drm/cma-helper: Move mmap to object functions
Subsystems (4)
Thomas Zimmermann
maintainercover drm/cma-helper: Move mmap to object functions
This patchset moves CMA's mmap into a GEM object function. This change allows for removing CMA's mmap and gem_prime_mmap callbacks in favor of the default implementation.
Tested with vc4 on RPi3.
Thomas Zimmermann (2):
drm/cma-helper: Remove prime infix from GEM object functions drm/cma-helper: Implement mmap as GEM CMA object functions
drivers/gpu/drm/drm_file.c | 3 +- drivers/gpu/drm/drm_gem_cma_helper.c | 141 +++++++++------------------ drivers/gpu/drm/pl111/pl111_drv.c | 2 +- drivers/gpu/drm/vc4/vc4_bo.c | 6 +- include/drm/drm_gem_cma_helper.h | 14 +-- 5 files changed, 58 insertions(+), 108 deletions(-)
--
2.29.2
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Thomas Zimmermann
maintainerPATCH 1/2 drm/cma-helper: Remove prime infix from GEM object functions
These functions are not directly related to PRIME interfaces any longer, but are now GEM object functions. Rename them accordingly and fix the docs.
--- drivers/gpu/drm/drm_gem_cma_helper.c | 20 ++++++++++---------- drivers/gpu/drm/vc4/vc4_bo.c | 4 ++-- include/drm/drm_gem_cma_helper.h | 4 ++-- 3 files changed, 14 insertions(+), 14 deletions(-)
drivers/gpu/drm/drm_gem_cma_helper.c +10 -10
@@ -36,8 +36,8 @@static const struct drm_gem_object_funcs drm_gem_cma_default_funcs = { .free = drm_gem_cma_free_object, .print_info = drm_gem_cma_print_info, - .get_sg_table = drm_gem_cma_prime_get_sg_table, - .vmap = drm_gem_cma_prime_vmap, + .get_sg_table = drm_gem_cma_get_sg_table, + .vmap = drm_gem_cma_vmap, .vm_ops = &drm_gem_cma_vm_ops, };
@@ -424,18 +424,18 @@ void drm_gem_cma_print_info(struct drm_printer *p, unsigned int indent,EXPORT_SYMBOL (drm_gem_cma_print_info);/** -* drm_gem_cma_prime_get_sg_table - provide a scatter/gather table of pinned +* drm_gem_cma_get_sg_table - provide a scatter/gather table of pinned * pages for a CMA GEM object * @obj: GEM object * -* This function exports a scatter/gather table suitable for PRIME usage by +* This function exports a scatter/gather table by * calling the standard DMA mapping API. Drivers using the CMA helpers should * set this as their &drm_gem_object_funcs.get_sg_table callback. * * Returns: * A pointer to the scatter/gather table of pinned pages or NULL on failure. */ -struct sg_table *drm_gem_cma_prime_get_sg_table (struct drm_gem_object *obj) +struct sg_table *drm_gem_cma_get_sg_table (struct drm_gem_object *obj) {struct drm_gem_cma_object *cma_obj =to_drm_gem_cma_obj (obj);struct sg_table *sgt;
@@ -456,7 +456,7 @@ struct sg_table *drm_gem_cma_prime_get_sg_table(struct drm_gem_object *obj)kfree (sgt);return ERR_PTR (ret); } -EXPORT_SYMBOL_GPL (drm_gem_cma_prime_get_sg_table); +EXPORT_SYMBOL_GPL (drm_gem_cma_get_sg_table);/** * drm_gem_cma_prime_import_sg_table - produce a CMA GEM object from another
@@ -528,13 +528,13 @@ int drm_gem_cma_prime_mmap(struct drm_gem_object *obj,EXPORT_SYMBOL_GPL (drm_gem_cma_prime_mmap);/** -* drm_gem_cma_prime_vmap - map a CMA GEM object into the kernel's virtual +* drm_gem_cma_vmap - map a CMA GEM object into the kernel's virtual * address space * @obj: GEM object * @map: Returns the kernel virtual address of the CMA GEM object's backing * store. * -* This function maps a buffer exported via DRM PRIME into the kernel's +* This function maps a buffer into the kernel's * virtual address space. Since the CMA buffers are already mapped into the * kernel virtual address space this simply returns the cached virtual * address. Drivers using the CMA helpers should set this as their DRM
@@ -543,7 +543,7 @@ EXPORT_SYMBOL_GPL(drm_gem_cma_prime_mmap);* Returns: * 0 on success, or a negative error code otherwise. */ -int drm_gem_cma_prime_vmap (struct drm_gem_object *obj,struct dma_buf_map *map) +int drm_gem_cma_vmap (struct drm_gem_object *obj,struct dma_buf_map *map) {struct drm_gem_cma_object *cma_obj =to_drm_gem_cma_obj (obj);
@@ -551,7 +551,7 @@ int drm_gem_cma_prime_vmap(struct drm_gem_object *obj, struct dma_buf_map *map)return 0 ; } -EXPORT_SYMBOL_GPL (drm_gem_cma_prime_vmap); +EXPORT_SYMBOL_GPL (drm_gem_cma_vmap);/** * drm_gem_cma_prime_import_sg_table_vmap - PRIME import another driver's
drivers/gpu/drm/vc4/vc4_bo.c +2 -2
@@ -385,7 +385,7 @@ static const struct vm_operations_struct vc4_vm_ops = {static const struct drm_gem_object_funcs vc4_gem_object_funcs = { .free = vc4_free_object, .export = vc4_prime_export, - .get_sg_table = drm_gem_cma_prime_get_sg_table, + .get_sg_table = drm_gem_cma_get_sg_table, .vmap = vc4_prime_vmap, .vm_ops = &vc4_vm_ops, };
@@ -794,7 +794,7 @@ int vc4_prime_vmap(struct drm_gem_object *obj, struct dma_buf_map *map)return -EINVAL ; } -return drm_gem_cma_prime_vmap (obj, map); +return drm_gem_cma_vmap (obj, map); } struct drm_gem_object *
include/drm/drm_gem_cma_helper.h +2 -2
@@ -96,14 +96,14 @@ unsigned long drm_gem_cma_get_unmapped_area(struct file *filp,void drm_gem_cma_print_info (struct drm_printer *p,unsigned int indent,const struct drm_gem_object *obj); -struct sg_table *drm_gem_cma_prime_get_sg_table (struct drm_gem_object *obj); +struct sg_table *drm_gem_cma_get_sg_table (struct drm_gem_object *obj);struct drm_gem_object *drm_gem_cma_prime_import_sg_table (struct drm_device *dev,struct dma_buf_attachment *attach,struct sg_table *sgt);int drm_gem_cma_prime_mmap (struct drm_gem_object *obj,struct vm_area_struct *vma); -int drm_gem_cma_prime_vmap (struct drm_gem_object *obj,struct dma_buf_map *map); +int drm_gem_cma_vmap (struct drm_gem_object *obj,struct dma_buf_map *map);/** * DRM_GEM_CMA_DRIVER_OPS_WITH_DUMB_CREATE - CMA GEM driver operations
Signature · 6 lines
2.29.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Thomas Zimmermann
maintainerPATCH 2/2 drm/cma-helper: Implement mmap as GEM CMA object functions
In reply to Thomas ZimmermannThe new GEM object function drm_gem_cma_mmap() sets the VMA flags and offset as in the old implementation and immediately maps in the buffer's memory pages.
Changing CMA helpers to use the GEM object function allows for the removal of the special implementations for mmap and gem_prime_mmap callbacks. The regular functions drm_gem_mmap() and drm_gem_prime_mmap() are now used.
