Anonymous Inode Allocations
Subsystems (12)
cover Anonymous Inode Allocations
Hi
This implements anon_inodes_new() to create anonymous inodes. Patch #1 describes the changes to anon_inodes.c and why DRM could make great use of this. Patch #2 converts DRM core to use anon_inodes_new() instead of delayed dev_mapping initialization (but kept simple, TTM can be converted later).
The idea is to get an anonymous backing inode for DRM devices without waiting for the first char-dev inode. We use it to unmap userspace mappings via unmap_mapping_range() if we have to evict buffers if GTT runs short or if the FW framebuffers are destroyed.
We need the inode only once the first user-space mapping was created, but the delayed initialization causes an ugly code base and keeps inodes of char-devs around just to preserve the address_space.
We actually only need the address_space, but mm core requires an corresponding file* and FS core requires mapping->host to be set (although I am not sure whether we can hit those paths with char-devs). So I went with the whole anonymous inode approach.
If anyone has an idea how to use an embedded "struct address_space" inside of "drm_device" and set "dev_mapping->host" to the shared anon_inode_inode, I'd be happy to implement it. However, I didn't succeed and I am actually not sure that separate "struct address_space" are actually supported. For instance, DRM core uses code like:
container_of(dev_mapping, struct inode, i_data)
So I didn't spent much time on that approach.
Cheers
David
David Herrmann (2):
anon_inodes: allow external inode allocations
DRM: use anon_inode instead of delayed inode init
drivers/gpu/drm/drm_drv.c | 1 - drivers/gpu/drm/drm_fops.c | 24 +++-------------------- drivers/gpu/drm/drm_stub.c | 12 +++++++++++- drivers/gpu/drm/i915/i915_gem.c | 4 ++-- drivers/gpu/drm/nouveau/nouveau_gem.c | 2 +- drivers/gpu/drm/omapdrm/omap_gem.c | 7 ++++--- drivers/gpu/drm/qxl/qxl_object.c | 2 +- drivers/gpu/drm/qxl/qxl_ttm.c | 2 +- drivers/gpu/drm/radeon/radeon_object.c | 2 +- drivers/gpu/drm/radeon/radeon_ttm.c | 2 +- drivers/gpu/drm/vmwgfx/vmwgfx_drv.c | 2 +- fs/anon_inodes.c | 36 +++++++++++++++++++++++++++------- include/drm/drmP.h | 2 +- include/linux/anon_inodes.h | 1 + 14 files changed, 57 insertions(+), 42 deletions(-)
Signature · 1 line
1.8.3.2
PATCH 1/2 anon_inodes: allow external inode allocations
DRM core shares a single address_space across all inodes that belong to the same DRM device. This allows efficient unmapping of user-space mappings during buffer destruction. However, there is no easy way to get a shared address_space for DRM devices during initialization. Therefore, we currently delay this step until the first ->open() and save the given inode for future use.
This causes ugly delayed initialization throughout the DRM core. TTM devices end up without a dev_mapping pointer and we have to carefully respect any underlying filesystem implementation so we don't corrupt the inode->i_mapping and inode->i_data fields.
We can avoid this if we were allowed to allocate an anonymous inode for each DRM device. We only have to set file->f_mapping during ->open() and no longer need to adjust inode mappings. As fs/anon_inodes.c already provides a minimal internal FS mount, we extend it to also provide anonymous inodes for use in drivers like DRM.
--- fs/anon_inodes.c | 36 +++++++++++++++++++++++++++++------- include/linux/anon_inodes.h | 1 + 2 files changed, 30 insertions(+), 7 deletions(-)
fs/anon_inodes.c +29 -7
@@ -25,6 +25,7 @@static struct vfsmount *anon_inode_mnt__read_mostly ;static struct inode *anon_inode_inode;static const struct file_operations anon_inode_fops; +static struct dentry *anon_inode_root;/* * anon_inodefs_dname() is called from d_path().
@@ -87,19 +88,18 @@ static struct inode *anon_inode_mkinode(struct super_block *s)static struct dentry *anon_inodefs_mount (struct file_system_type *fs_type,int flags,const char *dev_name,void *data) { -struct dentry *root; - root =mount_pseudo (fs_type,"anon_inode:" ,NULL , + anon_inode_root =mount_pseudo (fs_type,"anon_inode:" ,NULL , &anon_inodefs_dentry_operations,ANON_INODE_FS_MAGIC ); -if (!IS_ERR (root)) { -struct super_block *s = root->d_sb; +if (!IS_ERR (anon_inode_root)) { +struct super_block *s = anon_inode_root->d_sb; anon_inode_inode =anon_inode_mkinode (s);if (IS_ERR (anon_inode_inode)) { -dput (root); +dput (anon_inode_root);deactivate_locked_super (s); - root =ERR_CAST (anon_inode_inode); + anon_inode_root =ERR_CAST (anon_inode_inode); } } -return root; +return anon_inode_root; }static struct file_system_type anon_inode_fs_type = {
@@ -219,6 +219,28 @@ err_put_unused_fd: }EXPORT_SYMBOL_GPL (anon_inode_getfd); +/** +* anon_inode_new - create private anonymous inode +* +* Creates a new inode on the anonymous inode FS for driver's use. The inode has +* it's own address_space compared to the shared anon_inode_inode. It can be +* used in situations where user-space mappings have to be shared across +* different files but no backing inode is available. +* +* Call iput(inode) to release the inode. +* +* RETURNS: +* New inode on success, error pointer on failure. +*/ +struct inode *anon_inode_new (void ) +{ +if (IS_ERR (anon_inode_root)) +return ERR_CAST (anon_inode_root); + +return anon_inode_mkinode (anon_inode_root->d_sb); +} +EXPORT_SYMBOL_GPL (anon_inode_new); +static int __init anon_inode_init (void ) {int error;
include/linux/anon_inodes.h +1 -0
@@ -15,6 +15,7 @@ struct file *anon_inode_getfile(const char *name, void *priv, int flags);int anon_inode_getfd (const char *name,const struct file_operations *fops,void *priv,int flags); +struct inode *anon_inode_new (void );#endif /* _LINUX_ANON_INODES_H */
Signature · 1 line
1.8.3.2
Re: [PATCH 1/2] anon_inodes: allow external inode allocations
Quoted 6 lines · David Herrmann
DRM core shares a single
address_spaceacross all inodes that belong to the same DRM device. This allows efficient unmapping of user-space mappings during buffer destruction. However, there is no easy way to get a sharedaddress_spacefor DRM devices during initialization. Therefore, we currently delay this step until the first ->open() and save the given inode for future use.
