xref: /linux/Documentation/gpu/nova/core/todo.rst (revision 69050f8d6d075dc01af7a5f2f550a8067510366f)
1.. SPDX-License-Identifier: (GPL-2.0+ OR MIT)
2
3=========
4Task List
5=========
6
7Tasks may have the following fields:
8
9- ``Complexity``: Describes the required familiarity with Rust and / or the
10  corresponding kernel APIs or subsystems. There are four different complexities,
11  ``Beginner``, ``Intermediate``, ``Advanced`` and ``Expert``.
12- ``Reference``: References to other tasks.
13- ``Link``: Links to external resources.
14- ``Contact``: The person that can be contacted for further information about
15  the task.
16
17A task might have `[ABCD]` code after its name. This code can be used to grep
18into the code for `TODO` entries related to it.
19
20Enablement (Rust)
21=================
22
23Tasks that are not directly related to nova-core, but are preconditions in terms
24of required APIs.
25
26FromPrimitive API [FPRI]
27------------------------
28
29Sometimes the need arises to convert a number to a value of an enum or a
30structure.
31
32A good example from nova-core would be the ``Chipset`` enum type, which defines
33the value ``AD102``. When probing the GPU the value ``0x192`` can be read from a
34certain register indication the chipset AD102. Hence, the enum value ``AD102``
35should be derived from the number ``0x192``. Currently, nova-core uses a custom
36implementation (``Chipset::from_u32`` for this.
37
38Instead, it would be desirable to have something like the ``FromPrimitive``
39trait [1] from the num crate.
40
41Having this generalization also helps with implementing a generic macro that
42automatically generates the corresponding mappings between a value and a number.
43
44FromPrimitive support has been worked on in the past, but hasn't been followed
45since then [1].
46
47There also have been considerations of ToPrimitive [2].
48
49| Complexity: Beginner
50| Link: https://docs.rs/num/latest/num/trait.FromPrimitive.html
51| Link: https://lore.kernel.org/all/cover.1750689857.git.y.j3ms.n@gmail.com/ [1]
52| Link: https://rust-for-linux.zulipchat.com/#narrow/channel/288089-General/topic/Implement.20.60FromPrimitive.60.20trait.20.2B.20derive.20macro.20for.20nova-core/with/541971854 [2]
53
54Generic register abstraction [REGA]
55-----------------------------------
56
57Work out how register constants and structures can be automatically generated
58through generalized macros.
59
60Example:
61
62.. code-block:: rust
63
64	register!(BOOT0, 0x0, u32, pci::Bar<SIZE>, Fields [
65	   MINOR_REVISION(3:0, RO),
66	   MAJOR_REVISION(7:4, RO),
67	   REVISION(7:0, RO), // Virtual register combining major and minor rev.
68	])
69
70This could expand to something like:
71
72.. code-block:: rust
73
74	const BOOT0_OFFSET: usize = 0x00000000;
75	const BOOT0_MINOR_REVISION_SHIFT: u8 = 0;
76	const BOOT0_MINOR_REVISION_MASK: u32 = 0x0000000f;
77	const BOOT0_MAJOR_REVISION_SHIFT: u8 = 4;
78	const BOOT0_MAJOR_REVISION_MASK: u32 = 0x000000f0;
79	const BOOT0_REVISION_SHIFT: u8 = BOOT0_MINOR_REVISION_SHIFT;
80	const BOOT0_REVISION_MASK: u32 = BOOT0_MINOR_REVISION_MASK | BOOT0_MAJOR_REVISION_MASK;
81
82	struct Boot0(u32);
83
84	impl Boot0 {
85	   #[inline]
86	   fn read(bar: &RevocableGuard<'_, pci::Bar<SIZE>>) -> Self {
87	      Self(bar.readl(BOOT0_OFFSET))
88	   }
89
90	   #[inline]
91	   fn minor_revision(&self) -> u32 {
92	      (self.0 & BOOT0_MINOR_REVISION_MASK) >> BOOT0_MINOR_REVISION_SHIFT
93	   }
94
95	   #[inline]
96	   fn major_revision(&self) -> u32 {
97	      (self.0 & BOOT0_MAJOR_REVISION_MASK) >> BOOT0_MAJOR_REVISION_SHIFT
98	   }
99
100	   #[inline]
101	   fn revision(&self) -> u32 {
102	      (self.0 & BOOT0_REVISION_MASK) >> BOOT0_REVISION_SHIFT
103	   }
104	}
105
106Usage:
107
108.. code-block:: rust
109
110	let bar = bar.try_access().ok_or(ENXIO)?;
111
112	let boot0 = Boot0::read(&bar);
113	pr_info!("Revision: {}\n", boot0.revision());
114
115A work-in-progress implementation currently resides in
116`drivers/gpu/nova-core/regs/macros.rs` and is used in nova-core. It would be
117nice to improve it (possibly using proc macros) and move it to the `kernel`
118crate so it can be used by other components as well.