--- drivers/gpu/drm/drm_file.c | 3 +- drivers/gpu/drm/drm_gem_cma_helper.c | 121 +++++++++------------------ drivers/gpu/drm/pl111/pl111_drv.c | 2 +- drivers/gpu/drm/vc4/vc4_bo.c | 2 +- include/drm/drm_gem_cma_helper.h | 10 +-- 5 files changed, 44 insertions(+), 94 deletions(-)
drivers/gpu/drm/drm_file.c +1 -2
@@ -113,8 +113,7 @@ bool drm_dev_needs_global_mutex(struct drm_device *dev)* The memory mapping implementation will vary depending on how the driver * manages memory. Legacy drivers will use the deprecated drm_legacy_mmap() * function, modern drivers should use one of the provided memory-manager -* specific implementations. For GEM-based drivers this is drm_gem_mmap(), and -* for drivers which use the CMA GEM helpers it's drm_gem_cma_mmap(). +* specific implementations. For GEM-based drivers this is drm_gem_mmap(). * * No other file operations are supported by the DRM userspace API. Overall the * following is an example &file_operations structure::
drivers/gpu/drm/drm_gem_cma_helper.c +38 -83
145 lines not on this page · raw message
drivers/gpu/drm/pl111/pl111_drv.c +1 -1
@@ -228,7 +228,7 @@ static const struct drm_driver pl111_drm_driver = { .prime_handle_to_fd = drm_gem_prime_handle_to_fd, .prime_fd_to_handle = drm_gem_prime_fd_to_handle, .gem_prime_import_sg_table = pl111_gem_import_sg_table, - .gem_prime_mmap = drm_gem_cma_prime_mmap, + .gem_prime_mmap = drm_gem_prime_mmap,#if defined(CONFIG_DEBUG_FS ) .debugfs_init = pl111_debugfs_init,
drivers/gpu/drm/vc4/vc4_bo.c +1 -1
@@ -782,7 +782,7 @@ int vc4_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)return -EINVAL ; } -return drm_gem_cma_prime_mmap (obj, vma); +return drm_gem_prime_mmap (obj, vma); }int vc4_prime_vmap (struct drm_gem_object *obj,struct dma_buf_map *map)
include/drm/drm_gem_cma_helper.h +3 -7
@@ -59,7 +59,7 @@ struct drm_gem_cma_object { .poll = drm_poll,\ .read = drm_read,\ .llseek = noop_llseek,\ - .mmap = drm_gem_cma_mmap,\ + .mmap = drm_gem_mmap,\ DRM_GEM_CMA_UNMAPPED_AREA_FOPS \ }
@@ -76,9 +76,6 @@ int drm_gem_cma_dumb_create(struct drm_file *file_priv,struct drm_device *drm, struct drm_mode_create_dumb *args); -/* set vm_flags and we can change the VM attribute to other one at here */ -int drm_gem_cma_mmap (struct file *filp,struct vm_area_struct *vma); -/* allocate physical memory */ struct drm_gem_cma_object *drm_gem_cma_create (struct drm_device *drm,size_t size);
@@ -101,9 +98,8 @@ struct drm_gem_object * drm_gem_cma_prime_import_sg_table(struct drm_device *dev, struct dma_buf_attachment *attach, struct sg_table *sgt); -int drm_gem_cma_prime_mmap (struct drm_gem_object *obj, -struct vm_area_struct *vma);int drm_gem_cma_vmap (struct drm_gem_object *obj,struct dma_buf_map *map); +int drm_gem_cma_mmap (struct drm_gem_object *obj,struct vm_area_struct *vma);/** * DRM_GEM_CMA_DRIVER_OPS_WITH_DUMB_CREATE - CMA GEM driver operations
@@ -123,7 +119,7 @@ int drm_gem_cma_vmap(struct drm_gem_object *obj, struct dma_buf_map *map); .prime_handle_to_fd = drm_gem_prime_handle_to_fd, \ .prime_fd_to_handle = drm_gem_prime_fd_to_handle, \ .gem_prime_import_sg_table = drm_gem_cma_prime_import_sg_table, \ - .gem_prime_mmap = drm_gem_cma_prime_mmap + .gem_prime_mmap = drm_gem_prime_mmap/** * DRM_GEM_CMA_DRIVER_OPS - CMA GEM driver operations
Signature · 6 lines
2.29.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH 2/2] drm/cma-helper: Implement mmap as GEM CMA object functions
Hi Thomas,
Quoted 8 lines · Thomas Zimmermann
The new GEM object function
drm_gem_cma_mmap()sets the VMA flags and offset as in the old implementation and immediately maps in the buffer's memory pages.Changing CMA helpers to use the GEM object function allows for the removal of the special implementations for mmap and
gem_prime_mmapcallbacks. The regular functionsdrm_gem_mmap()anddrm_gem_prime_mmap()are now used.
I've encountered a memory leak regression in our Renesas R-Car DU tests, and git bisection has led me to this patch (as commit f5ca8eb6f9).
Running the tests sequentially, while grepping /proc/meminfo for Cma, it is evident that CMA memory is not released, until exhausted and the allocations fail (seen in [0]) shown by the error report:
Quoted 2 lines
self.fbs.append(pykms.DumbFramebuffer(self.card, mode.hdisplay, mode.vdisplay, "XR24"))
ValueError:DRM_IOCTL_MODE_CREATE_DUMBfailed: Cannot allocate memory
Failing tests at f5ca8eb6f9 can be seen at [0], while the tests pass successfully [1] on the commit previous to that (bc2532ab7c2):
Reverting f5ca8eb6f9 also produces a successful pass [2]
[0] https://paste.ubuntu.com/p/VjPGPgswxR/ # Failed at f5ca8eb6f9
[1] https://paste.ubuntu.com/p/78RRp2WpNR/ # Success at bc2532ab7c2
[2] https://paste.ubuntu.com/p/qJKjZZN2pt/ # Success with revert
I don't believe we handle mmap specially in the RCar-DU driver, so I wonder if this issue has hit anyone else as well?
Any ideas of a repair without a revert ? Or do we just need to submit a revert?
I've yet to fully understand the implications of the patch below.
Thanks
--
Regards
Kieran
Quoted 238 lines
Signed-off-by Thomas Zimmermann <[email protected]>--- drivers/gpu/drm/drm_file.c | 3 +- drivers/gpu/drm/drm_gem_cma_helper.c | 121 +++++++++------------------ drivers/gpu/drm/pl111/pl111_drv.c | 2 +- drivers/gpu/drm/vc4/vc4_bo.c | 2 +- include/drm/drm_gem_cma_helper.h | 10 +-- 5 files changed, 44 insertions(+), 94 deletions(-)drivers/gpu/drm/drm_file.c +1
-2@@ -113,8 +113,7 @@ bool drm_dev_needs_global_mutex(struct drm_device *dev)* The memory mapping implementation will vary depending on how the driver * manages memory. Legacy drivers will use the deprecated drm_legacy_mmap() * function, modern drivers should use one of the provided memory-manager -* specific implementations. For GEM-based drivers this is drm_gem_mmap(), and -* for drivers which use the CMA GEM helpers it's drm_gem_cma_mmap(). +* specific implementations. For GEM-based drivers this is drm_gem_mmap(). * * No other file operations are supported by the DRM userspace API. Overall the * following is an example &file_operations structure:: drivers/gpu/drm/drm_gem_cma_helper.c +38
-83@@ -38,6 +38,7 @@ static const struct drm_gem_object_funcs drm_gem_cma_default_funcs = { .print_info = drm_gem_cma_print_info, .get_sg_table = drm_gem_cma_get_sg_table, .vmap = drm_gem_cma_vmap, + .mmap = drm_gem_cma_mmap, .vm_ops = &drm_gem_cma_vm_ops, };@@ -277,62 +278,6 @@ const struct vm_operations_struct drm_gem_cma_vm_ops = { };EXPORT_SYMBOL_GPL (drm_gem_cma_vm_ops); -static int drm_gem_cma_mmap_obj (struct drm_gem_cma_object *cma_obj, -struct vm_area_struct *vma) -{ -int ret; - -/* -* Clear the VM_PFNMAP flag that was set by drm_gem_mmap(), and set the -* vm_pgoff (used as a fake buffer offset by DRM) to 0 as we want to map -* the whole buffer. -*/ - vma->vm_flags &= ~VM_PFNMAP ; - vma->vm_pgoff =0 ; - - ret =dma_mmap_wc (cma_obj->base.dev->dev, vma, cma_obj->vaddr, - cma_obj->paddr, vma->vm_end - vma->vm_start); -if (ret) -drm_gem_vm_close (vma); - -return ret; -} - -/** -* drm_gem_cma_mmap - memory-map a CMA GEM object -* @filp: file object -* @vma: VMA for the area to be mapped -* -* This function implements an augmented version of the GEM DRM file mmap -* operation for CMA objects: In addition to the usual GEM VMA setup it -* immediately faults in the entire object instead of using on-demaind -* faulting. Drivers which employ the CMA helpers should use this function -* as their ->mmap() handler in the DRM device file's