5 Modern Linux systems require large amount of graphics memory to store
6 frame buffers, textures, vertices and other graphics-related data. Given
7 the very dynamic nature of many of that data, managing graphics memory
8 efficiently is thus crucial for the graphics stack and plays a central
9 role in the DRM infrastructure.
11 The DRM core includes two memory managers, namely Translation Table Maps
12 (TTM) and Graphics Execution Manager (GEM). TTM was the first DRM memory
13 manager to be developed and tried to be a one-size-fits-them all
14 solution. It provides a single userspace API to accommodate the need of
15 all hardware, supporting both Unified Memory Architecture (UMA) devices
16 and devices with dedicated video RAM (i.e. most discrete video cards).
17 This resulted in a large, complex piece of code that turned out to be
18 hard to use for driver development.
20 GEM started as an Intel-sponsored project in reaction to TTM's
21 complexity. Its design philosophy is completely different: instead of
22 providing a solution to every graphics memory-related problems, GEM
23 identified common code between drivers and created a support library to
24 share it. GEM has simpler initialization and execution requirements than
25 TTM, but has no video RAM management capabilities and is thus limited to
28 The Translation Table Manager (TTM)
29 ===================================
31 .. kernel-doc:: drivers/gpu/drm/ttm/ttm_module.c
34 .. kernel-doc:: include/drm/ttm/ttm_caching.h
37 TTM device object reference
38 ---------------------------
40 .. kernel-doc:: include/drm/ttm/ttm_device.h
43 .. kernel-doc:: drivers/gpu/drm/ttm/ttm_device.c
46 TTM resource placement reference
47 --------------------------------
49 .. kernel-doc:: include/drm/ttm/ttm_placement.h
52 TTM resource object reference
53 -----------------------------
55 .. kernel-doc:: include/drm/ttm/ttm_resource.h
58 .. kernel-doc:: drivers/gpu/drm/ttm/ttm_resource.c
61 TTM TT object reference
62 -----------------------
64 .. kernel-doc:: include/drm/ttm/ttm_tt.h
67 .. kernel-doc:: drivers/gpu/drm/ttm/ttm_tt.c
70 TTM page pool reference
71 -----------------------
73 .. kernel-doc:: include/drm/ttm/ttm_pool.h
76 .. kernel-doc:: drivers/gpu/drm/ttm/ttm_pool.c
79 The Graphics Execution Manager (GEM)
80 ====================================
82 The GEM design approach has resulted in a memory manager that doesn't
83 provide full coverage of all (or even all common) use cases in its
84 userspace or kernel API. GEM exposes a set of standard memory-related
85 operations to userspace and a set of helper functions to drivers, and
86 let drivers implement hardware-specific operations with their own
89 The GEM userspace API is described in the `GEM - the Graphics Execution
90 Manager <http://lwn.net/Articles/283798/>`__ article on LWN. While
91 slightly outdated, the document provides a good overview of the GEM API
92 principles. Buffer allocation and read and write operations, described
93 as part of the common GEM API, are currently implemented using
94 driver-specific ioctls.
96 GEM is data-agnostic. It manages abstract buffer objects without knowing
97 what individual buffers contain. APIs that require knowledge of buffer
98 contents or purpose, such as buffer allocation or synchronization
99 primitives, are thus outside of the scope of GEM and must be implemented
100 using driver-specific ioctls.
102 On a fundamental level, GEM involves several operations:
104 - Memory allocation and freeing
106 - Aperture management at command execution time
108 Buffer object allocation is relatively straightforward and largely
109 provided by Linux's shmem layer, which provides memory to back each
112 Device-specific operations, such as command execution, pinning, buffer
113 read & write, mapping, and domain ownership transfers are left to
114 driver-specific ioctls.
119 Drivers that use GEM must set the DRIVER_GEM bit in the struct
120 :c:type:`struct drm_driver <drm_driver>` driver_features
121 field. The DRM core will then automatically initialize the GEM core
122 before calling the load operation. Behind the scene, this will create a
123 DRM Memory Manager object which provides an address space pool for
126 In a KMS configuration, drivers need to allocate and initialize a
127 command ring buffer following core GEM initialization if required by the
128 hardware. UMA devices usually have what is called a "stolen" memory
129 region, which provides space for the initial framebuffer and large,
130 contiguous memory regions required by the device. This space is
131 typically not managed by GEM, and must be initialized separately into
132 its own DRM MM object.