119
120Features desired before this happens:
121
122* Make I/O optional I/O (for field values that are not registers),
123* Support other sizes than `u32`,
124* Allow visibility control for registers and individual fields,
125* Use Rust slice syntax to express fields ranges.
126
127| Complexity: Advanced
128| Contact: Alexandre Courbot
129
130Numerical operations [NUMM]
131---------------------------
132
133Nova uses integer operations that are not part of the standard library (or not
134implemented in an optimized way for the kernel). These include:
135
136- The "Find Last Set Bit" (`fls` function of the C part of the kernel)
137  operation.
138
139A `num` core kernel module is being designed to provide these operations.
140
141| Complexity: Intermediate
142| Contact: Alexandre Courbot
143
144Page abstraction for foreign pages
145----------------------------------
146
147Rust abstractions for pages not created by the Rust page abstraction without
148direct ownership.
149
150There is active onging work from Abdiel Janulgue [1] and Lina [2].
151
152| Complexity: Advanced
153| Link: https://lore.kernel.org/linux-mm/20241119112408.779243-1-abdiel.janulgue@gmail.com/ [1]
154| Link: https://lore.kernel.org/rust-for-linux/20250202-rust-page-v1-0-e3170d7fe55e@asahilina.net/ [2]
155
156PCI MISC APIs
157-------------
158
159Extend the existing PCI device / driver abstractions by SR-IOV, capability, MSI
160API abstractions.
161
162SR-IOV [1] is work in progress.
163
164| Complexity: Beginner
165| Link: https://lore.kernel.org/all/20251119-rust-pci-sriov-v1-0-883a94599a97@redhat.com/ [1]
166
167GPU (general)
168=============
169
170Initial Devinit support
171-----------------------
172
173Implement BIOS Device Initialization, i.e. memory sizing, waiting, PLL
174configuration.
175
176| Contact: Dave Airlie
177| Complexity: Beginner
178
179MMU / PT management
180-------------------
181
182Work out the architecture for MMU / page table management.
183
184We need to consider that nova-drm will need rather fine-grained control,
185especially in terms of locking, in order to be able to implement asynchronous
186Vulkan queues.
187
188While generally sharing the corresponding code is desirable, it needs to be
189evaluated how (and if at all) sharing the corresponding code is expedient.
190
191| Complexity: Expert
192
193VRAM memory allocator
194---------------------
195
196Investigate options for a VRAM memory allocator.
197
198Some possible options:
199  - Rust abstractions for
200    - RB tree (interval tree) / drm_mm
201    - maple_tree
202  - native Rust collections
203
204There is work in progress for using drm_buddy [1].
205
206| Complexity: Advanced
207| Link: https://lore.kernel.org/all/20251219203805.1246586-4-joelagnelf@nvidia.com/ [1]
208
209Instance Memory
210---------------
211
212Implement support for instmem (bar2) used to store page tables.