file_operations -* structure. -* -* Instead of directly referencing this function, drivers should use the -* DEFINE_DRM_GEM_CMA_FOPS().macro. -* -* Returns: -* 0 on success or a negative error code on failure. -*/ -int drm_gem_cma_mmap (struct file *filp,struct vm_area_struct *vma) -{ -struct drm_gem_cma_object *cma_obj; -struct drm_gem_object *gem_obj; -int ret; - - ret =drm_gem_mmap (filp, vma); -if (ret) -return ret; - - gem_obj = vma->vm_private_data; - cma_obj =to_drm_gem_cma_obj (gem_obj); - -return drm_gem_cma_mmap_obj (cma_obj, vma); -} -EXPORT_SYMBOL_GPL (drm_gem_cma_mmap); -#ifndef CONFIG_MMU /** * drm_gem_cma_get_unmapped_area - propose address for mapping in noMMU cases @@ -500,33 +445,6 @@ drm_gem_cma_prime_import_sg_table(struct drm_device *dev, }EXPORT_SYMBOL_GPL (drm_gem_cma_prime_import_sg_table); -/** -* drm_gem_cma_prime_mmap - memory-map an exported CMA GEM object -* @obj: GEM object -* @vma: VMA for the area to be mapped -* -* This function maps a buffer imported via DRM PRIME into a userspace -* process's address space. Drivers that use the CMA helpers should set this -* as their &drm_driver.gem_prime_mmap callback. -* -* Returns: -* 0 on success or a negative error code on failure. -*/ -int drm_gem_cma_prime_mmap (struct drm_gem_object *obj, -struct vm_area_struct *vma) -{ -struct drm_gem_cma_object *cma_obj; -int ret; - - ret =drm_gem_mmap_obj (obj, obj->size, vma); -if (ret <0 ) -return ret; - - cma_obj =to_drm_gem_cma_obj (obj); -return drm_gem_cma_mmap_obj (cma_obj, vma); -} -EXPORT_SYMBOL_GPL (drm_gem_cma_prime_mmap); -/** * drm_gem_cma_vmap - map a CMA GEM object into the kernel's virtual * address space @@ -553,6 +471,43 @@ int drm_gem_cma_vmap(struct drm_gem_object *obj, struct dma_buf_map *map) }EXPORT_SYMBOL_GPL (drm_gem_cma_vmap); +/** +* drm_gem_cma_mmap - memory-map an exported CMA GEM object +* @obj: GEM object +* @vma: VMA for the area to be mapped +* +* This function maps a buffer into a userspace process's address space. +* In addition to the usual GEM VMA setup it immediately faults in the entire +* object instead of using on-demand faulting. Drivers that use the CMA +* helpers should set this as their &drm_gem_object_funcs.mmap callback. +* +* Returns: +* 0 on success or a negative error code on failure. +*/ +int drm_gem_cma_mmap (struct drm_gem_object *obj,struct vm_area_struct *vma) +{ +struct drm_gem_cma_object *cma_obj; +int ret; + +/* +* Clear the VM_PFNMAP flag that was set by drm_gem_mmap(), and set the +* vm_pgoff (used as a fake buffer offset by DRM) to 0 as we want to map +* the whole buffer. +*/ + vma->vm_pgoff -=drm_vma_node_start (&obj->vma_node); + vma->vm_flags &= ~VM_PFNMAP ; + + cma_obj =to_drm_gem_cma_obj (obj); + + ret =dma_mmap_wc (cma_obj->base.dev->dev, vma, cma_obj->vaddr, + cma_obj->paddr, vma->vm_end - vma->vm_start); +if (ret) +drm_gem_vm_close (vma); + +return ret; +} +EXPORT_SYMBOL_GPL (drm_gem_cma_mmap); +/** * drm_gem_cma_prime_import_sg_table_vmap - PRIME import another driver's * scatter/gather table and get the virtual address of the buffer drivers/gpu/drm/pl111/pl111_drv.c +1
-1@@ -228,7 +228,7 @@ static const struct drm_driver pl111_drm_driver = { .prime_handle_to_fd = drm_gem_prime_handle_to_fd, .prime_fd_to_handle = drm_gem_prime_fd_to_handle, .gem_prime_import_sg_table = pl111_gem_import_sg_table, - .gem_prime_mmap = drm_gem_cma_prime_mmap, + .gem_prime_mmap = drm_gem_prime_mmap,#if defined(CONFIG_DEBUG_FS ) .debugfs_init = pl111_debugfs_init,drivers/gpu/drm/vc4/vc4_bo.c +1
-1@@ -782,7 +782,7 @@ int vc4_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)return -EINVAL ; } -return drm_gem_cma_prime_mmap (obj, vma); +return drm_gem_prime_mmap (obj, vma); }int vc4_prime_vmap (struct drm_gem_object *obj,struct dma_buf_map *map)include/drm/drm_gem_cma_helper.h +2
-6@@ -59,7 +59,7 @@ struct drm_gem_cma_object { .poll = drm_poll,\ .read = drm_read,\ .llseek = noop_llseek,\ - .mmap = drm_gem_cma_mmap,\ + .mmap = drm_gem_mmap,\ DRM_GEM_CMA_UNMAPPED_AREA_FOPS \ }@@ -76,9 +76,6 @@ int drm_gem_cma_dumb_create(struct drm_file *file_priv,struct drm_device *drm, struct drm_mode_create_dumb *args); -/* set vm_flags and we can change the VM attribute to other one at here */ -int drm_gem_cma_mmap (struct file *filp,struct vm_area_struct *vma); -/* allocate physical memory */ struct drm_gem_cma_object *drm_gem_cma_create (struct drm_device *drm,size_t size);@@ -101,9 +98,8 @@ struct drm_gem_object * drm_gem_cma_prime_import_sg_table(struct drm_device *dev, struct dma_buf_attachment *attach, struct sg_table *sgt); -int drm_gem_cma_prime_mmap (struct drm_gem_object *obj, -struct vm_area_struct *vma);int drm_gem_cma_vmap (struct drm_gem_object *obj,struct dma_buf_map *map); +int drm_gem_cma_mmap (struct drm_gem_object *obj,struct vm_area_struct *vma);/** * DRM_GEM_CMA_DRIVER_OPS_WITH_DUMB_CREATE - CMA GEM driver operations
include/drm/drm_gem_cma_helper.h int drm_gem_cma_vmap(struct drm_gem_object *obj, struct dma_buf_map *map); · line 125
Quoted 8 lines
include/drm/drm_gem_cma_helper.h +1
-1@@ -123,7 +119,7 @@ int drm_gem_cma_vmap(struct drm_gem_object *obj, struct dma_buf_map *map); .prime_handle_to_fd = drm_gem_prime_handle_to_fd, \ .prime_fd_to_handle = drm_gem_prime_fd_to_handle, \ .gem_prime_import_sg_table = drm_gem_cma_prime_import_sg_table, \ - .gem_prime_mmap = drm_gem_cma_prime_mmap + .gem_prime_mmap = drm_gem_prime_mmap/** * DRM_GEM_CMA_DRIVER_OPS - CMA GEM driver operations
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Thomas Zimmermann
maintainerRe: [PATCH 2/2] drm/cma-helper: Implement mmap as GEM CMA object functions
Hi Kieran
Am 14.01.21 um 13:51 schrieb Kieran Bingham:
Quoted 33 lines
Hi Thomas,
Quoted 8 lines · Thomas Zimmermann
The new GEM object function
drm_gem_cma_mmap()sets the VMA flags and offset as in the old implementation and immediately maps in the buffer's memory pages.Changing CMA helpers to use the GEM object function allows for the removal of the special implementations for mmap and
gem_prime_mmapcallbacks. The regular functionsdrm_gem_mmap()anddrm_gem_prime_mmap()are now used.I've encountered a memory leak regression in our Renesas R-Car DU tests, and git bisection has led me to this patch (as commit
f5ca8eb6f9).Running the tests sequentially, while grepping
/proc/meminfofor Cma, it is evident that CMA memory is not released, until exhausted and the allocations fail (seen in [0]) shown by the error report:Quoted 2 lines
self.fbs.append(pykms.DumbFramebuffer(self.card, mode.hdisplay, mode.vdisplay, "XR24"))
ValueError:DRM_IOCTL_MODE_CREATE_DUMBfailed: Cannot allocate memoryFailing tests at
f5ca8eb6f9can be seen at [0], while the tests pass successfully [1] on the commit previous to that (bc2532ab7c2):Reverting
f5ca8eb6f9also produces a successful pass [2][0] https://paste.ubuntu.com/p/VjPGPgswxR/ # Failed at
f5ca8eb6f9
[1] https://paste.ubuntu.com/p/78RRp2WpNR/ # Success atbc2532ab7c2
[2] https://paste.ubuntu.com/p/qJKjZZN2pt/ # Success with revert
Quoted 5 lines
I don't believe we handle mmap specially in the RCar-DU driver, so I wonder if this issue has hit anyone else as well?
Any ideas of a repair without a revert ? Or do we just need to submit a revert?
I think we might not be setting the VMA ops and therefore not finalize the BO correctly. Could you please apply the attched (quick-and-dirty) patch and try again?
Best regards
Thomas
Quoted 263 lines
I've yet to fully understand the implications of the patch below.
