1==========================================
2I915 VM_BIND feature design and use cases
3==========================================
4
5VM_BIND feature
6================
7DRM_I915_GEM_VM_BIND/UNBIND ioctls allows UMD to bind/unbind GEM buffer
8objects (BOs) or sections of a BOs at specified GPU virtual addresses on a
9specified address space (VM). These mappings (also referred to as persistent
10mappings) will be persistent across multiple GPU submissions (execbuf calls)
11issued by the UMD, without user having to provide a list of all required
12mappings during each submission (as required by older execbuf mode).
13
14The VM_BIND/UNBIND calls allow UMDs to request a timeline out fence for
15signaling the completion of bind/unbind operation.
16
17VM_BIND feature is advertised to user via I915_PARAM_VM_BIND_VERSION.
18User has to opt-in for VM_BIND mode of binding for an address space (VM)
19during VM creation time via I915_VM_CREATE_FLAGS_USE_VM_BIND extension.
20
21VM_BIND/UNBIND ioctl calls executed on different CPU threads concurrently are
22not ordered. Furthermore, parts of the VM_BIND/UNBIND operations can be done
23asynchronously, when valid out fence is specified.
24
25VM_BIND features include:
26
27* Multiple Virtual Address (VA) mappings can map to the same physical pages
28  of an object (aliasing).
29* VA mapping can map to a partial section of the BO (partial binding).
30* Support capture of persistent mappings in the dump upon GPU error.
31* Support for userptr gem objects (no special uapi is required for this).
32
33TLB flush consideration
34------------------------
35The i915 driver flushes the TLB for each submission and when an object's
36pages are released. The VM_BIND/UNBIND operation will not do any additional
37TLB flush. Any VM_BIND mapping added will be in the working set for subsequent
38submissions on that VM and will not be in the working set for currently running
39batches (which would require additional TLB flushes, which is not supported).
40
41Execbuf ioctl in VM_BIND mode
42-------------------------------
43A VM in VM_BIND mode will not support older execbuf mode of binding.
44The execbuf ioctl handling in VM_BIND mode differs significantly from the
45older execbuf2 ioctl (See struct drm_i915_gem_execbuffer2).
46Hence, a new execbuf3 ioctl has been added to support VM_BIND mode. (See
47struct drm_i915_gem_execbuffer3). The execbuf3 ioctl will not accept any
48execlist. Hence, no support for implicit sync. It is expected that the below
49work will be able to support requirements of object dependency setting in all
50use cases:
51
52"dma-buf: Add an API for exporting sync files"
53(https://lwn.net/Articles/859290/)
54
55The new execbuf3 ioctl only works in VM_BIND mode and the VM_BIND mode only
56works with execbuf3 ioctl for submission. All BOs mapped on that VM (through
57VM_BIND call) at the time of execbuf3 call are deemed required for that
58submission.
59
60The execbuf3 ioctl directly specifies the batch addresses instead of as
61object handles as in execbuf2 ioctl. The execbuf3 ioctl will also not
62support many of the older features like in/out/submit fences, fence array,
63default gem context and many more (See struct drm_i915_gem_execbuffer3).
64
65In VM_BIND mode, VA allocation is completely managed by the user instead of
66the i915 driver. Hence all VA assignment, eviction are not applicable in
67VM_BIND mode. Also, for determining object activeness, VM_BIND mode will not
68be using the i915_vma active reference tracking. It will instead use dma-resv
69object for that (See `VM_BIND dma_resv usage`_).
70
71So, a lot of existing code supporting execbuf2 ioctl, like relocations, VA
72evictions, vma lookup table, implicit sync, vma active reference tracking etc.,
73are not applicable for execbuf3 ioctl. Hence, all execbuf3 specific handling
74should be in a separate file and only functionalities common to these ioctls
75can be the shared code where possible.
76
77VM_PRIVATE objects
78-------------------
79By default, BOs can be mapped on multiple VMs and can also be dma-buf
80exported. Hence these BOs are referred to as Shared BOs.
81During each execbuf submission, the request fence must be added to the
82dma-resv fence list of all shared BOs mapped on the VM.
83
84VM_BIND feature introduces an optimization where user can create BO which
85is private to a specified VM via I915_GEM_CREATE_EXT_VM_PRIVATE flag during
86BO creation. Unlike Shared BOs, these VM private BOs can only be mapped on
87the VM they are private to and can't be dma-buf exported.
88All private BOs of a VM share the dma-resv object. Hence during each execbuf
89submission, they need only one dma-resv fence list updated. Thus, the fast
90path (where required mappings are already bound) submission latency is O(1)
91w.r.t the number of VM private BOs.
92
93VM_BIND locking hierarchy
94-------------------------
95The locking design here supports the older (execlist based) execbuf mode, the
96newer VM_BIND mode, the VM_BIND mode with GPU page faults and possible future
97system allocator support (See `Shared Virtual Memory (SVM) support`_).
98The older execbuf mode and the newer VM_BIND mode without page faults manages
99residency of backing storage using dma_fence. The VM_BIND mode with page faults
100and the system allocator support do not use any dma_fence at all.
101
102VM_BIND locking order is as below.
103
1041) Lock-A: A vm_bind mutex will protect vm_bind lists. This lock is taken in
105   vm_bind/vm_unbind ioctl calls, in the execbuf path and while releasing the
106   mapping.
107
108   In future, when GPU page faults are supported, we can potentially use a
109   rwsem instead, so that multiple page fault handlers can take the read side
110   lock to lookup the mapping and hence can run in parallel.
111   The older execbuf mode of binding do not need this lock.
112
1132) Lock-B: The object's dma-resv lock will protect i915_vma state and needs to
114   be held while binding/unbinding a vma in the async worker and while updating
115   dma-resv fence list of an object. Note that private BOs of a VM will all
116   share a dma-resv object.