Quick correction for the drm use-case: We don't need the address space at buffer destruction, since each vma holds a reference to the backing gem object. So buffer destruction can only happen once there's guaranteed to be no mmapping around any more.
But we make extensive use of the address_space to shoot down ptes with unmap_mapping_range before we evict a buffer from the mmio gart. In the page fault handler we can then move the buffer object back to a place userspace can get at it and set up new ptes.
Quoted 5 lines
This causes ugly delayed initialization throughout the DRM core. TTM devices end up without a
dev_mappingpointer and we have to carefully respect any underlying filesystem implementation so we don't corrupt theinode->i_mappingandinode->i_datafields.
Quoted 7 lines
We can avoid this if we were allowed to allocate an anonymous inode for each DRM device. We only have to set
file->f_mappingduring ->open() and no longer need to adjust inode mappings. Asfs/anon_inodes.calready provides a minimal internal FS mount, we extend it to also provide anonymous inodes for use in drivers like DRM.Signed-off-by David Herrmann <[email protected]>
I'm not an fs guy so no comment on the patch itself, but having an inode/address space available at driver load time will allow us to get rid of tons of ulgy
if (dev_mapping) unmap_mapping_rang(dev_mapping, ...);
code sprinkled all over drm core and drivers. So
Quoted 87 lines
--- fs/anon_inodes.c | 36 +++++++++++++++++++++++++++++------- include/linux/anon_inodes.h | 1 + 2 files changed, 30 insertions(+), 7 deletions(-)fs/anon_inodes.c +29
-7@@ -25,6 +25,7 @@static struct vfsmount *anon_inode_mnt__read_mostly ;static struct inode *anon_inode_inode;static const struct file_operations anon_inode_fops; +static struct dentry *anon_inode_root;/* * anon_inodefs_dname() is called from d_path(). @@ -87,19 +88,18 @@ static struct inode *anon_inode_mkinode(struct super_block *s)static struct dentry *anon_inodefs_mount (struct file_system_type *fs_type,int flags,const char *dev_name,void *data) { -struct dentry *root; - root =mount_pseudo (fs_type,"anon_inode:" ,NULL , + anon_inode_root =mount_pseudo (fs_type,"anon_inode:" ,NULL , &anon_inodefs_dentry_operations,ANON_INODE_FS_MAGIC ); -if (!IS_ERR (root)) { -struct super_block *s = root->d_sb; +if (!IS_ERR (anon_inode_root)) { +struct super_block *s = anon_inode_root->d_sb; anon_inode_inode =anon_inode_mkinode (s);if (IS_ERR (anon_inode_inode)) { -dput (root); +dput (anon_inode_root);deactivate_locked_super (s); - root =ERR_CAST (anon_inode_inode); + anon_inode_root =ERR_CAST (anon_inode_inode); } } -return root; +return anon_inode_root; }static struct file_system_type anon_inode_fs_type = {@@ -219,6 +219,28 @@ err_put_unused_fd: }EXPORT_SYMBOL_GPL (anon_inode_getfd); +/** +* anon_inode_new - create private anonymous inode +* +* Creates a new inode on the anonymous inode FS for driver's use. The inode has +* it's own address_space compared to the shared anon_inode_inode. It can be +* used in situations where user-space mappings have to be shared across +* different files but no backing inode is available. +* +* Call iput(inode) to release the inode. +* +* RETURNS: +* New inode on success, error pointer on failure. +*/ +struct inode *anon_inode_new (void ) +{ +if (IS_ERR (anon_inode_root)) +return ERR_CAST (anon_inode_root); + +return anon_inode_mkinode (anon_inode_root->d_sb); +} +EXPORT_SYMBOL_GPL (anon_inode_new); +static int __init anon_inode_init (void ) {int error;include/linux/anon_inodes.h +1
-0@@ -15,6 +15,7 @@ struct file *anon_inode_getfile(const char *name, void *priv, int flags);int anon_inode_getfd (const char *name,const struct file_operations *fops,void *priv,int flags); +struct inode *anon_inode_new (void );#endif /* _LINUX_ANON_INODES_H */ Signature · 1 line
1.8.3.2
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [PATCH 1/2] anon_inodes: allow external inode allocations
In reply to David HerrmannHi Al
Any comment on this?