Thanks
--
RegardsKieran
Quoted 238 lines
Signed-off-by Thomas Zimmermann <[email protected]>--- drivers/gpu/drm/drm_file.c | 3 +- drivers/gpu/drm/drm_gem_cma_helper.c | 121 +++++++++------------------ drivers/gpu/drm/pl111/pl111_drv.c | 2 +- drivers/gpu/drm/vc4/vc4_bo.c | 2 +- include/drm/drm_gem_cma_helper.h | 10 +-- 5 files changed, 44 insertions(+), 94 deletions(-)drivers/gpu/drm/drm_file.c +1
-2@@ -113,8 +113,7 @@ bool drm_dev_needs_global_mutex(struct drm_device *dev)* The memory mapping implementation will vary depending on how the driver * manages memory. Legacy drivers will use the deprecated drm_legacy_mmap() * function, modern drivers should use one of the provided memory-manager -* specific implementations. For GEM-based drivers this is drm_gem_mmap(), and -* for drivers which use the CMA GEM helpers it's drm_gem_cma_mmap(). +* specific implementations. For GEM-based drivers this is drm_gem_mmap(). * * No other file operations are supported by the DRM userspace API. Overall the * following is an example &file_operations structure:: drivers/gpu/drm/drm_gem_cma_helper.c +38
-83@@ -38,6 +38,7 @@ static const struct drm_gem_object_funcs drm_gem_cma_default_funcs = { .print_info = drm_gem_cma_print_info, .get_sg_table = drm_gem_cma_get_sg_table, .vmap = drm_gem_cma_vmap, + .mmap = drm_gem_cma_mmap, .vm_ops = &drm_gem_cma_vm_ops, };@@ -277,62 +278,6 @@ const struct vm_operations_struct drm_gem_cma_vm_ops = { };EXPORT_SYMBOL_GPL (drm_gem_cma_vm_ops); -static int drm_gem_cma_mmap_obj (struct drm_gem_cma_object *cma_obj, -struct vm_area_struct *vma) -{ -int ret; - -/* -* Clear the VM_PFNMAP flag that was set by drm_gem_mmap(), and set the -* vm_pgoff (used as a fake buffer offset by DRM) to 0 as we want to map -* the whole buffer. -*/ - vma->vm_flags &= ~VM_PFNMAP ; - vma->vm_pgoff =0 ; - - ret =dma_mmap_wc (cma_obj->base.dev->dev, vma, cma_obj->vaddr, - cma_obj->paddr, vma->vm_end - vma->vm_start); -if (ret) -drm_gem_vm_close (vma); - -return ret; -} - -/** -* drm_gem_cma_mmap - memory-map a CMA GEM object -* @filp: file object -* @vma: VMA for the area to be mapped -* -* This function implements an augmented version of the GEM DRM file mmap -* operation for CMA objects: In addition to the usual GEM VMA setup it -* immediately faults in the entire object instead of using on-demaind -* faulting. Drivers which employ the CMA helpers should use this function -* as their ->mmap() handler in the DRM device file's file_operations -* structure. -* -* Instead of directly referencing this function, drivers should use the -* DEFINE_DRM_GEM_CMA_FOPS().macro. -* -* Returns: -* 0 on success or a negative error code on failure. -*/ -int drm_gem_cma_mmap (struct file *filp,struct vm_area_struct *vma) -{ -struct drm_gem_cma_object *cma_obj; -struct drm_gem_object *gem_obj; -int ret; - - ret =drm_gem_mmap (filp, vma); -if (ret) -return ret; - - gem_obj = vma->vm_private_data; - cma_obj =to_drm_gem_cma_obj (gem_obj); - -return drm_gem_cma_mmap_obj (cma_obj, vma); -} -EXPORT_SYMBOL_GPL (drm_gem_cma_mmap); -#ifndef CONFIG_MMU /** * drm_gem_cma_get_unmapped_area - propose address for mapping in noMMU cases @@ -500,33 +445,6 @@ drm_gem_cma_prime_import_sg_table(struct drm_device *dev, }EXPORT_SYMBOL_GPL (drm_gem_cma_prime_import_sg_table); -/** -* drm_gem_cma_prime_mmap - memory-map an exported CMA GEM object -* @obj: GEM object -* @vma: VMA for the area to be mapped -* -* This function maps a buffer imported via DRM PRIME into a userspace -* process's address space. Drivers that use the CMA helpers should set this -* as their &drm_driver.gem_prime_mmap callback. -* -* Returns: -* 0 on success or a negative error code on failure. -*/ -int drm_gem_cma_prime_mmap (struct drm_gem_object *obj, -struct vm_area_struct *vma) -{ -struct drm_gem_cma_object *cma_obj; -int ret; - - ret =drm_gem_mmap_obj (obj, obj->size, vma); -if (ret <0 ) -return ret; - - cma_obj =to_drm_gem_cma_obj (obj); -return drm_gem_cma_mmap_obj (cma_obj, vma); -} -EXPORT_SYMBOL_GPL (drm_gem_cma_prime_mmap); -/** * drm_gem_cma_vmap - map a CMA GEM object into the kernel's virtual * address space @@ -553,6 +471,43 @@ int drm_gem_cma_vmap(struct drm_gem_object *obj, struct dma_buf_map *map) }EXPORT_SYMBOL_GPL (drm_gem_cma_vmap); +/** +* drm_gem_cma_mmap - memory-map an exported CMA GEM object +* @obj: GEM object +* @vma: VMA for the area to be mapped +* +* This function maps a buffer into a userspace process's address space. +* In addition to the usual GEM VMA setup it immediately faults in the entire +* object instead of using on-demand faulting. Drivers that use the CMA +* helpers should set this as their &drm_gem_object_funcs.mmap callback. +* +* Returns: +* 0 on success or a negative error code on failure. +*/ +int drm_gem_cma_mmap (struct drm_gem_object *obj,struct vm_area_struct *vma) +{ +struct drm_gem_cma_object *cma_obj; +int ret; + +/* +* Clear the VM_PFNMAP flag that was set by drm_gem_mmap(), and set the +* vm_pgoff (used as a fake buffer offset by DRM) to 0 as we want to map +* the whole buffer. +*/ + vma->vm_pgoff -=drm_vma_node_start (&obj->vma_node); + vma->vm_flags &= ~VM_PFNMAP ; + + cma_obj =to_drm_gem_cma_obj (obj); + + ret =dma_mmap_wc (cma_obj->base.dev->dev, vma, cma_obj->vaddr, + cma_obj->paddr, vma->vm_end - vma->vm_start); +if (ret) +drm_gem_vm_close (vma); + +return ret; +} +EXPORT_SYMBOL_GPL (drm_gem_cma_mmap); +/** * drm_gem_cma_prime_import_sg_table_vmap - PRIME import another driver's * scatter/gather table and get the virtual address of the buffer drivers/gpu/drm/pl111/pl111_drv.c +1
-1@@ -228,7 +228,7 @@ static const struct drm_driver pl111_drm_driver = { .prime_handle_to_fd = drm_gem_prime_handle_to_fd, .prime_fd_to_handle = drm_gem_prime_fd_to_handle, .gem_prime_import_sg_table = pl111_gem_import_sg_table, - .gem_prime_mmap = drm_gem_cma_prime_mmap, + .gem_prime_mmap = drm_gem_prime_mmap,#if defined(CONFIG_DEBUG_FS ) .debugfs_init = pl111_debugfs_init,drivers/gpu/drm/vc4/vc4_bo.c +1
-1@@ -782,7 +782,7 @@ int vc4_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)return -EINVAL ; } -return drm_gem_cma_prime_mmap (obj, vma); +return drm_gem_prime_mmap (obj, vma); }int vc4_prime_vmap (struct drm_gem_object *obj,struct dma_buf_map *map)include/drm/drm_gem_cma_helper.h +2
-6@@ -59,7 +59,7 @@ struct drm_gem_cma_object { .poll = drm_poll,\ .read = drm_read,\ .llseek = noop_llseek,\ - .mmap = drm_gem_cma_mmap,\ + .mmap = drm_gem_mmap,\ DRM_GEM_CMA_UNMAPPED_AREA_FOPS \ }@@ -76,9 +76,6 @@ int drm_gem_cma_dumb_create(struct drm_file *file_priv,struct drm_device *drm, struct drm_mode_create_dumb *args); -/* set vm_flags and we can change the VM attribute to other one at here */ -int drm_gem_cma_mmap (struct file *filp,struct vm_area_struct *vma); -/* allocate physical memory */ struct drm_gem_cma_object *drm_gem_cma_create (struct drm_device *drm,size_t size);@@ -101,9 +98,8 @@ struct drm_gem_object * drm_gem_cma_prime_import_sg_table(struct drm_device *dev, struct dma_buf_attachment *attach, struct sg_table *sgt); -int drm_gem_cma_prime_mmap (struct drm_gem_object *obj, -struct vm_area_struct *vma);int drm_gem_cma_vmap (struct drm_gem_object *obj,struct dma_buf_map *map); +int drm_gem_cma_mmap (struct drm_gem_object *obj,struct vm_area_struct *vma);/** * DRM_GEM_CMA_DRIVER_OPS_WITH_DUMB_CREATE - CMA GEM driver operations include/drm/drm_gem_cma_helper.h int drm_gem_cma_vmap(struct drm_gem_object *obj, struct dma_buf_map *map); · line 125
Quoted 8 lines
include/drm/drm_gem_cma_helper.h +1
-1@@ -123,7 +119,7 @@ int drm_gem_cma_vmap(struct drm_gem_object *obj, struct dma_buf_map *map); .prime_handle_to_fd = drm_gem_prime_handle_to_fd, \ .prime_fd_to_handle = drm_gem_prime_fd_to_handle, \ .gem_prime_import_sg_table = drm_gem_cma_prime_import_sg_table, \ - .gem_prime_mmap = drm_gem_cma_prime_mmap + .gem_prime_mmap = drm_gem_prime_mmap/** * DRM_GEM_CMA_DRIVER_OPS - CMA GEM driver operations _______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Signature · 6 lines
Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Felix Imendörffer
From d0583fe22cd0cd29749ff679e46e13b58de325cb Mon Sep 17 00:00:00 2001
From: Thomas Zimmermann <[email protected]>
Date: Thu, 14 Jan 2021 14:21:51 +0100
Subject: [PATCH] drm/cma: Set vma ops in mmap function
--- drivers/gpu/drm/drm_gem_cma_helper.c | 2 ++ 1 file changed, 2 insertions(+)
drivers/gpu/drm/drm_gem_cma_helper.c +2 -0
@@ -489,6 +489,8 @@ int drm_gem_cma_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)struct drm_gem_cma_object *cma_obj;int ret; + vma->vm_ops = obj->funcs->vm_ops; +/* * Clear the VM_PFNMAP flag that was set by drm_gem_mmap(), and set the * vm_pgoff (used as a fake buffer offset by DRM) to 0 as we want to map
Signature · 1 line
2.29.2
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Kieran Bingham
Tested-byRe: [PATCH 2/2] drm/cma-helper: Implement mmap as GEM CMA object functions
Hi Thomas,
Quoted 40 lines · Thomas Zimmermann
Hi Kieran
Am 14.01.21 um 13:51 schrieb Kieran Bingham:
Quoted 37 lines
Hi Thomas,
Quoted 8 lines · Thomas Zimmermann
The new GEM object function
drm_gem_cma_mmap()sets the VMA flags and offset as in the old implementation and immediately maps in the buffer's memory pages.Changing CMA helpers to use the GEM object function allows for the removal of the special implementations for mmap and
gem_prime_mmapcallbacks. The regular functionsdrm_gem_mmap()anddrm_gem_prime_mmap()are now used.I've encountered a memory leak regression in our Renesas R-Car DU tests, and git bisection has led me to this patch (as commit
f5ca8eb6f9).Running the tests sequentially, while grepping
/proc/meminfofor Cma, it is evident that CMA memory is not released, until exhausted and the allocations fail (seen in [0]) shown by the error report:Quoted 3 lines
self.fbs.append(pykms.DumbFramebuffer(self.card, mode.hdisplay, mode.vdisplay, "XR24"))
ValueError:DRM_IOCTL_MODE_CREATE_DUMBfailed: Cannot allocate memoryFailing tests at
f5ca8eb6f9can be seen at [0], while the tests pass successfully [1] on the commit previous to that (bc2532ab7c2):Reverting
f5ca8eb6f9also produces a successful pass [2][0] https://paste.ubuntu.com/p/VjPGPgswxR/ # Failed at
f5ca8eb6f9
[1] https://paste.ubuntu.com/p/78RRp2WpNR/ # Success atbc2532ab7c2
[2] https://paste.ubuntu.com/p/qJKjZZN2pt/ # Success with revertI don't believe we handle mmap specially in the RCar-DU driver, so I wonder if this issue has hit anyone else as well?
