev->anon_inode->i_mapping,
+				  DRM_FILE_PAGE_OFFSET,
 				  bits <= 32 ? true : false);
 	if (ret) {
 		NV_ERROR(drm, "error initialising bo driver, %d\n", ret);
diff --git a/drivers/gpu/drm/qxl/qxl_object.c b/drivers/gpu/drm/qxl/qxl_object.c
index 7e20c21..b95f144 100644
--- a/drivers/gpu/drm/qxl/qxl_object.c
+++ b/drivers/gpu/drm/qxl/qxl_object.c
@@ -82,7 +82,6 @@ int qxl_bo_create(struct qxl_device *qdev,
 	enum ttm_bo_type type;
 	int r;
 
-	qdev->mman.bdev.dev_mapping = qdev->ddev->anon_inode->i_mapping;
 	if (kernel)
 		type = ttm_bo_type_kernel;
 	else
diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
index 78cbc40..29c02e0 100644
--- a/drivers/gpu/drm/qxl/qxl_ttm.c
+++ b/drivers/gpu/drm/qxl/qxl_ttm.c
@@ -493,7 +493,9 @@ int qxl_ttm_init(struct qxl_device *qdev)
 	/* No others user of address space so set it to 0 */
 	r = ttm_bo_device_init(&qdev->mman.bdev,
 			       qdev->mman.bo_global_ref.ref.object,
-			       &qxl_bo_driver, DRM_FILE_PAGE_OFFSET, 0);
+			       &qxl_bo_driver,
+			       qdev->ddev->anon_inode->i_mapping,
+			       DRM_FILE_PAGE_OFFSET, 0);
 	if (r) {
 		DRM_ERROR("failed initializing buffer object driver(%d).\n", r);
 		return r;
@@ -518,7 +520,6 @@ int qxl_ttm_init(struct qxl_device *qdev)
 		 ((unsigned)num_io_pages * PAGE_SIZE) / (1024 * 1024));
 	DRM_INFO("qxl: %uM of Surface memory size\n",
 		 (unsigned)qdev->surfaceram_size / (1024 * 1024));
-	qdev->mman.bdev.dev_mapping = qdev->ddev->anon_inode->i_mapping;
 	r = qxl_ttm_debugfs_init(qdev);
 	if (r) {
 		DRM_ERROR("Failed to init debugfs\n");
diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
index 0900b36..c2f728c 100644
--- a/drivers/gpu/drm/radeon/radeon_object.c
+++ b/drivers/gpu/drm/radeon/radeon_object.c
@@ -120,7 +120,6 @@ int radeon_bo_create(struct radeon_device *rdev,
 
 	size = ALIGN(size, PAGE_SIZE);
 
-	rdev->mman.bdev.dev_mapping = rdev->ddev->anon_inode->i_mapping;
 	if (kernel) {
 		type = ttm_bo_type_kernel;
 	} else if (sg) {
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index 669c448..12188de 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -699,7 +699,9 @@ int radeon_ttm_init(struct radeon_device *rdev)
 	/* No others user of address space so set it to 0 */
 	r = ttm_bo_device_init(&rdev->mman.bdev,
 			       rdev->mman.bo_global_ref.ref.object,
-			       &radeon_bo_driver, DRM_FILE_PAGE_OFFSET,
+			       &radeon_bo_driver,
+			       rdev->ddev->anon_inode->i_mapping,
+			       DRM_FILE_PAGE_OFFSET,
 			       rdev->need_dma32);
 	if (r) {
 		DRM_ERROR("failed initializing buffer object driver(%d).\n", r);
@@ -737,7 +739,6 @@ int radeon_ttm_init(struct radeon_device *rdev)
 	}
 	DRM_INFO("radeon: %uM of GTT memory ready.\n",
 		 (unsigned)(rdev->mc.gtt_size / (1024 * 1024)));
-	rdev->mman.bdev.dev_mapping = rdev->ddev->anon_inode->i_mapping;
 