Regards
David
Quoted 106 lines · David Herrmann
DRM core shares a single
address_spaceacross all inodes that belong to the same DRM device. This allows efficient unmapping of user-space mappings during buffer destruction. However, there is no easy way to get a sharedaddress_spacefor DRM devices during initialization. Therefore, we currently delay this step until the first ->open() and save the given inode for future use.This causes ugly delayed initialization throughout the DRM core. TTM devices end up without a
dev_mappingpointer and we have to carefully respect any underlying filesystem implementation so we don't corrupt theinode->i_mappingandinode->i_datafields.We can avoid this if we were allowed to allocate an anonymous inode for each DRM device. We only have to set
file->f_mappingduring ->open() and no longer need to adjust inode mappings. Asfs/anon_inodes.calready provides a minimal internal FS mount, we extend it to also provide anonymous inodes for use in drivers like DRM.Signed-off-by David Herrmann <[email protected]>--- fs/anon_inodes.c | 36 +++++++++++++++++++++++++++++------- include/linux/anon_inodes.h | 1 + 2 files changed, 30 insertions(+), 7 deletions(-)fs/anon_inodes.c +29
-7@@ -25,6 +25,7 @@static struct vfsmount *anon_inode_mnt__read_mostly ;static struct inode *anon_inode_inode;static const struct file_operations anon_inode_fops; +static struct dentry *anon_inode_root;/* * anon_inodefs_dname() is called from d_path(). @@ -87,19 +88,18 @@ static struct inode *anon_inode_mkinode(struct super_block *s)static struct dentry *anon_inodefs_mount (struct file_system_type *fs_type,int flags,const char *dev_name,void *data) { -struct dentry *root; - root =mount_pseudo (fs_type,"anon_inode:" ,NULL , + anon_inode_root =mount_pseudo (fs_type,"anon_inode:" ,NULL , &anon_inodefs_dentry_operations,ANON_INODE_FS_MAGIC ); -if (!IS_ERR (root)) { -struct super_block *s = root->d_sb; +if (!IS_ERR (anon_inode_root)) { +struct super_block *s = anon_inode_root->d_sb; anon_inode_inode =anon_inode_mkinode (s);if (IS_ERR (anon_inode_inode)) { -dput (root); +dput (anon_inode_root);deactivate_locked_super (s); - root =ERR_CAST (anon_inode_inode); + anon_inode_root =ERR_CAST (anon_inode_inode); } } -return root; +return anon_inode_root; }static struct file_system_type anon_inode_fs_type = {@@ -219,6 +219,28 @@ err_put_unused_fd: }EXPORT_SYMBOL_GPL (anon_inode_getfd); +/** +* anon_inode_new - create private anonymous inode +* +* Creates a new inode on the anonymous inode FS for driver's use. The inode has +* it's own address_space compared to the shared anon_inode_inode. It can be +* used in situations where user-space mappings have to be shared across +* different files but no backing inode is available. +* +* Call iput(inode) to release the inode. +* +* RETURNS: +* New inode on success, error pointer on failure. +*/ +struct inode *anon_inode_new (void ) +{ +if (IS_ERR (anon_inode_root)) +return ERR_CAST (anon_inode_root); + +return anon_inode_mkinode (anon_inode_root->d_sb); +} +EXPORT_SYMBOL_GPL (anon_inode_new); +static int __init anon_inode_init (void ) {int error;include/linux/anon_inodes.h +1
-1@@ -15,6 +15,7 @@ struct file *anon_inode_getfile(const char *name, void *priv, int flags);int anon_inode_getfd (const char *name,const struct file_operations *fops,void *priv,int flags); +struct inode *anon_inode_new (void );#endif /* _LINUX_ANON_INODES_H */ --1.8.3.2
PATCH 2/2 DRM: use anon_inode instead of delayed inode init
In reply to David HerrmannInstead of delaying inode initialization until first ->open(), we can use an anonymous inode. This avoids modifying FS internal inode fields and provides us a private address_space right during initialization.
Delayed TTM dev_mapping initialization is currently left untouched to keep this simple. But we could now safely provide the address_space during ttm_bo_device_init() instead of delaying until first buffer ->mmap().
Note that this also fixes several bugs:
- We currently call iput(container_of(..dev_mapping..)) before drm_lastclose(), but we reset dev_mapping to zero at the end of drm_lastclose(). This fails if dev_mapping points to an address_space other than the current inode and the char-dev got already removed.
- We also drop dev_mapping during any drm_lastclose() call. So if user-space still has VMAs to our buffers, we will be unable to unmap them if the next ->firstopen() is on another inode. dev_mapping will then point to a new address_space and we leak mappings that we no longer control.
- We ignore inode->i_mapping completely. It is unlikely that a FS uses it to overwrite inode->i_data for char-devs, but it definitely doesn't look very nice to ignore it silently.
Regarding legacy drivers: We no longer reset the address_space during drm_lastclose() to avoid re-allocating inodes all the time. However, legacy UMS drivers are weird and it is not clear to me whether some of the old drivers might depend on this (for what reason?), but I remember Daniel told me that i810 might.
Tested with nouveau on x86_64.