137 GEM splits creation of GEM objects and allocation of the memory that
138 backs them in two distinct operations.
140 GEM objects are represented by an instance of struct :c:type:`struct
141 drm_gem_object <drm_gem_object>`. Drivers usually need to
142 extend GEM objects with private information and thus create a
143 driver-specific GEM object structure type that embeds an instance of
144 struct :c:type:`struct drm_gem_object <drm_gem_object>`.
146 To create a GEM object, a driver allocates memory for an instance of its
147 specific GEM object type and initializes the embedded struct
148 :c:type:`struct drm_gem_object <drm_gem_object>` with a call
149 to drm_gem_object_init(). The function takes a pointer
150 to the DRM device, a pointer to the GEM object and the buffer object
153 GEM uses shmem to allocate anonymous pageable memory.
154 drm_gem_object_init() will create an shmfs file of the
155 requested size and store it into the struct :c:type:`struct
156 drm_gem_object <drm_gem_object>` filp field. The memory is
157 used as either main storage for the object when the graphics hardware
158 uses system memory directly or as a backing store otherwise.
160 Drivers are responsible for the actual physical pages allocation by
161 calling shmem_read_mapping_page_gfp() for each page.
162 Note that they can decide to allocate pages when initializing the GEM
163 object, or to delay allocation until the memory is needed (for instance
164 when a page fault occurs as a result of a userspace memory access or
165 when the driver needs to start a DMA transfer involving the memory).
167 Anonymous pageable memory allocation is not always desired, for instance
168 when the hardware requires physically contiguous system memory as is
169 often the case in embedded devices. Drivers can create GEM objects with
170 no shmfs backing (called private GEM objects) by initializing them with a call
171 to drm_gem_private_object_init() instead of drm_gem_object_init(). Storage for
172 private GEM objects must be managed by drivers.
177 All GEM objects are reference-counted by the GEM core. References can be
178 acquired and release by calling drm_gem_object_get() and drm_gem_object_put()
181 When the last reference to a GEM object is released the GEM core calls
182 the :c:type:`struct drm_gem_object_funcs <gem_object_funcs>` free
183 operation. That operation is mandatory for GEM-enabled drivers and must
184 free the GEM object and all associated resources.
186 void (\*free) (struct drm_gem_object \*obj); Drivers are
187 responsible for freeing all GEM object resources. This includes the
188 resources created by the GEM core, which need to be released with
189 drm_gem_object_release().
194 Communication between userspace and the kernel refers to GEM objects
195 using local handles, global names or, more recently, file descriptors.
196 All of those are 32-bit integer values; the usual Linux kernel limits
197 apply to the file descriptors.
199 GEM handles are local to a DRM file. Applications get a handle to a GEM
200 object through a driver-specific ioctl, and can use that handle to refer
201 to the GEM object in other standard or driver-specific ioctls. Closing a
202 DRM file handle frees all its GEM handles and dereferences the
203 associated GEM objects.
205 To create a handle for a GEM object drivers call drm_gem_handle_create(). The
206 function takes a pointer to the DRM file and the GEM object and returns a
207 locally unique handle. When the handle is no longer needed drivers delete it
208 with a call to drm_gem_handle_delete(). Finally the GEM object associated with a
209 handle can be retrieved by a call to drm_gem_object_lookup().
211 Handles don't take ownership of GEM objects, they only take a reference
212 to the object that will be dropped when the handle is destroyed. To
213 avoid leaking GEM objects, drivers must make sure they drop the
214 reference(s) they own (such as the initial reference taken at object
215 creation time) as appropriate, without any special consideration for the
216 handle. For example, in the particular case of combined GEM object and
217 handle creation in the implementation of the dumb_create operation,
218 drivers must drop the initial reference to the GEM object before
219 returning the handle.
221 GEM names are similar in purpose to handles but are not local to DRM
222 files. They can be passed between processes to reference a GEM object
223 globally. Names can't be used directly to refer to objects in the DRM
224 API, applications must convert handles to names and names to handles
225 using the DRM_IOCTL_GEM_FLINK and DRM_IOCTL_GEM_OPEN ioctls
226 respectively. The conversion is handled by the DRM core without any
227 driver-specific support.
229 GEM also supports buffer sharing with dma-buf file descriptors through
230 PRIME. GEM-based drivers must use the provided helpers functions to
231 implement the exporting and importing correctly. See ?. Since sharing
232 file descriptors is inherently more secure than the easily guessable and
233 global GEM names it is the preferred buffer sharing mechanism. Sharing
234 buffers through GEM names is only supported for legacy userspace.