 	r = radeon_ttm_debugfs_init(rdev);
 	if (r) {
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index 07e02c4..23cf47d 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -1449,6 +1449,7 @@ EXPORT_SYMBOL(ttm_bo_device_release);
 int ttm_bo_device_init(struct ttm_bo_device *bdev,
 		       struct ttm_bo_global *glob,
 		       struct ttm_bo_driver *driver,
+		       struct address_space *mapping,
 		       uint64_t file_page_offset,
 		       bool need_dma32)
 {
@@ -1470,7 +1471,7 @@ int ttm_bo_device_init(struct ttm_bo_device *bdev,
 				    0x10000000);
 	INIT_DELAYED_WORK(&bdev->wq, ttm_bo_delayed_workqueue);
 	INIT_LIST_HEAD(&bdev->ddestroy);
-	bdev->dev_mapping = NULL;
+	bdev->dev_mapping = mapping;
 	bdev->glob = glob;
 	bdev->need_dma32 = need_dma32;
 	bdev->val_seq = 0;
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
index 218b711..7c32303 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
@@ -632,7 +632,9 @@ static int vmw_driver_load(struct drm_device *dev, unsigned long chipset)
 
 	ret = ttm_bo_device_init(&dev_priv->bdev,
 				 dev_priv->bo_global_ref.ref.object,
-				 &vmw_bo_driver, VMWGFX_FILE_PAGE_OFFSET,
+				 &vmw_bo_driver,
+				 dev->anon_inode->i_mapping,
+				 VMWGFX_FILE_PAGE_OFFSET,
 				 false);
 	if (unlikely(ret != 0)) {
 		DRM_ERROR("Failed initializing TTM buffer object driver.\n");
@@ -860,7 +862,6 @@ static int vmw_driver_open(struct drm_device *dev, struct drm_file *file_priv)
 		goto out_no_tfile;
 
 	file_priv->driver_priv = vmw_fp;
-	dev_priv->bdev.dev_mapping = dev->anon_inode->i_mapping;
 
 	return 0;
 
diff --git a/include/drm/drm_vma_manager.h b/include/drm/drm_vma_manager.h
index c18a593..8cd402c7 100644
--- a/include/drm/drm_vma_manager.h
+++ b/include/drm/drm_vma_manager.h
@@ -221,8 +221,8 @@ static inline __u64 drm_vma_node_offset_addr(struct drm_vma_offset_node *node)
  * @file_mapping: Address space to unmap @node from
  *
  * Unmap all userspace mappings for a given offset node. The mappings must be
- * associated with the @file_mapping address-space. If no offset exists or
- * the address-space is invalid, nothing is done.
+ * associated with the @file_mapping address-space. If no offset exists
+ * nothing is done.
  *
  * This call is unlocked. The caller must guarantee that drm_vma_offset_remove()
  * is not called on this node concurrently.
@@ -230,7 +230,7 @@ static inline __u64 drm_vma_node_offset_addr(struct drm_vma_offset_node *node)
 static inline void drm_vma_node_unmap(struct drm_vma_offset_node *node,
 				      struct address_space *file_mapping)
 {
-	if (file_mapping && drm_vma_node_has_offset(node))
+	if (drm_vma_node_has_offset(node))
 		unmap_mapping_range(file_mapping,
 				    drm_vma_node_offset_addr(node),
 				    drm_vma_node_size(node) << PAGE_SHIFT, 1);
diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h
index 8639c85..9261e42 100644
--- a/include/drm/ttm/ttm_bo_driver.h
+++ b/include/drm/ttm/ttm_bo_driver.h
@@ -738,6 +738,7 @@ extern int ttm_bo_device_release(struct ttm_bo_device *bdev);
  * @bdev: A pointer to a struct ttm_bo_device to initialize.
  * @glob: A pointer to an initialized struct ttm_bo_global.
  * @driver: A pointer to a struct ttm_bo_driver set up by the caller.
+ * @mapping: The address space to use for this bo.
  * @file_page_offset: Offset into the device address space that is available
  * for buffer data. This ensures compatibility with other users of the
  * address space.
@@ -749,6 +750,7 @@ extern int ttm_bo_device_release(struct ttm_bo_device *bdev);
 extern int ttm_bo_device_init(struct ttm_bo_device *bdev,
 			      struct ttm_bo_global *glob,
 			      struct ttm_bo_driver *driver,
+			      struct address_space *mapping,
 			      uint64_t file_page_offset, bool need_dma32);
 