117
118   The future system allocator support will use the HMM prescribed locking
119   instead.
120
1213) Lock-C: Spinlock/s to protect some of the VM's lists like the list of
122   invalidated vmas (due to eviction and userptr invalidation) etc.
123
124When GPU page faults are supported, the execbuf path do not take any of these
125locks. There we will simply smash the new batch buffer address into the ring and
126then tell the scheduler run that. The lock taking only happens from the page
127fault handler, where we take lock-A in read mode, whichever lock-B we need to
128find the backing storage (dma_resv lock for gem objects, and hmm/core mm for
129system allocator) and some additional locks (lock-D) for taking care of page
130table races. Page fault mode should not need to ever manipulate the vm lists,
131so won't ever need lock-C.
132
133VM_BIND LRU handling
134---------------------
135We need to ensure VM_BIND mapped objects are properly LRU tagged to avoid
136performance degradation. We will also need support for bulk LRU movement of
137VM_BIND objects to avoid additional latencies in execbuf path.
138
139The page table pages are similar to VM_BIND mapped objects (See
140`Evictable page table allocations`_) and are maintained per VM and needs to
141be pinned in memory when VM is made active (ie., upon an execbuf call with
142that VM). So, bulk LRU movement of page table pages is also needed.
143
144VM_BIND dma_resv usage
145-----------------------
146Fences needs to be added to all VM_BIND mapped objects. During each execbuf
147submission, they are added with DMA_RESV_USAGE_BOOKKEEP usage to prevent
148over sync (See enum dma_resv_usage). One can override it with either
149DMA_RESV_USAGE_READ or DMA_RESV_USAGE_WRITE usage during explicit object
150dependency setting.
151
152Note that DRM_I915_GEM_WAIT and DRM_I915_GEM_BUSY ioctls do not check for
153DMA_RESV_USAGE_BOOKKEEP usage and hence should not be used for end of batch
154check. Instead, the execbuf3 out fence should be used for end of batch check
155(See struct drm_i915_gem_execbuffer3).
156
157Also, in VM_BIND mode, use dma-resv apis for determining object activeness
158(See dma_resv_test_signaled() and dma_resv_wait_timeout()) and do not use the
159older i915_vma active reference tracking which is deprecated. This should be
160easier to get it working with the current TTM backend.
161
162Mesa use case
163--------------
164VM_BIND can potentially reduce the CPU overhead in Mesa (both Vulkan and Iris),
165hence improving performance of CPU-bound applications. It also allows us to
166implement Vulkan's Sparse Resources. With increasing GPU hardware performance,
167reducing CPU overhead becomes more impactful.
168
169
170Other VM_BIND use cases
171========================
172
173Long running Compute contexts
174------------------------------
175Usage of dma-fence expects that they complete in reasonable amount of time.
176Compute on the other hand can be long running. Hence it is appropriate for
177compute to use user/memory fence (See `User/Memory Fence`_) and dma-fence usage
178must be limited to in-kernel consumption only.
179
180Where GPU page faults are not available, kernel driver upon buffer invalidation
181will initiate a suspend (preemption) of long running context, finish the
182invalidation, revalidate the BO and then resume the compute context. This is
183done by having a per-context preempt fence which is enabled when someone tries
184to wait on it and triggers the context preemption.
185
186User/Memory Fence
187~~~~~~~~~~~~~~~~~~
188User/Memory fence is a <address, value> pair. To signal the user fence, the
189specified value will be written at the specified virtual address and wakeup the
190waiting process. User fence can be signaled either by the GPU or kernel async
191worker (like upon bind completion). User can wait on a user fence with a new
192user fence wait ioctl.
193
194Here is some prior work on this:
195https://patchwork.freedesktop.org/patch/349417/
196
197Low Latency Submission
198~~~~~~~~~~~~~~~~~~~~~~~
199Allows compute UMD to directly submit GPU jobs instead of through execbuf
200ioctl. This is made possible by VM_BIND is not being synchronized against
201execbuf. VM_BIND allows bind/unbind of mappings required for the directly
202submitted jobs.
203
204Debugger
205---------
206With debug event interface user space process (debugger) is able to keep track
207of and act upon resources created by another process (debugged) and attached
208to GPU via vm_bind interface.
209
210GPU page faults
211----------------
212GPU page faults when supported (in future), will only be supported in the
213VM_BIND mode. While both the older execbuf mode and the newer VM_BIND mode of
214binding will require using dma-fence to ensure residency, the GPU page faults
215mode when supported, will not use any dma-fence as residency is purely managed
216by installing and removing/invalidating page table entries.
217
218Page level hints settings
219--------------------------
220VM_BIND allows any hints setting per mapping instead of per BO. Possible hints
221include placement and atomicity. Sub-BO level placement hint will be even more
222relevant with upcoming GPU on-demand page fault support.
223
224Page level Cache/CLOS settings
225-------------------------------
226VM_BIND allows cache/CLOS settings per mapping instead of per BO.
227
228Evictable page table allocations
229---------------------------------
230Make pagetable allocations evictable and manage them similar to VM_BIND
231mapped objects. Page table pages are similar to persistent mappings of a
232VM (difference here are that the page table pages will not have an i915_vma
233structure and after swapping pages back in, parent page link needs to be
234updated).
235
236Shared Virtual Memory (SVM) support
237------------------------------------
238VM_BIND interface can be used to map system memory directly (without gem BO
239abstraction) using the HMM interface. SVM is only supported with GPU page
240faults enabled.
241
242VM_BIND UAPI
243=============
244
245.. kernel-doc:: Documentation/gpu/rfc/i915_vm_bind.h
246