235 Furthermore PRIME also allows cross-device buffer sharing since it is
241 Because mapping operations are fairly heavyweight GEM favours
242 read/write-like access to buffers, implemented through driver-specific
243 ioctls, over mapping buffers to userspace. However, when random access
244 to the buffer is needed (to perform software rendering for instance),
245 direct access to the object can be more efficient.
247 The mmap system call can't be used directly to map GEM objects, as they
248 don't have their own file handle. Two alternative methods currently
249 co-exist to map GEM objects to userspace. The first method uses a
250 driver-specific ioctl to perform the mapping operation, calling
251 do_mmap() under the hood. This is often considered
252 dubious, seems to be discouraged for new GEM-enabled drivers, and will
253 thus not be described here.
255 The second method uses the mmap system call on the DRM file handle. void
256 \*mmap(void \*addr, size_t length, int prot, int flags, int fd, off_t
257 offset); DRM identifies the GEM object to be mapped by a fake offset
258 passed through the mmap offset argument. Prior to being mapped, a GEM
259 object must thus be associated with a fake offset. To do so, drivers
260 must call drm_gem_create_mmap_offset() on the object.
262 Once allocated, the fake offset value must be passed to the application
263 in a driver-specific way and can then be used as the mmap offset
266 The GEM core provides a helper method drm_gem_mmap() to
267 handle object mapping. The method can be set directly as the mmap file
268 operation handler. It will look up the GEM object based on the offset
269 value and set the VMA operations to the :c:type:`struct drm_driver
270 <drm_driver>` gem_vm_ops field. Note that drm_gem_mmap() doesn't map memory to
271 userspace, but relies on the driver-provided fault handler to map pages
274 To use drm_gem_mmap(), drivers must fill the struct :c:type:`struct drm_driver
275 <drm_driver>` gem_vm_ops field with a pointer to VM operations.
277 The VM operations is a :c:type:`struct vm_operations_struct <vm_operations_struct>`
278 made up of several fields, the more interesting ones being:
282 struct vm_operations_struct {
283 void (*open)(struct vm_area_struct * area);
284 void (*close)(struct vm_area_struct * area);
285 vm_fault_t (*fault)(struct vm_fault *vmf);
289 The open and close operations must update the GEM object reference
290 count. Drivers can use the drm_gem_vm_open() and drm_gem_vm_close() helper
291 functions directly as open and close handlers.
293 The fault operation handler is responsible for mapping individual pages
294 to userspace when a page fault occurs. Depending on the memory
295 allocation scheme, drivers can allocate pages at fault time, or can
296 decide to allocate memory for the GEM object at the time the object is
299 Drivers that want to map the GEM object upfront instead of handling page
300 faults can implement their own mmap file operation handler.
302 For platforms without MMU the GEM core provides a helper method
303 drm_gem_cma_get_unmapped_area(). The mmap() routines will call this to get a
304 proposed address for the mapping.
306 To use drm_gem_cma_get_unmapped_area(), drivers must fill the struct
307 :c:type:`struct file_operations <file_operations>` get_unmapped_area field with
308 a pointer on drm_gem_cma_get_unmapped_area().
310 More detailed information about get_unmapped_area can be found in
311 Documentation/admin-guide/mm/nommu-mmap.rst
316 When mapped to the device or used in a command buffer, backing pages for
317 an object are flushed to memory and marked write combined so as to be
318 coherent with the GPU. Likewise, if the CPU accesses an object after the
319 GPU has finished rendering to the object, then the object must be made
320 coherent with the CPU's view of memory, usually involving GPU cache
321 flushing of various kinds. This core CPU<->GPU coherency management is
322 provided by a device-specific ioctl, which evaluates an object's current
323 domain and performs any necessary flushing or synchronization to put the
324 object into the desired coherency domain (note that the object may be
325 busy, i.e. an active render target; in that case, setting the domain
326 blocks the client and waits for rendering to complete before performing
327 any necessary flushing operations).
332 Perhaps the most important GEM function for GPU devices is providing a
333 command execution interface to clients. Client programs construct
334 command buffers containing references to previously allocated memory
335 objects, and then submit them to GEM. At that point, GEM takes care to
336 bind all the objects into the GTT, execute the buffer, and provide
337 necessary synchronization between clients accessing the same buffers.