 /**
-- 
1.8.5.2

Re: [RFC v2 0/3] DRM Anonymous Inodes

[David Herrmann]

Hi

On Fri, Jan 3, 2014 at 3:41 PM, David Herrmann <dh.herrmann@gmail.com> wrote:
> Hi
>
> With 3.13-rc1 the required VFS core changes for anonymous backing inodes in DRM
> got merged. This series reworks the previous patches (I think from early
> August '13) and finally replaces the ugly drm_device->dev_mapping hack.
>
> The patches should be fairly obvious. As I currently only have nouveau here, I'd
> be happy to get some more testers with other drivers.
>
> Patch #1 adds an internal private mount point to DRM core.
> Patch #2 replaces ->dev_mapping with a private anonymous inode for each device.
> Patch #3 removes the delayed dev_mapping initialization in TTM.
>
> The internal mnt-point is based on fs/aio.c and fs/anon_inode.c which do the
> same. It's not clear to me why we cannot share the mnt-point between all these,
> but Al Viro made clear that he will reject any patches on anon_inode.c which do
> that. So lets follow his recommendation and use our own internal mnt-point.

Also tested on i915 ivb now (before: nouveau+TTM). I think this is
ready for 3.14 if nobody finds any issues.

Thanks
David

> Thanks
> David
>
> David Herrmann (3):
>   drm: add pseudo filesystem for shared inodes
>   drm: use anon-inode instead of relying on cdevs
>   drm: init TTM dev_mapping in ttm_bo_device_init()
>
>  drivers/gpu/drm/ast/ast_ttm.c          |  5 +++--
>  drivers/gpu/drm/cirrus/cirrus_ttm.c    |  5 +++--
>  drivers/gpu/drm/drm_drv.c              | 37 ++++++++++++++++++++++++++++++++++
>  drivers/gpu/drm/drm_fops.c             | 25 +++--------------------
>  drivers/gpu/drm/drm_stub.c             | 12 ++++++++++-
>  drivers/gpu/drm/i915/i915_gem.c        |  3 ++-
>  drivers/gpu/drm/mgag200/mgag200_ttm.c  |  5 +++--
>  drivers/gpu/drm/nouveau/nouveau_gem.c  |  2 --
>  drivers/gpu/drm/nouveau/nouveau_ttm.c  |  4 +++-
>  drivers/gpu/drm/omapdrm/omap_gem.c     | 34 +++++++++++++++----------------
>  drivers/gpu/drm/qxl/qxl_object.c       |  2 --
>  drivers/gpu/drm/qxl/qxl_ttm.c          |  6 +++---
>  drivers/gpu/drm/radeon/radeon_object.c |  1 -
>  drivers/gpu/drm/radeon/radeon_ttm.c    |  5 +++--
>  drivers/gpu/drm/ttm/ttm_bo.c           |  3 ++-
>  drivers/gpu/drm/vmwgfx/vmwgfx_drv.c    |  5 +++--
>  include/drm/drmP.h                     |  3 ++-
>  include/drm/drm_vma_manager.h          |  6 +++---
>  include/drm/ttm/ttm_bo_driver.h        |  2 ++
>  19 files changed, 100 insertions(+), 65 deletions(-)
>
> --
> 1.8.5.2
>

Re: [RFC v2 0/3] DRM Anonymous Inodes

[Thomas Hellstrom]

David,

I haven't had time to test on vmwgfx yet, but looking through the code
it looks good to me.