--- drivers/gpu/drm/drm_drv.c | 1 - drivers/gpu/drm/drm_fops.c | 24 +++--------------------- drivers/gpu/drm/drm_stub.c | 12 +++++++++++- drivers/gpu/drm/i915/i915_gem.c | 4 ++-- drivers/gpu/drm/nouveau/nouveau_gem.c | 2 +- drivers/gpu/drm/omapdrm/omap_gem.c | 7 ++++--- drivers/gpu/drm/qxl/qxl_object.c | 2 +- drivers/gpu/drm/qxl/qxl_ttm.c | 2 +- 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 +- 12 files changed, 27 insertions(+), 35 deletions(-)
drivers/gpu/drm/drm_drv.c +0 -1
@@ -232,7 +232,6 @@ int drm_lastclose(struct drm_device * dev) !drm_core_check_feature(dev, DRIVER_MODESET))drm_dma_takedown (dev); - dev->dev_mapping =NULL ;mutex_unlock (&dev->struct_mutex);DRM_DEBUG ("lastclose completed\n" );
drivers/gpu/drm/drm_fops.c +3 -21
@@ -122,8 +122,6 @@ 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)
@@ -137,16 +135,9 @@ 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); + +/* set address_space for shared mappings */ + filp->f_mapping = dev->anon_inode->i_mapping; retcode =drm_open_helper (inode, filp, dev);if (retcode)
@@ -160,12 +151,6 @@ 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; }
@@ -543,9 +528,6 @@ 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 */ drm_master_put (&file_priv->master); file_priv->is_master =0 ;
drivers/gpu/drm/drm_stub.c +11 -1
@@ -31,6 +31,7 @@* DEALINGS IN THE SOFTWARE. */ +#include <linux/anon_inodes.h> #include <linux/module.h> #include <linux/moduleparam.h> #include <linux/slab.h>
@@ -267,8 +268,14 @@ int drm_fill_in_dev(struct drm_device *dev,mutex_init (&dev->struct_mutex);mutex_init (&dev->ctxlist_mutex); +/* create private address_space on anon inode */ + dev->anon_inode =anon_inode_new (); +if (IS_ERR (dev->anon_inode)) +return PTR_ERR (dev->anon_inode); +if (drm_ht_create (&dev->map_hash,12 )) { -return -ENOMEM ; + retcode = -ENOMEM ; + goto err_inode; }/* the DRM has 6 basic counters */
@@ -309,6 +316,8 @@ int drm_fill_in_dev(struct drm_device *dev, error_out_unreg:drm_lastclose (dev); +err_inode: +iput (dev->anon_inode);return retcode; }EXPORT_SYMBOL (drm_fill_in_dev);
@@ -478,6 +487,7 @@ void drm_put_dev(struct drm_device *dev)drm_put_minor (&dev->primary); +iput (dev->anon_inode);list_del (&dev->driver_item);kfree (dev->devname);kfree (dev);
drivers/gpu/drm/i915/i915_gem.c +2 -2
@@ -1425,8 +1425,8 @@ i915_gem_release_mmap(struct drm_i915_gem_object *obj)if (!obj->fault_mappable)return ; -if (obj->base.dev->dev_mapping) -unmap_mapping_range (obj->base.dev->dev_mapping, +if (obj->base.dev->anon_inode->i_mapping) +unmap_mapping_range (obj->base.dev->anon_inode->i_mapping, (loff_t )obj->base.map_list.hash.key<<PAGE_SHIFT , obj->base.size,1 );
drivers/gpu/drm/nouveau/nouveau_gem.c +1 -1
@@ -243,7 +243,7 @@ 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);
drivers/gpu/drm/omapdrm/omap_gem.c +4 -3
@@ -152,7 +152,7 @@ static struct {static void evict_entry (struct drm_gem_object *obj,enum tiler_fmt fmt,struct usergart_entry *entry) { -if (obj->dev->dev_mapping) { +if (obj->dev->anon_inode->i_mapping) {struct omap_gem_object *omap_obj =to_omap_bo (obj);int n = usergart[fmt].height;size_t size =PAGE_SIZE * n;
@@ -163,12 +163,13 @@ static void evict_entry(struct drm_gem_object *obj,int i;/* if stride > than PAGE_SIZE then sparse mapping: */ for (i = n; i >0 ; i--) { -unmap_mapping_range (obj->dev->dev_mapping, +unmap_mapping_range (obj->dev->anon_inode->i_mapping, off,PAGE_SIZE ,1 ); off +=PAGE_SIZE * m; } }else { -unmap_mapping_range (obj->dev->dev_mapping, off, size,1 ); +unmap_mapping_range (obj->dev->anon_inode->i_mapping, + off, size,1 ); } }
drivers/gpu/drm/qxl/qxl_object.c +1 -1
@@ -82,7 +82,7 @@ int qxl_bo_create(struct qxl_device *qdev,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
drivers/gpu/drm/qxl/qxl_ttm.c +1 -1
@@ -515,7 +515,7 @@ int qxl_ttm_init(struct qxl_device *qdev)DRM_INFO ("qxl: %luM of IO pages memory ready (VRAM domain)\n" , ((unsigned )num_io_pages *PAGE_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" );
drivers/gpu/drm/radeon/radeon_object.c +1 -1
@@ -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) {
drivers/gpu/drm/radeon/radeon_ttm.c +1 -1
@@ -735,7 +735,7 @@ 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) {
drivers/gpu/drm/vmwgfx/vmwgfx_drv.c +1 -1
@@ -760,7 +760,7 @@ 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 ;
include/drm/drmP.h +1 -1
@@ -1198,7 +1198,7 @@ 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;
Signature · 1 line
1.8.3.2
Re: [PATCH 2/2] DRM: use anon_inode instead of delayed inode init
Quoted 8 lines · David Herrmann
Instead of delaying inode initialization until first ->open(), we can use an anonymous inode. This avoids modifying FS internal inode fields and provides us a private
address_spaceright during initialization.Delayed TTM
dev_mappinginitialization is currently left untouched to keep this simple. But we could now safely provide theaddress_spaceduringttm_bo_device_init()instead of delaying until first buffer ->mmap().
Quoted 10 lines
Note that this also fixes several bugs:
- We currently call iput(container_of(..dev_mapping..)) beforedrm_lastclose(), but we resetdev_mappingto zero at the end ofdrm_lastclose(). This fails ifdev_mappingpoints to anaddress_spaceother than the current inode and the char-dev got already removed.
- We also dropdev_mappingduring anydrm_lastclose()call. So if user-space still has VMAs to our buffers, we will be unable to unmap them if the next ->firstopen() is on another inode.dev_mappingwill then point to a newaddress_spaceand we leak mappings that we no longer control.
It's certainly ugly, but I don't think we have a real problem here. vma grabs a reference to the open file at mmap time and we grab a reference to the underlying gem object. So it shouldn't be possible to observe a non-NULL dev_mapping while the inode refcount has already reached 0 anywhere we actually care. At least in drm/i915 since we never call unmap_mapping_range if we know that there's no ptes around pointing at this specific object (which we accurately keep track of in our fault handler).
TTM might be different, and it's certainly good to rid us of this code.
Quoted 3 lines
- We ignore
inode->i_mappingcompletely. It is unlikely that a FS uses it to overwriteinode->i_datafor char-devs, but it definitely doesn't look very nice to ignore it silently.
Tbh I have no idea what the rules are here ... But since core vfs code uses the filp->f_mapping at mmap time not frobbing inode->i_mapping looks like the sane to do.