Quoted 6 lines
Quoted 2 lines
Any ideas of a repair without a revert ? Or do we just need to submit a revert?
I think we might not be setting the VMA ops and therefore not finalize the BO correctly. Could you please apply the attched (quick-and-dirty) patch and try again?
Thanks for the quick response.
I can confirm the quick-and-dirty patch resolves the issue:
https://paste.ubuntu.com/p/sKDp3dNvwV/
You can add a
if it stays like that, but I suspect there might be a better place to initialise the ops rather than in the mmap call itself.
Happy to do further testing as required.
--
Regards
Kieran
Quoted 291 lines
Best regards
ThomasQuoted 287 lines
I've yet to fully understand the implications of the patch below.
Thanks
Signature · 3 lines
Regards KieranQuoted 262 lines
Signed-off-by Thomas Zimmermann <[email protected]>---
drivers/gpu/drm/drm_file.c | 3 +-drivers/gpu/drm/drm_gem_cma_helper.c| 121 +++++++++------------------ drivers/gpu/drm/pl111/pl111_drv.c | 2 +- drivers/gpu/drm/vc4/vc4_bo.c | 2 +- include/drm/drm_gem_cma_helper.h | 10 +--5 files changed, 44 insertions(+), 94 deletions(-)drivers/gpu/drm/drm_file.c +0
-0@@ -113,8 +113,7 @@ bool drm_dev_needs_global_mutex(struct drm_device *dev) * The memory mapping implementation will vary depending on how thedriver
* manages memory. Legacy drivers will use the deprecateddrm_legacy_mmap()
* function, modern drivers should use one of the provided memory-manager
- * specific implementations. For GEM-based drivers this isdrm_gem_mmap(), and
- * for drivers which use the CMA GEM helpers it'sdrm_gem_cma_mmap().
+ * specific implementations. For GEM-based drivers this isdrm_gem_mmap().
*
* No other file operations are supported by the DRM userspace API. Overall the
* following is an example&file_operationsstructure::drivers/gpu/drm/drm_gem_cma_helper.c +1
-0@@ -38,6 +38,7 @@ static const struct drm_gem_object_funcs drm_gem_cma_default_funcs = { .print_info = drm_gem_cma_print_info, .get_sg_table = drm_gem_cma_get_sg_table, .vmap = drm_gem_cma_vmap, + .mmap = drm_gem_cma_mmap, .vm_ops = &drm_gem_cma_vm_ops, };@@ -277,62 +278,6 @@ const struct
vm_operations_structdrm_gem_cma_vm_ops= {
};
EXPORT_SYMBOL_GPL(drm_gem_cma_vm_ops);
-static intdrm_gem_cma_mmap_obj(struct drm_gem_cma_object *cma_obj, - struct vm_area_struct *vma)drivers/gpu/drm/drm_gem_cma_helper.c +0
-5-{ - int ret; - - /* -* Clear the VM_PFNMAP flag that was set by drm_gem_mmap(), and set the
- *vm_pgoff(used as a fake buffer offset by DRM) to 0 as we want to map
- * the whole buffer.
- */
-vma->vm_flags&= ~VM_PFNMAP;
-vma->vm_pgoff= 0;drivers/gpu/drm/drm_gem_cma_helper.c +0
-15- - ret = dma_mmap_wc(cma_obj->base.dev->dev, vma, cma_obj->vaddr, - cma_obj->paddr, vma->vm_end - vma->vm_start); - if (ret) - drm_gem_vm_close(vma); - - return ret; -} - -/** -* drm_gem_cma_mmap - memory-map a CMA GEM object -* @filp: file object -* @vma: VMA for the area to be mapped -* -* This function implements an augmented version of the GEM DRM file mmap
- * operation for CMA objects: In addition to the usual GEM VMA setup it
- * immediately faults in the entire object instead of using on-demaind
- * faulting. Drivers which employ the CMA helpers should use this function
- * as their ->mmap() handler in the DRM device file'sfile_operations
- * structure.
- *
- * Instead of directly referencing this function, drivers should use the
- * DEFINE_DRM_GEM_CMA_FOPS().macro.
- *
- * Returns:
- * 0 on success or a negative error code on failure.
- */drivers/gpu/drm/drm_gem_cma_helper.c +0
-17-int drm_gem_cma_mmap (struct file *filp,struct vm_area_struct *vma) -{ - struct drm_gem_cma_object *cma_obj; - struct drm_gem_object *gem_obj; - int ret; - - ret = drm_gem_mmap(filp, vma); - if (ret) - return ret; - - gem_obj = vma->vm_private_data; - cma_obj = to_drm_gem_cma_obj(gem_obj); - - return drm_gem_cma_mmap_obj(cma_obj, vma); -} -EXPORT_SYMBOL_GPL (drm_gem_cma_mmap); -#ifndef
CONFIG_MMU
/**
*drm_gem_cma_get_unmapped_area- propose address for mapping in noMMU casesdrivers/gpu/drm/drm_gem_cma_helper.c +0
-0@@ -500,33 +445,6 @@ drm_gem_cma_prime_import_sg_table(structdrm_device *dev, } EXPORT_SYMBOL_GPL(drm_gem_cma_prime_import_sg_table);-/**
- *drm_gem_cma_prime_mmap- memory-map an exported CMA GEM object
- * @obj: GEM object
- * @vma: VMA for the area to be mapped
- *
- * This function maps a buffer imported via DRM PRIME into a userspace
- * process's address space. Drivers that use the CMA helpers should set this
- * as their &drm_driver.gem_prime_mmap callback.
- *
- * Returns:
- * 0 on success or a negative error code on failure.
- */drivers/gpu/drm/drm_gem_cma_helper.c +0
-15-int drm_gem_cma_prime_mmap (struct drm_gem_object *obj, - struct vm_area_struct *vma) -{ - struct drm_gem_cma_object *cma_obj; - int ret; - - ret = drm_gem_mmap_obj(obj, obj->size, vma); - if (ret < 0) - return ret; - - cma_obj = to_drm_gem_cma_obj(obj); - return drm_gem_cma_mmap_obj(cma_obj, vma); -} -EXPORT_SYMBOL_GPL (drm_gem_cma_prime_mmap); -/**
*drm_gem_cma_vmap- map a CMA GEM object into the kernel's virtual * address spacedrivers/gpu/drm/drm_gem_cma_helper.c +0
-0@@ -553,6 +471,43 @@ int drm_gem_cma_vmap(struct drm_gem_object *obj, struct dma_buf_map *map) } EXPORT_SYMBOL_GPL(drm_gem_cma_vmap);+/**
+ *drm_gem_cma_mmap- memory-map an exported CMA GEM object
+ * @obj: GEM object
+ * @vma: VMA for the area to be mapped
+ *
+ * This function maps a buffer into a userspace process's address space.