Acked-by: Thomas Hellstrom <thellstrom@vmware.com>

/Thomas

On 01/16/2014 12:29 PM, David Herrmann wrote:
> Hi
>
> On Fri, Jan 3, 2014 at 3:41 PM, David Herrmann <dh.herrmann@gmail.com> wrote:
>> Hi
>>
>> With 3.13-rc1 the required VFS core changes for anonymous backing inodes in DRM
>> got merged. This series reworks the previous patches (I think from early
>> August '13) and finally replaces the ugly drm_device->dev_mapping hack.
>>
>> The patches should be fairly obvious. As I currently only have nouveau here, I'd
>> be happy to get some more testers with other drivers.
>>
>> Patch #1 adds an internal private mount point to DRM core.
>> Patch #2 replaces ->dev_mapping with a private anonymous inode for each device.
>> Patch #3 removes the delayed dev_mapping initialization in TTM.
>>
>> The internal mnt-point is based on fs/aio.c and fs/anon_inode.c which do the
>> same. It's not clear to me why we cannot share the mnt-point between all these,
>> but Al Viro made clear that he will reject any patches on anon_inode.c which do
>> that. So lets follow his recommendation and use our own internal mnt-point.
> Also tested on i915 ivb now (before: nouveau+TTM). I think this is
> ready for 3.14 if nobody finds any issues.
>
> Thanks
> David
>
>> Thanks
>> David
>>
>> David Herrmann (3):
>>   drm: add pseudo filesystem for shared inodes
>>   drm: use anon-inode instead of relying on cdevs
>>   drm: init TTM dev_mapping in ttm_bo_device_init()
>>
>>  drivers/gpu/drm/ast/ast_ttm.c          |  5 +++--
>>  drivers/gpu/drm/cirrus/cirrus_ttm.c    |  5 +++--
>>  drivers/gpu/drm/drm_drv.c              | 37 ++++++++++++++++++++++++++++++++++
>>  drivers/gpu/drm/drm_fops.c             | 25 +++--------------------
>>  drivers/gpu/drm/drm_stub.c             | 12 ++++++++++-
>>  drivers/gpu/drm/i915/i915_gem.c        |  3 ++-
>>  drivers/gpu/drm/mgag200/mgag200_ttm.c  |  5 +++--
>>  drivers/gpu/drm/nouveau/nouveau_gem.c  |  2 --
>>  drivers/gpu/drm/nouveau/nouveau_ttm.c  |  4 +++-
>>  drivers/gpu/drm/omapdrm/omap_gem.c     | 34 +++++++++++++++----------------
>>  drivers/gpu/drm/qxl/qxl_object.c       |  2 --
>>  drivers/gpu/drm/qxl/qxl_ttm.c          |  6 +++---
>>  drivers/gpu/drm/radeon/radeon_object.c |  1 -
>>  drivers/gpu/drm/radeon/radeon_ttm.c    |  5 +++--
>>  drivers/gpu/drm/ttm/ttm_bo.c           |  3 ++-
>>  drivers/gpu/drm/vmwgfx/vmwgfx_drv.c    |  5 +++--
>>  include/drm/drmP.h                     |  3 ++-
>>  include/drm/drm_vma_manager.h          |  6 +++---
>>  include/drm/ttm/ttm_bo_driver.h        |  2 ++
>>  19 files changed, 100 insertions(+), 65 deletions(-)
>>
>> --
>> 1.8.5.2
>>
> _______________________________________________
> dri-devel mailing list
> dri-devel@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: [RFC v2 0/3] DRM Anonymous Inodes

[Dave Airlie]

>> Hi
>>
>> On Fri, Jan 3, 2014 at 3:41 PM, David Herrmann <dh.herrmann@gmail.com> wrote:
>>> Hi
>>>
>>> With 3.13-rc1 the required VFS core changes for anonymous backing inodes in DRM
>>> got merged. This series reworks the previous patches (I think from early
>>> August '13) and finally replaces the ugly drm_device->dev_mapping hack.
>>>
>>> The patches should be fairly obvious. As I currently only have nouveau here, I'd
>>> be happy to get some more testers with other drivers.
>>>
>>> Patch #1 adds an internal private mount point to DRM core.
>>> Patch #2 replaces ->dev_mapping with a private anonymous inode for each device.
>>> Patch #3 removes the delayed dev_mapping initialization in TTM.
>>>
>>> The internal mnt-point is based on fs/aio.c and fs/anon_inode.c which do the
>>> same. It's not clear to me why we cannot share the mnt-point between all these,
>>> but Al Viro made clear that he will reject any patches on anon_inode.c which do
>>> that. So lets follow his recommendation and use our own internal mnt-point.
>> Also tested on i915 ivb now (before: nouveau+TTM). I think this is
>> ready for 3.14 if nobody finds any issues.
>