Quoted 5 lines
Regarding legacy drivers: We no longer reset the
address_spaceduringdrm_lastclose()to avoid re-allocating inodes all the time. However, legacy UMS drivers are weird and it is not clear to me whether some of the old drivers might depend on this (for what reason?), but I remember Daniel told me that i810 might.
i915 in gem+ums mode might. i810 is a different level of horror show entirely, but since it doesn't do gem we can ignore it here.
Quoted 4 lines
Tested with nouveau on
x86_64.Signed-off-by David Herrmann <[email protected]>
I guess it makes sense to merge that after your drm vma offset manager changes with the little pte shootdown helper since that'll reduce the diff?
-Daniel
Quoted 266 lines
--- drivers/gpu/drm/drm_drv.c | 1 - drivers/gpu/drm/drm_fops.c | 24 +++--------------------- drivers/gpu/drm/drm_stub.c | 12 +++++++++++- drivers/gpu/drm/i915/i915_gem.c | 4 ++-- drivers/gpu/drm/nouveau/nouveau_gem.c | 2 +- drivers/gpu/drm/omapdrm/omap_gem.c | 7 ++++--- drivers/gpu/drm/qxl/qxl_object.c | 2 +- drivers/gpu/drm/qxl/qxl_ttm.c | 2 +- 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 +- 12 files changed, 27 insertions(+), 35 deletions(-)drivers/gpu/drm/drm_drv.c +0
-1@@ -232,7 +232,6 @@ int drm_lastclose(struct drm_device * dev) !drm_core_check_feature(dev, DRIVER_MODESET))drm_dma_takedown (dev); - dev->dev_mapping =NULL ;mutex_unlock (&dev->struct_mutex);DRM_DEBUG ("lastclose completed\n" );drivers/gpu/drm/drm_fops.c +3
-21@@ -122,8 +122,6 @@ 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)@@ -137,16 +135,9 @@ 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); + +/* set address_space for shared mappings */ + filp->f_mapping = dev->anon_inode->i_mapping; retcode =drm_open_helper (inode, filp, dev);if (retcode)@@ -160,12 +151,6 @@ 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; }@@ -543,9 +528,6 @@ 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 */ drm_master_put (&file_priv->master); file_priv->is_master =0 ;drivers/gpu/drm/drm_stub.c +11
-1@@ -31,6 +31,7 @@* DEALINGS IN THE SOFTWARE. */ +#include <linux/anon_inodes.h> #include <linux/module.h> #include <linux/moduleparam.h> #include <linux/slab.h> @@ -267,8 +268,14 @@ int drm_fill_in_dev(struct drm_device *dev,mutex_init (&dev->struct_mutex);mutex_init (&dev->ctxlist_mutex); +/* create private address_space on anon inode */ + dev->anon_inode =anon_inode_new (); +if (IS_ERR (dev->anon_inode)) +return PTR_ERR (dev->anon_inode); +if (drm_ht_create (&dev->map_hash,12 )) { -return -ENOMEM ; + retcode = -ENOMEM ; + goto err_inode; }/* the DRM has 6 basic counters */ @@ -309,6 +316,8 @@ int drm_fill_in_dev(struct drm_device *dev, error_out_unreg:drm_lastclose (dev); +err_inode: +iput (dev->anon_inode);return retcode; }EXPORT_SYMBOL (drm_fill_in_dev);@@ -478,6 +487,7 @@ void drm_put_dev(struct drm_device *dev)drm_put_minor (&dev->primary); +iput (dev->anon_inode);list_del (&dev->driver_item);kfree (dev->devname);kfree (dev);drivers/gpu/drm/i915/i915_gem.c +2
-2@@ -1425,8 +1425,8 @@ i915_gem_release_mmap(struct drm_i915_gem_object *obj)if (!obj->fault_mappable)return ; -if (obj->base.dev->dev_mapping) -unmap_mapping_range (obj->base.dev->dev_mapping, +if (obj->base.dev->anon_inode->i_mapping) +unmap_mapping_range (obj->base.dev->anon_inode->i_mapping, (loff_t )obj->base.map_list.hash.key<<PAGE_SHIFT , obj->base.size,1 );drivers/gpu/drm/nouveau/nouveau_gem.c +1
-1@@ -243,7 +243,7 @@ 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);drivers/gpu/drm/omapdrm/omap_gem.c +4
-3@@ -152,7 +152,7 @@ static struct {static void evict_entry (struct drm_gem_object *obj,enum tiler_fmt fmt,struct usergart_entry *entry) { -if (obj->dev->dev_mapping) { +if (obj->dev->anon_inode->i_mapping) {struct omap_gem_object *omap_obj =to_omap_bo (obj);int n = usergart[fmt].height;size_t size =PAGE_SIZE * n;@@ -163,12 +163,13 @@ static void evict_entry(struct drm_gem_object *obj,int i;/* if stride > than PAGE_SIZE then sparse mapping: */ for (i = n; i >0 ; i--) { -unmap_mapping_range (obj->dev->dev_mapping, +unmap_mapping_range (obj->dev->anon_inode->i_mapping, off,PAGE_SIZE ,1 ); off +=PAGE_SIZE * m; } }else { -unmap_mapping_range (obj->dev->dev_mapping, off, size,1 ); +unmap_mapping_range (obj->dev->anon_inode->i_mapping, + off, size,1 ); } }drivers/gpu/drm/qxl/qxl_object.c +1
-1@@ -82,7 +82,7 @@ int qxl_bo_create(struct qxl_device *qdev,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 drivers/gpu/drm/qxl/qxl_ttm.c +1
-1@@ -515,7 +515,7 @@ int qxl_ttm_init(struct qxl_device *qdev)DRM_INFO ("qxl: %luM of IO pages memory ready (VRAM domain)\n" , ((unsigned )num_io_pages *PAGE_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" );drivers/gpu/drm/radeon/radeon_object.c +1
-1@@ -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) {drivers/gpu/drm/radeon/radeon_ttm.c +1
-1@@ -735,7 +735,7 @@ 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) {drivers/gpu/drm/vmwgfx/vmwgfx_drv.c +1
-1@@ -760,7 +760,7 @@ 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 ;include/drm/drmP.h +1
-1@@ -1198,7 +1198,7 @@ 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;Signature · 1 line
1.8.3.2
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
v2 cover Anonymous Inode Allocations
In reply to David HerrmannHi
This is v2 of my anon_inodes allocation series. I would really like some comments on #1 (and maybe #2) from VFS maintainers/developers so we can continue with this work. Patches 1 and 2 are unchanged, I added 3 and 4 to clean up DRM delayed initializations based on the new anon_inode.