213
214| Complexity: Intermediate
215| Contact: Dave Airlie
216
217GPU System Processor (GSP)
218==========================
219
220Export GSP log buffers
221----------------------
222
223Recent patches from Timur Tabi [1] added support to expose GSP-RM log buffers
224(even after failure to probe the driver) through debugfs.
225
226This is also an interesting feature for nova-core, especially in the early days.
227
228| Link: https://lore.kernel.org/nouveau/20241030202952.694055-2-ttabi@nvidia.com/ [1]
229| Reference: Debugfs abstractions
230| Complexity: Intermediate
231
232GSP firmware abstraction
233------------------------
234
235The GSP-RM firmware API is unstable and may incompatibly change from version to
236version, in terms of data structures and semantics.
237
238This problem is one of the big motivations for using Rust for nova-core, since
239it turns out that Rust's procedural macro feature provides a rather elegant way
240to address this issue:
241
2421. generate Rust structures from the C headers in a separate namespace per version
2432. build abstraction structures (within a generic namespace) that implement the
244   firmware interfaces; annotate the differences in implementation with version
245   identifiers
2463. use a procedural macro to generate the actual per version implementation out
247   of this abstraction
2484. instantiate the correct version type one on runtime (can be sure that all
249   have the same interface because it's defined by a common trait)
250
251There is a PoC implementation of this pattern, in the context of the nova-core
252PoC driver.
253
254This task aims at refining the feature and ideally generalize it, to be usable
255by other drivers as well.
256
257| Complexity: Expert
258
259GSP message queue
260-----------------
261
262Implement low level GSP message queue (command, status) for communication
263between the kernel driver and GSP.
264
265| Complexity: Advanced
266| Contact: Dave Airlie
267
268Bootstrap GSP
269-------------
270
271Call the boot firmware to boot the GSP processor; execute initial control
272messages.
273
274| Complexity: Intermediate
275| Contact: Dave Airlie
276
277Client / Device APIs
278--------------------
279
280Implement the GSP message interface for client / device allocation and the
281corresponding client and device allocation APIs.
282
283| Complexity: Intermediate
284| Contact: Dave Airlie
285
286Bar PDE handling
287----------------
288
289Synchronize page table handling for BARs between the kernel driver and GSP.
290
291| Complexity: Beginner
292| Contact: Dave Airlie
293
294FIFO engine
295-----------
296
297Implement support for the FIFO engine, i.e. the corresponding GSP message
298interface and provide an API for chid allocation and channel handling.
299
300| Complexity: Advanced
301| Contact: Dave Airlie
302
303GR engine
304---------
305
306Implement support for the graphics engine, i.e. the corresponding GSP message
307interface and provide an API for (golden) context creation and promotion.
308
309| Complexity: Advanced
310| Contact: Dave Airlie
311
312CE engine
313---------
314
315Implement support for the copy engine, i.e. the corresponding GSP message
316interface.
317
318| Complexity: Intermediate
319| Contact: Dave Airlie
320
321VFN IRQ controller
322------------------
323
324Support for the VFN interrupt controller.
325
326| Complexity: Intermediate
327| Contact: Dave Airlie
328
329External APIs
330=============
331
332nova-core base API
333------------------
334
335Work out the common pieces of the API to connect 2nd level drivers, i.e. vGPU
336manager and nova-drm.
337
338| Complexity: Advanced
339
340vGPU manager API
341----------------
342
343Work out the API parts required by the vGPU manager, which are not covered by
344the base API.
345
346| Complexity: Advanced
347
348nova-core C API
349---------------
350
351Implement a C wrapper for the APIs required by the vGPU manager driver.
352
353| Complexity: Intermediate
354
355Testing
356=======
357
358CI pipeline
359-----------
360
361Investigate option for continuous integration testing.
362
363This can go from as simple as running KUnit tests over running (graphics) CTS to
364booting up (multiple) guest VMs to test VFIO use-cases.
365
366It might also be worth to consider the introduction of a new test suite directly
367sitting on top of the uAPI for more targeted testing and debugging. There may be
368options for collaboration / shared code with the Mesa project.
369
370| Complexity: Advanced
371