I talked to Al,

7:13 < viro> umm...
17:14 < viro> you do realize that your module will be impossible to unload?
17:14 < viro> kern_mount will pin that fs down
17:14 < viro> bumping module refcount
17:15 < viro> so kern_unmount() in module_exit is not going to be called at all
17:15 -!- rafael__ [^rafael@aerd124.neoplus.adsl.tpnet.pl] has quit
[Ping timeout: 608 seconds]
17:16 < viro> what you want is some variation on simple_pin_fs()
17:17 < viro> done from drm_dev_alloc()
17:18 < viro> with simple_release_fs() from drm_dev_free(), right after iput()
17:20 < viro> the export itself is OK (albeit we might be better off
exporting a variant of mount_pseudo() instead, so that aio and
              this sucker would both just call mount_dname_pseudo())
17:21 < viro> the problem I see there is with kern_mount() use
17:47 -!- sameo [~samuel@192.55.55.41] has quit [Remote host closed
the connection]

So maybe we should think about it a bit more :-)

Dave.

Re: [RFC v2 0/3] DRM Anonymous Inodes

[David Herrmann]

Hi

On Mon, Jan 20, 2014 at 3:00 AM, Dave Airlie <airlied@gmail.com> wrote:
>>> Hi
>>>
>>> On Fri, Jan 3, 2014 at 3:41 PM, David Herrmann <dh.herrmann@gmail.com> wrote:
>>>> Hi
>>>>
>>>> With 3.13-rc1 the required VFS core changes for anonymous backing inodes in DRM
>>>> got merged. This series reworks the previous patches (I think from early
>>>> August '13) and finally replaces the ugly drm_device->dev_mapping hack.
>>>>
>>>> The patches should be fairly obvious. As I currently only have nouveau here, I'd
>>>> be happy to get some more testers with other drivers.
>>>>
>>>> Patch #1 adds an internal private mount point to DRM core.
>>>> Patch #2 replaces ->dev_mapping with a private anonymous inode for each device.
>>>> Patch #3 removes the delayed dev_mapping initialization in TTM.
>>>>
>>>> The internal mnt-point is based on fs/aio.c and fs/anon_inode.c which do the
>>>> same. It's not clear to me why we cannot share the mnt-point between all these,
>>>> but Al Viro made clear that he will reject any patches on anon_inode.c which do
>>>> that. So lets follow his recommendation and use our own internal mnt-point.
>>> Also tested on i915 ivb now (before: nouveau+TTM). I think this is
>>> ready for 3.14 if nobody finds any issues.
>>
>
> I talked to Al,
>
> 7:13 < viro> umm...
> 17:14 < viro> you do realize that your module will be impossible to unload?
> 17:14 < viro> kern_mount will pin that fs down
> 17:14 < viro> bumping module refcount
> 17:15 < viro> so kern_unmount() in module_exit is not going to be called at all
> 17:15 -!- rafael__ [^rafael@aerd124.neoplus.adsl.tpnet.pl] has quit
> [Ping timeout: 608 seconds]
> 17:16 < viro> what you want is some variation on simple_pin_fs()
> 17:17 < viro> done from drm_dev_alloc()
> 17:18 < viro> with simple_release_fs() from drm_dev_free(), right after iput()
> 17:20 < viro> the export itself is OK (albeit we might be better off
> exporting a variant of mount_pseudo() instead, so that aio and
>               this sucker would both just call mount_dname_pseudo())
> 17:21 < viro> the problem I see there is with kern_mount() use
> 17:47 -!- sameo [~samuel@192.55.55.41] has quit [Remote host closed
> the connection]
>
> So maybe we should think about it a bit more :-)

So we call simple_pin_fs() in drm_alloc_inode() and add a
drm_free_inode() which calls iput() followed by simple_release_fs()? I
will write up v3 and test with module re-loading this time.

Thanks
David