+ * In addition to the usual GEM VMA setup it immediately faults in the entire
+ * object instead of using on-demand faulting. Drivers that use the CMA
+ * helpers should set this as their &drm_gem_object_funcs.mmap callback.
+ *
+ * Returns:
+ * 0 on success or a negative error code on failure.
+ */
+intdrm_gem_cma_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)drivers/gpu/drm/drm_gem_cma_helper.c +6
-0+{ + struct drm_gem_cma_object *cma_obj; + int ret; + + /* +* Clear the VM_PFNMAP flag that was set by drm_gem_mmap(), and set the
+ *vm_pgoff(used as a fake buffer offset by DRM) to 0 as we want to map
+ * the whole buffer.
+ */
+vma->vm_pgoff-=drm_vma_node_start(&obj->vma_node); +vma->vm_flags&= ~VM_PFNMAP;drivers/gpu/drm/drm_gem_cma_helper.c +12
-0+ + cma_obj = to_drm_gem_cma_obj(obj); + + ret = dma_mmap_wc(cma_obj->base.dev->dev, vma, cma_obj->vaddr, + cma_obj->paddr, vma->vm_end - vma->vm_start); + if (ret) + drm_gem_vm_close(vma); + + return ret; +} +EXPORT_SYMBOL_GPL (drm_gem_cma_mmap); +/**
*drm_gem_cma_prime_import_sg_table_vmap- PRIME import another driver's
* scatter/gather table and get the virtual address of the bufferdrivers/gpu/drm/pl111/pl111_drv.c +1
-1@@ -228,7 +228,7 @@ static const struct drm_driver pl111_drm_driver = { .prime_handle_to_fd = drm_gem_prime_handle_to_fd, .prime_fd_to_handle = drm_gem_prime_fd_to_handle, .gem_prime_import_sg_table = pl111_gem_import_sg_table, - .gem_prime_mmap = drm_gem_cma_prime_mmap, + .gem_prime_mmap = drm_gem_prime_mmap, #if defined(CONFIG_DEBUG_FS) .debugfs_init = pl111_debugfs_init,drivers/gpu/drm/vc4/vc4_bo.c +1
-0@@ -782,7 +782,7 @@ int vc4_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) return -EINVAL; } - return drm_gem_cma_prime_mmap(obj, vma); + return drm_gem_prime_mmap(obj, vma); }int
vc4_prime_vmap(struct drm_gem_object *obj, struct dma_buf_map *map)include/drm/drm_gem_cma_helper.h +1
-1@@ -59,7 +59,7 @@ struct drm_gem_cma_object { .poll = drm_poll,\ .read = drm_read,\ .llseek = noop_llseek,\ - .mmap = drm_gem_cma_mmap,\ + .mmap = drm_gem_mmap,\ DRM_GEM_CMA_UNMAPPED_AREA_FOPS \ }@@ -76,9 +76,6 @@ int drm_gem_cma_dumb_create(struct
drm_file*file_priv,
structdrm_device*drm,
structdrm_mode_create_dumb*args);
-/* setvm_flagsand we can change the VM attribute to other one at here */include/drm/drm_gem_cma_helper.h +0
-2-int drm_gem_cma_mmap (struct file *filp,struct vm_area_struct *vma); -/* allocate physical memory */
structdrm_gem_cma_object*drm_gem_cma_create(structdrm_device*drm,size_tsize);include/drm/drm_gem_cma_helper.h +0
-2@@ -101,9 +98,8 @@ struct drm_gem_object * drm_gem_cma_prime_import_sg_table(struct drm_device *dev, struct dma_buf_attachment *attach, struct sg_table *sgt); -int drm_gem_cma_prime_mmap (struct drm_gem_object *obj, - struct vm_area_struct *vma);int
drm_gem_cma_vmap(struct drm_gem_object *obj, struct dma_buf_map *map);
+intdrm_gem_cma_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma);
/**
*DRM_GEM_CMA_DRIVER_OPS_WITH_DUMB_CREATE- CMA GEM driver operationsinclude/drm/drm_gem_cma_helper.h int drm_gem_cma_vmap(struct drm_gem_object *obj, · line 125
Quoted 8 lines
include/drm/drm_gem_cma_helper.h +1
-1@@ -123,7 +119,7 @@ int drm_gem_cma_vmap(struct drm_gem_object *obj, struct dma_buf_map *map); .prime_handle_to_fd = drm_gem_prime_handle_to_fd, \ .prime_fd_to_handle = drm_gem_prime_fd_to_handle, \ .gem_prime_import_sg_table = drm_gem_cma_prime_import_sg_table, \ - .gem_prime_mmap = drm_gem_cma_prime_mmap + .gem_prime_mmap = drm_gem_prime_mmap /*** DRM_GEM_CMA_DRIVER_OPS - CMA GEM driver operations _______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Thomas Zimmermann
maintainerRe: [PATCH 2/2] drm/cma-helper: Implement mmap as GEM CMA object functions
Hi
Am 14.01.21 um 15:34 schrieb Kieran Bingham:
Quoted 50 lines
Hi Thomas,
Quoted 40 lines · Thomas Zimmermann
Hi Kieran
Am 14.01.21 um 13:51 schrieb Kieran Bingham:
Quoted 37 lines
Hi Thomas,
Quoted 8 lines · Thomas Zimmermann
The new GEM object function
drm_gem_cma_mmap()sets the VMA flags and offset as in the old implementation and immediately maps in the buffer's memory pages.Changing CMA helpers to use the GEM object function allows for the removal of the special implementations for mmap and
gem_prime_mmapcallbacks. The regular functionsdrm_gem_mmap()anddrm_gem_prime_mmap()are now used.I've encountered a memory leak regression in our Renesas R-Car DU tests, and git bisection has led me to this patch (as commit
f5ca8eb6f9).Running the tests sequentially, while grepping
/proc/meminfofor Cma, it is evident that CMA memory is not released, until exhausted and the allocations fail (seen in [0]) shown by the error report:Quoted 3 lines
self.fbs.append(pykms.DumbFramebuffer(self.card, mode.hdisplay,
mode.vdisplay, "XR24"))
ValueError:DRM_IOCTL_MODE_CREATE_DUMBfailed: Cannot allocate memoryFailing tests at
f5ca8eb6f9can be seen at [0], while the tests pass successfully [1] on the commit previous to that (bc2532ab7c2):Reverting
f5ca8eb6f9also produces a successful pass [2][0] https://paste.ubuntu.com/p/VjPGPgswxR/ # Failed at
f5ca8eb6f9
[1] https://paste.ubuntu.com/p/78RRp2WpNR/ # Success atbc2532ab7c2
[2] https://paste.ubuntu.com/p/qJKjZZN2pt/ # Success with revertI don't believe we handle mmap specially in the RCar-DU driver, so I wonder if this issue has hit anyone else as well?
Quoted 6 lines
Quoted 2 lines
Any ideas of a repair without a revert ? Or do we just need to submit a revert?
I think we might not be setting the VMA ops and therefore not finalize the BO correctly. Could you please apply the attched (quick-and-dirty) patch and try again?
Quoted 7 lines
Thanks for the quick response.
I can confirm the quick-and-dirty patch resolves the issue:
https://paste.ubuntu.com/p/sKDp3dNvwV/You can add a
Tested-by Kieran Bingham <[email protected]>
Great! If you don't mind, I'd also add you in the Reported-by tag.
Quoted 3 lines
if it stays like that, but I suspect there might be a better place to initialise the ops rather than in the mmap call itself.
I think that's the fix, basically. We could put such a line as a fall-back somewhere into the DRM core code. But I don't know if this really works with all drivers. Maybe there's one that requires vm_ops to be NULL.
Thanks for reporting this issue and testing quickly.
Best regards
Thomas
Quoted 301 lines
Happy to do further testing as required.
--
RegardsKieran
Quoted 291 lines
Best regards
ThomasQuoted 287 lines
I've yet to fully understand the implications of the patch below.
Thanks
Signature · 3 lines
Regards KieranQuoted 262 lines
Signed-off-by Thomas Zimmermann <[email protected]>--- drivers/gpu/drm/drm_file.c | 3 +- drivers/gpu/drm/drm_gem_cma_helper.c | 121 +++++++++------------------ drivers/gpu/drm/pl111/pl111_drv.c | 2 +- drivers/gpu/drm/vc4/vc4_bo.c | 2 +- include/drm/drm_gem_cma_helper.h | 10 +-- 5 files changed, 44 insertions(+), 94 deletions(-)drivers/gpu/drm/drm_file.c +0
-0@@ -113,8 +113,7 @@ bool drm_dev_needs_global_mutex(struct drm_device *dev) * The memory mapping implementation will vary depending on how thedriver
* manages memory. Legacy drivers will use the deprecateddrm_legacy_mmap()
* function, modern drivers should use one of the provided
memory-manager
- * specific implementations. For GEM-based drivers this isdrm_gem_mmap(), and
- * for drivers which use the CMA GEM helpers it'sdrm_gem_cma_mmap().
+ * specific implementations. For GEM-based drivers this isdrm_gem_mmap().
*
* No other file operations are supported by the DRM userspace API.