This implements anon_inodes_new() to create anonymous inodes. Patch #1 describes the changes to anon_inodes.c and why DRM could make great use of this. Patch #2 converts DRM core to use anon_inodes_new() instead of delayed dev_mapping initialization. Patches #3 and #4 simplify the various driver bits to drop unused code paths.
The idea is to get an anonymous backing inode for DRM devices without waiting for the first char-dev inode. We use it to unmap userspace mappings via unmap_mapping_range() if we have to evict buffers if GTT runs short or if the FW framebuffers are destroyed.
We need the inode only once the first user-space mapping was created. But the delayed initialization causes an ugly code base and keeps inodes of char-devs around just to preserve the address_space.
We actually only need the address_space, but mm core requires a corresponding file* and FS core requires mapping->host to be set (although I am not sure whether we can hit those paths with char-devs). So I went with the whole anonymous inode approach.
If anyone has an idea how to use an embedded "struct address_space" inside of "drm_device" and set "dev_mapping->host" to the shared anon_inode_inode, I'd be happy to implement it. However, I didn't succeed and I am actually not sure that separate "struct address_space" are actually supported. For instance, DRM core uses code like:
container_of(dev_mapping, struct inode, i_data)
which would break in this case.
Cheers
David
David Herrmann (4):
anon_inodes: allow external inode allocations
DRM: use anon_inode instead of delayed inode init
drm: init TTM dev_mapping in ttm_bo_device_init() drm/omap: remove useless if() in evict_entry()
drivers/gpu/drm/ast/ast_ttm.c | 4 +++- drivers/gpu/drm/cirrus/cirrus_ttm.c | 4 +++- drivers/gpu/drm/drm_drv.c | 1 - drivers/gpu/drm/drm_fops.c | 24 +++-------------------- drivers/gpu/drm/drm_stub.c | 12 +++++++++++- drivers/gpu/drm/i915/i915_gem.c | 3 ++- drivers/gpu/drm/mgag200/mgag200_ttm.c | 4 +++- drivers/gpu/drm/nouveau/nouveau_gem.c | 2 -- drivers/gpu/drm/nouveau/nouveau_ttm.c | 4 +++- drivers/gpu/drm/omapdrm/omap_gem.c | 33 +++++++++++++++---------------- 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 +++-- fs/anon_inodes.c | 36 +++++++++++++++++++++++++++------- include/drm/drmP.h | 2 +- include/drm/drm_vma_manager.h | 6 +++--- include/drm/ttm/ttm_bo_driver.h | 2 ++ include/linux/anon_inodes.h | 1 + 21 files changed, 91 insertions(+), 69 deletions(-)
Signature · 1 line
1.8.3.4
PATCH v2 1/4 anon_inodes: allow external inode allocations
DRM core shares a single address_space across all inodes that belong to the same DRM device. This allows efficient unmapping of user-space mappings during buffer eviction. However, there is no easy way to get a shared address_space for DRM devices during initialization. Therefore, we currently delay this step until the first ->open() and save the given inode for future use.
This causes ugly delayed initialization throughout the DRM core. TTM devices end up without a dev_mapping pointer and we have to carefully respect any underlying filesystem implementation so we don't corrupt the inode->i_mapping and inode->i_data fields.
We can avoid this if we were allowed to allocate an anonymous inode for each DRM device. We only have to set file->f_mapping during ->open() and no longer need to adjust inode mappings. As fs/anon_inodes.c already provides a minimal internal FS mount, we extend it to also provide anonymous inodes for use in drivers like DRM.
--- fs/anon_inodes.c | 36 +++++++++++++++++++++++++++++------- include/linux/anon_inodes.h | 1 + 2 files changed, 30 insertions(+), 7 deletions(-)
fs/anon_inodes.c +29 -7
@@ -25,6 +25,7 @@static struct vfsmount *anon_inode_mnt__read_mostly ;static struct inode *anon_inode_inode;static const struct file_operations anon_inode_fops; +static struct dentry *anon_inode_root;/* * anon_inodefs_dname() is called from d_path().