338 This often involves evicting some objects from the GTT and re-binding
339 others (a fairly expensive operation), and providing relocation support
340 which hides fixed GTT offsets from clients. Clients must take care not
341 to submit command buffers that reference more objects than can fit in
342 the GTT; otherwise, GEM will reject them and no rendering will occur.
343 Similarly, if several objects in the buffer require fence registers to
344 be allocated for correct rendering (e.g. 2D blits on pre-965 chips),
345 care must be taken not to require more fence registers than are
346 available to the client. Such resource management should be abstracted
347 from the client in libdrm.
349 GEM Function Reference
350 ----------------------
352 .. kernel-doc:: include/drm/drm_gem.h
355 .. kernel-doc:: drivers/gpu/drm/drm_gem.c
358 GEM CMA Helper Functions Reference
359 ----------------------------------
361 .. kernel-doc:: drivers/gpu/drm/drm_gem_cma_helper.c
364 .. kernel-doc:: include/drm/drm_gem_cma_helper.h
367 .. kernel-doc:: drivers/gpu/drm/drm_gem_cma_helper.c
370 GEM SHMEM Helper Function Reference
371 -----------------------------------
373 .. kernel-doc:: drivers/gpu/drm/drm_gem_shmem_helper.c
376 .. kernel-doc:: include/drm/drm_gem_shmem_helper.h
379 .. kernel-doc:: drivers/gpu/drm/drm_gem_shmem_helper.c
382 GEM VRAM Helper Functions Reference
383 -----------------------------------
385 .. kernel-doc:: drivers/gpu/drm/drm_gem_vram_helper.c
388 .. kernel-doc:: include/drm/drm_gem_vram_helper.h
391 .. kernel-doc:: drivers/gpu/drm/drm_gem_vram_helper.c
394 GEM TTM Helper Functions Reference
395 -----------------------------------
397 .. kernel-doc:: drivers/gpu/drm/drm_gem_ttm_helper.c
400 .. kernel-doc:: drivers/gpu/drm/drm_gem_ttm_helper.c
406 .. kernel-doc:: drivers/gpu/drm/drm_vma_manager.c
407 :doc: vma offset manager
409 .. kernel-doc:: include/drm/drm_vma_manager.h
412 .. kernel-doc:: drivers/gpu/drm/drm_vma_manager.c
415 .. _prime_buffer_sharing:
420 PRIME is the cross device buffer sharing framework in drm, originally
421 created for the OPTIMUS range of multi-gpu platforms. To userspace PRIME
422 buffers are dma-buf based file descriptors.
424 Overview and Lifetime Rules
425 ---------------------------
427 .. kernel-doc:: drivers/gpu/drm/drm_prime.c
428 :doc: overview and lifetime rules
430 PRIME Helper Functions
431 ----------------------
433 .. kernel-doc:: drivers/gpu/drm/drm_prime.c
436 PRIME Function References
437 -------------------------
439 .. kernel-doc:: include/drm/drm_prime.h
442 .. kernel-doc:: drivers/gpu/drm/drm_prime.c
445 DRM MM Range Allocator
446 ======================
451 .. kernel-doc:: drivers/gpu/drm/drm_mm.c
454 LRU Scan/Eviction Support
455 -------------------------
457 .. kernel-doc:: drivers/gpu/drm/drm_mm.c
458 :doc: lru scan roster
460 DRM MM Range Allocator Function References
461 ------------------------------------------
463 .. kernel-doc:: include/drm/drm_mm.h
466 .. kernel-doc:: drivers/gpu/drm/drm_mm.c
469 DRM Cache Handling and Fast WC memcpy()
470 =======================================
472 .. kernel-doc:: drivers/gpu/drm/drm_cache.c
476 ===========================
478 .. kernel-doc:: drivers/gpu/drm/drm_syncobj.c
481 .. kernel-doc:: include/drm/drm_syncobj.h
484 .. kernel-doc:: drivers/gpu/drm/drm_syncobj.c
493 .. kernel-doc:: drivers/gpu/drm/scheduler/sched_main.c
496 Scheduler Function References
497 -----------------------------
499 .. kernel-doc:: include/drm/gpu_scheduler.h
502 .. kernel-doc:: drivers/gpu/drm/scheduler/sched_main.c
505 .. kernel-doc:: drivers/gpu/drm/scheduler/sched_entity.c