Overall the
* following is an example&file_operationsstructure::drivers/gpu/drm/drm_gem_cma_helper.c +1
-0@@ -38,6 +38,7 @@ static const struct drm_gem_object_funcs drm_gem_cma_default_funcs = { .print_info = drm_gem_cma_print_info, .get_sg_table = drm_gem_cma_get_sg_table, .vmap = drm_gem_cma_vmap, + .mmap = drm_gem_cma_mmap, .vm_ops = &drm_gem_cma_vm_ops, };@@ -277,62 +278,6 @@ const struct
vm_operations_structdrm_gem_cma_vm_ops= {
};
EXPORT_SYMBOL_GPL(drm_gem_cma_vm_ops);
-static int drm_gem_cma_mmap_obj(structdrm_gem_cma_object*cma_obj,
- structvm_area_struct*vma)drivers/gpu/drm/drm_gem_cma_helper.c +0
-5-{ - int ret; - - /* -* Clear the VM_PFNMAP flag that was set by drm_gem_mmap(), and set the
- *vm_pgoff(used as a fake buffer offset by DRM) to 0 as we want to map
- * the whole buffer.
- */
-vma->vm_flags&= ~VM_PFNMAP;
-vma->vm_pgoff= 0;drivers/gpu/drm/drm_gem_cma_helper.c +0
-15- - ret = dma_mmap_wc(cma_obj->base.dev->dev, vma, cma_obj->vaddr, - cma_obj->paddr, vma->vm_end - vma->vm_start); - if (ret) - drm_gem_vm_close(vma); - - return ret; -} - -/** -* drm_gem_cma_mmap - memory-map a CMA GEM object -* @filp: file object -* @vma: VMA for the area to be mapped -* -* This function implements an augmented version of the GEM DRM file mmap
- * operation for CMA objects: In addition to the usual GEM VMA setup it
- * immediately faults in the entire object instead of using on-demaind
- * faulting. Drivers which employ the CMA helpers should use this function
- * as their ->mmap() handler in the DRM device file'sfile_operations
- * structure.
- *
- * Instead of directly referencing this function, drivers should use the
- * DEFINE_DRM_GEM_CMA_FOPS().macro.
- *
- * Returns:
- * 0 on success or a negative error code on failure.
- */drivers/gpu/drm/drm_gem_cma_helper.c +0
-17-int drm_gem_cma_mmap (struct file *filp,struct vm_area_struct *vma) -{ - struct drm_gem_cma_object *cma_obj; - struct drm_gem_object *gem_obj; - int ret; - - ret = drm_gem_mmap(filp, vma); - if (ret) - return ret; - - gem_obj = vma->vm_private_data; - cma_obj = to_drm_gem_cma_obj(gem_obj); - - return drm_gem_cma_mmap_obj(cma_obj, vma); -} -EXPORT_SYMBOL_GPL (drm_gem_cma_mmap); - #ifndef CONFIG_MMU /*** drm_gem_cma_get_unmapped_area - propose address for mapping in noMMU cases
drivers/gpu/drm/drm_gem_cma_helper.c +0
-6@@ -500,33 +445,6 @@ drm_gem_cma_prime_import_sg_table(structdrm_device *dev, } EXPORT_SYMBOL_GPL(drm_gem_cma_prime_import_sg_table); -/** -* drm_gem_cma_prime_mmap - memory-map an exported CMA GEM object -* @obj: GEM object -* @vma: VMA for the area to be mapped -* -* This function maps a buffer imported via DRM PRIME into a userspace -* process's address space. Drivers that use the CMA helpers should set this
- * as their &drm_driver.gem_prime_mmap callback.
- *
- * Returns:
- * 0 on success or a negative error code on failure.
- */drivers/gpu/drm/drm_gem_cma_helper.c +5
-15-int drm_gem_cma_prime_mmap (struct drm_gem_object *obj, - struct vm_area_struct *vma) -{ - struct drm_gem_cma_object *cma_obj; - int ret; - - ret = drm_gem_mmap_obj(obj, obj->size, vma); - if (ret < 0) - return ret; - - cma_obj = to_drm_gem_cma_obj(obj); - return drm_gem_cma_mmap_obj(cma_obj, vma); -} -EXPORT_SYMBOL_GPL (drm_gem_cma_prime_mmap); - /*** drm_gem_cma_vmap - map a CMA GEM object into the kernel's virtual * address space @@ -553,6 +471,43 @@ int drm_gem_cma_vmap(struct drm_gem_object *obj, struct dma_buf_map *map) } EXPORT_SYMBOL_GPL(drm_gem_cma_vmap); +/** +* drm_gem_cma_mmap - memory-map an exported CMA GEM object +* @obj: GEM object +* @vma: VMA for the area to be mapped +* +* This function maps a buffer into a userspace process's address space.
+ * In addition to the usual GEM VMA setup it immediately faults in the entire
+ * object instead of using on-demand faulting. Drivers that use the CMA
+ * helpers should set this as their &drm_gem_object_funcs.mmap callback.
+ *
+ * Returns:
+ * 0 on success or a negative error code on failure.
+ */
+intdrm_gem_cma_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)drivers/gpu/drm/drm_gem_cma_helper.c +6
-0+{ + struct drm_gem_cma_object *cma_obj; + int ret; + + /* +* Clear the VM_PFNMAP flag that was set by drm_gem_mmap(), and set the
+ *vm_pgoff(used as a fake buffer offset by DRM) to 0 as we want to map
+ * the whole buffer.
+ */
+vma->vm_pgoff-=drm_vma_node_start(&obj->vma_node); +vma->vm_flags&= ~VM_PFNMAP;drivers/gpu/drm/drm_gem_cma_helper.c +12
-0+ + cma_obj = to_drm_gem_cma_obj(obj); + + ret = dma_mmap_wc(cma_obj->base.dev->dev, vma, cma_obj->vaddr, + cma_obj->paddr, vma->vm_end - vma->vm_start); + if (ret) + drm_gem_vm_close(vma); + + return ret; +} +EXPORT_SYMBOL_GPL (drm_gem_cma_mmap); + /*** drm_gem_cma_prime_import_sg_table_vmap - PRIME import another driver's
* scatter/gather table and get the virtual address of the bufferdrivers/gpu/drm/pl111/pl111_drv.c +1
-1@@ -228,7 +228,7 @@ static const struct drm_driver pl111_drm_driver = { .prime_handle_to_fd = drm_gem_prime_handle_to_fd, .prime_fd_to_handle = drm_gem_prime_fd_to_handle, .gem_prime_import_sg_table = pl111_gem_import_sg_table, - .gem_prime_mmap = drm_gem_cma_prime_mmap, + .gem_prime_mmap = drm_gem_prime_mmap, #if defined(CONFIG_DEBUG_FS) .debugfs_init = pl111_debugfs_init,drivers/gpu/drm/vc4/vc4_bo.c +1
-0@@ -782,7 +782,7 @@ int vc4_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) return -EINVAL; } - return drm_gem_cma_prime_mmap(obj, vma); + return drm_gem_prime_mmap(obj, vma); } int vc4_prime_vmap(struct drm_gem_object *obj, struct dma_buf_map*map)
include/drm/drm_gem_cma_helper.h +1
-1@@ -59,7 +59,7 @@ struct drm_gem_cma_object { .poll = drm_poll,\ .read = drm_read,\ .llseek = noop_llseek,\ - .mmap = drm_gem_cma_mmap,\ + .mmap = drm_gem_mmap,\ DRM_GEM_CMA_UNMAPPED_AREA_FOPS \ } @@ -76,9 +76,6 @@ int drm_gem_cma_dumb_create(struct drm_file
*file_priv,
structdrm_device*drm,
structdrm_mode_create_dumb*args);
-/* setvm_flagsand we can change the VM attribute to other one at
here */include/drm/drm_gem_cma_helper.h +1
-4-int drm_gem_cma_mmap (struct file *filp,struct vm_area_struct *vma); - /* allocate physical memory */ struct drm_gem_cma_object *drm_gem_cma_create(struct drm_device *drm, size_t size);@@ -101,9 +98,8 @@ struct drm_gem_object * drm_gem_cma_prime_import_sg_table(struct drm_device *dev, struct dma_buf_attachment *attach, struct sg_table *sgt); -int drm_gem_cma_prime_mmap (struct drm_gem_object *obj, - struct vm_area_struct *vma); int drm_gem_cma_vmap(struct drm_gem_object *obj, struct dma_buf_map *map); +int drm_gem_cma_mmap (struct drm_gem_object *obj,struct vm_area_struct *vma); /***
DRM_GEM_CMA_DRIVER_OPS_WITH_DUMB_CREATE- CMA GEM driver operationsinclude/drm/drm_gem_cma_helper.h int drm_gem_cma_vmap(struct drm_gem_object *obj, · line 125
Quoted 8 lines
include/drm/drm_gem_cma_helper.h +1
-1@@ -123,7 +119,7 @@ int drm_gem_cma_vmap(struct drm_gem_object *obj, struct dma_buf_map *map); .prime_handle_to_fd = drm_gem_prime_handle_to_fd, \ .prime_fd_to_handle = drm_gem_prime_fd_to_handle, \ .gem_prime_import_sg_table = drm_gem_cma_prime_import_sg_table, \ - .gem_prime_mmap = drm_gem_cma_prime_mmap + .gem_prime_mmap = drm_gem_prime_mmap /*** DRM_GEM_CMA_DRIVER_OPS - CMA GEM driver operations _______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Signature · 6 lines
Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Felix Imendörffer
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH 2/2] drm/cma-helper: Implement mmap as GEM CMA object functions
Hi Thomas,
Quoted 47 lines · Thomas Zimmermann
Quoted 47 lines
Quoted 38 lines
Quoted 38 lines · Thomas Zimmermann
Quoted 9 lines
The new GEM object function
drm_gem_cma_mmap()sets the VMA flags and offset as in the old implementation and immediately maps in the buffer's memory pages.Changing CMA helpers to use the GEM object function allows for the removal of the special implementations for mmap and
gem_prime_mmapcallbacks. The regular functionsdrm_gem_mmap()anddrm_gem_prime_mmap()
are now used.I've encountered a memory leak regression in our Renesas R-Car DU tests,
and git bisection has led me to this patch (as commitf5ca8eb6f9).Running the tests sequentially, while grepping
/proc/meminfofor Cma, it
is evident that CMA memory is not released, until exhausted and the allocations fail (seen in [0]) shown by the error report:Quoted 3 lines
self.fbs.append(pykms.DumbFramebuffer(self.card, mode.hdisplay, mode.vdisplay, "XR24"))
ValueError:DRM_IOCTL_MODE_CREATE_DUMBfailed: Cannot allocate memoryFailing tests at
f5ca8eb6f9can be seen at [0], while the tests pass successfully [1] on the commit previous to that (bc2532ab7c2):Reverting
f5ca8eb6f9also produces a successful pass [2][0] https://paste.ubuntu.com/p/VjPGPgswxR/ # Failed at
f5ca8eb6f9
[1] https://paste.ubuntu.com/p/78RRp2WpNR/ # Success atbc2532ab7c2
[2] https://paste.ubuntu.com/p/qJKjZZN2pt/ # Success with revertI don't believe we handle mmap specially in the RCar-DU driver, so I wonder if this issue has hit anyone else as well?