@@ -87,19 +88,18 @@ static struct inode *anon_inode_mkinode(struct super_block *s)static struct dentry *anon_inodefs_mount (struct file_system_type *fs_type,int flags,const char *dev_name,void *data) { -struct dentry *root; - root =mount_pseudo (fs_type,"anon_inode:" ,NULL , + anon_inode_root =mount_pseudo (fs_type,"anon_inode:" ,NULL , &anon_inodefs_dentry_operations,ANON_INODE_FS_MAGIC ); -if (!IS_ERR (root)) { -struct super_block *s = root->d_sb; +if (!IS_ERR (anon_inode_root)) { +struct super_block *s = anon_inode_root->d_sb; anon_inode_inode =anon_inode_mkinode (s);if (IS_ERR (anon_inode_inode)) { -dput (root); +dput (anon_inode_root);deactivate_locked_super (s); - root =ERR_CAST (anon_inode_inode); + anon_inode_root =ERR_CAST (anon_inode_inode); } } -return root; +return anon_inode_root; }static struct file_system_type anon_inode_fs_type = {
@@ -219,6 +219,28 @@ err_put_unused_fd: }EXPORT_SYMBOL_GPL (anon_inode_getfd); +/** +* anon_inode_new - create private anonymous inode +* +* Creates a new inode on the anonymous inode FS for driver's use. The inode has +* it's own address_space compared to the shared anon_inode_inode. It can be +* used in situations where user-space mappings have to be shared across +* different files but no backing inode is available. +* +* Call iput(inode) to release the inode. +* +* RETURNS: +* New inode on success, error pointer on failure. +*/ +struct inode *anon_inode_new (void ) +{ +if (IS_ERR (anon_inode_root)) +return ERR_CAST (anon_inode_root); + +return anon_inode_mkinode (anon_inode_root->d_sb); +} +EXPORT_SYMBOL_GPL (anon_inode_new); +static int __init anon_inode_init (void ) {int error;
include/linux/anon_inodes.h +1 -0
@@ -15,6 +15,7 @@ struct file *anon_inode_getfile(const char *name, void *priv, int flags);int anon_inode_getfd (const char *name,const struct file_operations *fops,void *priv,int flags); +struct inode *anon_inode_new (void );#endif /* _LINUX_ANON_INODES_H */
Signature · 1 line
1.8.3.4
PATCH v2 2/4 DRM: use anon_inode instead of delayed inode init
In reply to David HerrmannInstead of delaying inode initialization until first ->open(), we can use an anonymous inode. This avoids modifying FS internal inode fields and provides us a private address_space right during initialization.
Delayed TTM dev_mapping initialization is currently left untouched to keep this simple. But we could now safely provide the address_space during ttm_bo_device_init() instead of delaying until first buffer ->mmap().
Note that this also fixes several bugs:
- We currently call iput(container_of(..dev_mapping..)) before drm_lastclose(), but we reset dev_mapping to zero at the end of drm_lastclose(). This fails if dev_mapping points to an address_space other than the current inode and the char-dev got already removed.
- We also drop dev_mapping during any drm_lastclose() call. So if user-space still has VMAs to our buffers, we will be unable to unmap them if the next ->firstopen() is on another inode. dev_mapping will then point to a new address_space and we leak mappings that we no longer control.
- We ignore inode->i_mapping completely. It is unlikely that a FS uses it to overwrite inode->i_data for char-devs, but it definitely doesn't look very nice to ignore it silently.
Regarding legacy drivers: We no longer reset the address_space during drm_lastclose() to avoid re-allocating inodes all the time. However, legacy UMS drivers are weird and it is not clear to me whether some of the old drivers might depend on this (for what reason?), but I remember Daniel told me that i810 might.
Tested with nouveau on x86_64.
--- drivers/gpu/drm/drm_drv.c | 1 - drivers/gpu/drm/drm_fops.c | 24 +++--------------------- drivers/gpu/drm/drm_stub.c | 12 +++++++++++- drivers/gpu/drm/i915/i915_gem.c | 3 ++- drivers/gpu/drm/nouveau/nouveau_gem.c | 2 +- drivers/gpu/drm/omapdrm/omap_gem.c | 7 ++++--- drivers/gpu/drm/qxl/qxl_object.c | 2 +- drivers/gpu/drm/qxl/qxl_ttm.c | 2 +- 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 +- 12 files changed, 27 insertions(+), 34 deletions(-)
drivers/gpu/drm/drm_drv.c +0 -1
@@ -213,7 +213,6 @@ int drm_lastclose(struct drm_device * dev) !drm_core_check_feature(dev, DRIVER_MODESET))drm_dma_takedown (dev); - dev->dev_mapping =NULL ;mutex_unlock (&dev->struct_mutex);DRM_DEBUG ("lastclose completed\n" );
drivers/gpu/drm/drm_fops.c +3 -21
@@ -112,8 +112,6 @@ 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)
@@ -127,16 +125,9 @@ 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); + +/* set address_space for shared mappings */ + filp->f_mapping = dev->anon_inode->i_mapping; retcode =drm_open_helper (inode, filp, dev);if (retcode)
@@ -150,12 +141,6 @@ 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; }
@@ -533,9 +518,6 @@ 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 */ drm_master_put (&file_priv->master); file_priv->is_master =0 ;
drivers/gpu/drm/drm_stub.c +11 -1
@@ -31,6 +31,7 @@* DEALINGS IN THE SOFTWARE. */ +#include <linux/anon_inodes.h> #include <linux/module.h> #include <linux/moduleparam.h> #include <linux/slab.h>
@@ -267,8 +268,14 @@ int drm_fill_in_dev(struct drm_device *dev,mutex_init (&dev->struct_mutex);mutex_init (&dev->ctxlist_mutex); +/* create private address_space on anon inode */ + dev->anon_inode =anon_inode_new (); +if (IS_ERR (dev->anon_inode)) +return PTR_ERR (dev->anon_inode); +if (drm_ht_create (&dev->map_hash,12 )) { -return -ENOMEM ; + retcode = -ENOMEM ; + goto err_inode; }/* the DRM has 6 basic counters */
@@ -309,6 +316,8 @@ int drm_fill_in_dev(struct drm_device *dev, error_out_unreg:drm_lastclose (dev); +err_inode: +iput (dev->anon_inode);return retcode; }EXPORT_SYMBOL (drm_fill_in_dev);
@@ -473,6 +482,7 @@ void drm_put_dev(struct drm_device *dev)drm_put_minor (&dev->primary); +iput (dev->anon_inode);list_del (&dev->driver_item);kfree (dev->devname);kfree (dev);
drivers/gpu/drm/i915/i915_gem.c +2 -1
@@ -1420,7 +1420,8 @@ 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; }
drivers/gpu/drm/nouveau/nouveau_gem.c +1 -1
@@ -237,7 +237,7 @@ 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);
drivers/gpu/drm/omapdrm/omap_gem.c +4 -3
@@ -153,7 +153,7 @@ static struct {static void evict_entry (struct drm_gem_object *obj,enum tiler_fmt fmt,struct usergart_entry *entry) { -if (obj->dev->dev_mapping) { +if (obj->dev->anon_inode->i_mapping) {struct omap_gem_object *omap_obj =to_omap_bo (obj);int n = usergart[fmt].height;size_t size =PAGE_SIZE * n;
@@ -164,12 +164,13 @@ static void evict_entry(struct drm_gem_object *obj,int i;/* if stride > than PAGE_SIZE then sparse mapping: */ for (i = n; i >0 ; i--) { -unmap_mapping_range (obj->dev->dev_mapping, +unmap_mapping_range (obj->dev->anon_inode->i_mapping, off,PAGE_SIZE ,1 ); off +=PAGE_SIZE * m; } }else { -unmap_mapping_range (obj->dev->dev_mapping, off, size,1 ); +unmap_mapping_range (obj->dev->anon_inode->i_mapping, + off, size,1 ); } }
drivers/gpu/drm/qxl/qxl_object.c +1 -1
@@ -82,7 +82,7 @@ int qxl_bo_create(struct qxl_device *qdev,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
drivers/gpu/drm/qxl/qxl_ttm.c +1 -1
@@ -515,7 +515,7 @@ int qxl_ttm_init(struct qxl_device *qdev)DRM_INFO ("qxl: %luM of IO pages memory ready (VRAM domain)\n" , ((unsigned )num_io_pages *PAGE_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" );
drivers/gpu/drm/radeon/radeon_object.c +1 -1
@@ -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) {
drivers/gpu/drm/radeon/radeon_ttm.c +1 -1
@@ -735,7 +735,7 @@ 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) {
drivers/gpu/drm/vmwgfx/vmwgfx_drv.c +1 -1
@@ -760,7 +760,7 @@ 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 ;
include/drm/drmP.h +1 -1
@@ -1187,7 +1187,7 @@ 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;
Signature · 1 line
1.8.3.4
PATCH v2 3/4 drm: init TTM dev_mapping in ttm_bo_device_init()
In reply to David HerrmannWith 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.
--- drivers/gpu/drm/ast/ast_ttm.c | 4 +++- drivers/gpu/drm/cirrus/cirrus_ttm.c | 4 +++- drivers/gpu/drm/mgag200/mgag200_ttm.c | 4 +++- drivers/gpu/drm/nouveau/nouveau_gem.c | 2 -- drivers/gpu/drm/nouveau/nouveau_ttm.c | 4 +++- 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/drm_vma_manager.h | 6 +++--- include/drm/ttm/ttm_bo_driver.h | 2 ++ 13 files changed, 28 insertions(+), 20 deletions(-)
drivers/gpu/drm/ast/ast_ttm.c +3 -1
@@ -257,7 +257,9 @@ 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);
drivers/gpu/drm/cirrus/cirrus_ttm.c +3 -1
@@ -257,7 +257,9 @@ 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);
drivers/gpu/drm/mgag200/mgag200_ttm.c +3 -1
@@ -257,7 +257,9 @@ 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);
drivers/gpu/drm/nouveau/nouveau_gem.c +0 -2
@@ -237,8 +237,6 @@ 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 ;
drivers/gpu/drm/nouveau/nouveau_ttm.c +3 -1
@@ -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);
drivers/gpu/drm/qxl/qxl_object.c +0 -2
@@ -81,8 +81,6 @@ 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->anon_inode->i_mapping;if (kernel) type = ttm_bo_type_kernel;else
drivers/gpu/drm/qxl/qxl_ttm.c +3 -3
@@ -491,7 +491,9 @@ 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;
@@ -514,8 +516,6 @@ int qxl_ttm_init(struct qxl_device *qdev) (unsigned)qdev->vram_size / (1024 * 1024));DRM_INFO ("qxl: %luM of IO pages memory ready (VRAM domain)\n" , ((unsigned )num_io_pages *PAGE_SIZE ) / (1024 *1024 )); -if (unlikely (qdev->mman.bdev.dev_mapping ==NULL )) - 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" );
drivers/gpu/drm/radeon/radeon_object.c +0 -1
@@ -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) {
drivers/gpu/drm/radeon/radeon_ttm.c +3 -2
@@ -697,7 +697,9 @@ 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);
@@ -735,7 +737,6 @@ 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) {
drivers/gpu/drm/ttm/ttm_bo.c +2 -1
@@ -1428,6 +1428,7 @@ 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) {
@@ -1449,7 +1450,7 @@ 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 ;
drivers/gpu/drm/vmwgfx/vmwgfx_drv.c +3 -2
@@ -543,7 +543,9 @@ 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" );
@@ -760,7 +762,6 @@ 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 ;
include/drm/drm_vma_manager.h +3 -3
@@ -206,8 +206,8 @@ 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.
@@ -215,7 +215,7 @@ 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 );
include/drm/ttm/ttm_bo_driver.h +2 -0
@@ -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);/**
Signature · 1 line
1.8.3.4
PATCH v2 4/4 drm/omap: remove useless if() in evict_entry()
In reply to David Herrmannanon_inode->i_mapping is always set. This was needed before we introduced anon_inodes. We can safely remove it now.
--- drivers/gpu/drm/omapdrm/omap_gem.c | 32 +++++++++++++++----------------- 1 file changed, 15 insertions(+), 17 deletions(-)
drivers/gpu/drm/omapdrm/omap_gem.c +15 -17
@@ -153,25 +153,23 @@ static struct {static void evict_entry (struct drm_gem_object *obj,enum tiler_fmt fmt,struct usergart_entry *entry) { -if (obj->dev->anon_inode->i_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->anon_inode->i_mapping, - off,PAGE_SIZE ,1 ); - off +=PAGE_SIZE * m; - } - }else { +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, size,1 ); + off,PAGE_SIZE ,1 ); + off +=PAGE_SIZE * m; } + }else { +unmap_mapping_range (obj->dev->anon_inode->i_mapping, + off, size,1 ); } entry->obj =NULL ;
Signature · 1 line
1.8.3.4