Quoted 6 lines
Quoted 2 lines
Any ideas of a repair without a revert ? Or do we just need to submit a revert?
I think we might not be setting the VMA ops and therefore not finalize the BO correctly. Could you please apply the attched (quick-and-dirty) patch and try again?
Thanks for the quick response.
Quoted 7 lines
Quoted 5 lines
I can confirm the quick-and-dirty patch resolves the issue:
https://paste.ubuntu.com/p/sKDp3dNvwV/You can add a
Tested-by Kieran Bingham <[email protected]>Great! If you don't mind, I'd also add you in the Reported-by tag.
Certainly!
Quoted 8 lines
Quoted 3 lines
if it stays like that, but I suspect there might be a better place to initialise the ops rather than in the mmap call itself.
I think that's the fix, basically. We could put such a line as a fall-back somewhere into the DRM core code. But I don't know if this really works with all drivers. Maybe there's one that requires
vm_opsto be NULL.
Ok, that's reaching beyond code I've explored, so I'll leave it to you.
Quoted 1 line
Thanks for reporting this issue and testing quickly.
Thanks for fixing so quickly :-)
Regards
Kieran
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Thomas Zimmermann
maintainerRe: [PATCH 2/2] drm/cma-helper: Implement mmap as GEM CMA object functions
Hi
Am 14.01.21 um 17:28 schrieb Kieran Bingham:
Quoted 64 lines
Hi Thomas,
Quoted 47 lines · Thomas Zimmermann
Quoted 47 lines
Quoted 38 lines
Quoted 38 lines · Thomas Zimmermann
Quoted 9 lines
The new GEM object function
drm_gem_cma_mmap()sets the VMA flags and offset as in the old implementation and immediately maps in the buffer's memory pages.Changing CMA helpers to use the GEM object function allows for the removal of the special implementations for mmap and
gem_prime_mmapcallbacks. The regular functionsdrm_gem_mmap()anddrm_gem_prime_mmap()
are now used.I've encountered a memory leak regression in our Renesas R-Car DU tests,
and git bisection has led me to this patch (as commitf5ca8eb6f9).Running the tests sequentially, while grepping
/proc/meminfofor Cma, it
is evident that CMA memory is not released, until exhausted and the allocations fail (seen in [0]) shown by the error report:Quoted 3 lines
self.fbs.append(pykms.DumbFramebuffer(self.card, mode.hdisplay,
mode.vdisplay, "XR24"))
ValueError:DRM_IOCTL_MODE_CREATE_DUMBfailed: Cannot allocate memoryFailing tests at
f5ca8eb6f9can be seen at [0], while the tests pass successfully [1] on the commit previous to that (bc2532ab7c2):Reverting
f5ca8eb6f9also produces a successful pass [2][0] https://paste.ubuntu.com/p/VjPGPgswxR/ # Failed at
f5ca8eb6f9
[1] https://paste.ubuntu.com/p/78RRp2WpNR/ # Success atbc2532ab7c2
[2] https://paste.ubuntu.com/p/qJKjZZN2pt/ # Success with revertI don't believe we handle mmap specially in the RCar-DU driver, so I wonder if this issue has hit anyone else as well?
Quoted 6 lines
Quoted 2 lines
Any ideas of a repair without a revert ? Or do we just need to submit a revert?
I think we might not be setting the VMA ops and therefore not finalize the BO correctly. Could you please apply the attched (quick-and-dirty) patch and try again?
Thanks for the quick response.
Quoted 7 lines
Quoted 5 lines
I can confirm the quick-and-dirty patch resolves the issue:
https://paste.ubuntu.com/p/sKDp3dNvwV/You can add a
Tested-by Kieran Bingham <[email protected]>Great! If you don't mind, I'd also add you in the Reported-by tag.
Certainly!
Quoted 4 lines
Quoted 3 lines
if it stays like that, but I suspect there might be a better place to initialise the ops rather than in the mmap call itself.
Quoted 6 lines
Quoted 4 lines
I think that's the fix, basically. We could put such a line as a fall-back somewhere into the DRM core code. But I don't know if this really works with all drivers. Maybe there's one that requires
vm_opsto be NULL.Ok, that's reaching beyond code I've explored, so I'll leave it to you.
Daniel asked for a fix in the DRM core, so I'll go with the alternative approach. If you have the time, I'd appreciate another test run.
Best regards
Thomas
Quoted 10 lines
Quoted 3 lines
Thanks for reporting this issue and testing quickly.
Thanks for fixing so quickly :-)
Regards
Kieran
Signature · 6 lines
Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Felix Imendörffer
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH 2/2] drm/cma-helper: Implement mmap as GEM CMA object functions
In reply to Thomas ZimmermannQuoted 15 lines · Thomas Zimmermann
From
d0583fe22cd0cd29749ff679e46e13b58de325cbMon Sep 17 00:00:00 2001
From: Thomas Zimmermann <[email protected]>
Date: Thu, 14 Jan 2021 14:21:51 +0100
Subject: [PATCH] drm/cma: Set vma ops in mmap functionSigned-off-by Thomas Zimmermann <[email protected]>--- drivers/gpu/drm/drm_gem_cma_helper.c | 2 ++ 1 file changed, 2 insertions(+)drivers/gpu/drm/drm_gem_cma_helper.c +0
-0
drivers/gpu/drm/drm_gem_cma_helper.c int drm_gem_cma_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) · line 492
Quoted 4 lines
drivers/gpu/drm/drm_gem_cma_helper.c +1
-0@@ -489,6 +489,8 @@ int drm_gem_cma_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)struct drm_gem_cma_object *cma_obj;int ret; + vma->vm_ops = obj->funcs->vm_ops;
I think this should be done in core, otherwise we have tons of refcount leaks. I think this was an oversight when we've done that refactoring.
Also drivers can easily overwrite this one if they really have to, but not assigned this is a clear bug.
-Daniel
Quoted 7 lines
drivers/gpu/drm/drm_gem_cma_helper.c +1
-0+/* * Clear the VM_PFNMAP flag that was set by drm_gem_mmap(), and set the * vm_pgoff (used as a fake buffer offset by DRM) to 0 as we want to map Signature · 1 line
2.29.2
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Quoted 5 lines · Thomas Zimmermann
This patchset moves CMA's mmap into a GEM object function. This change allows for removing CMA's mmap and
gem_prime_mmapcallbacks in favor of the default implementation.Tested with vc4 on RPi3.
For both patches,
Maxime
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Thomas Zimmermann
maintainerAm 23.11.20 um 13:32 schrieb Maxime Ripard:
Quoted 9 lines · Thomas Zimmermann
Quoted 5 lines
This patchset moves CMA's mmap into a GEM object function. This change allows for removing CMA's mmap and
gem_prime_mmapcallbacks in favor of the default implementation.Tested with vc4 on RPi3.
For both patches,
Acked-by Maxime Ripard <[email protected]>
Thanks, Maxime!
Quoted 3 lines
Maxime
Signature · 6 lines
Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Felix Imendörffer
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel