1 .. SPDX-License-Identifier: GPL-2.0
3 ===========================================
4 Shared Virtual Addressing (SVA) with ENQCMD
5 ===========================================
10 Shared Virtual Addressing (SVA) allows the processor and device to use the
11 same virtual addresses avoiding the need for software to translate virtual
12 addresses to physical addresses. SVA is what PCIe calls Shared Virtual
15 In addition to the convenience of using application virtual addresses
16 by the device, it also doesn't require pinning pages for DMA.
17 PCIe Address Translation Services (ATS) along with Page Request Interface
18 (PRI) allow devices to function much the same way as the CPU handling
19 application page-faults. For more information please refer to the PCIe
20 specification Chapter 10: ATS Specification.
22 Use of SVA requires IOMMU support in the platform. IOMMU is also
23 required to support the PCIe features ATS and PRI. ATS allows devices
24 to cache translations for virtual addresses. The IOMMU driver uses the
25 mmu_notifier() support to keep the device TLB cache and the CPU cache in
26 sync. When an ATS lookup fails for a virtual address, the device should
27 use the PRI in order to request the virtual address to be paged into the
28 CPU page tables. The device must use ATS again in order the fetch the
29 translation before use.
31 Shared Hardware Workqueues
32 ==========================
34 Unlike Single Root I/O Virtualization (SR-IOV), Scalable IOV (SIOV) permits
35 the use of Shared Work Queues (SWQ) by both applications and Virtual
36 Machines (VM's). This allows better hardware utilization vs. hard
37 partitioning resources that could result in under utilization. In order to
38 allow the hardware to distinguish the context for which work is being
39 executed in the hardware by SWQ interface, SIOV uses Process Address Space
40 ID (PASID), which is a 20-bit number defined by the PCIe SIG.
42 PASID value is encoded in all transactions from the device. This allows the
43 IOMMU to track I/O on a per-PASID granularity in addition to using the PCIe
44 Resource Identifier (RID) which is the Bus/Device/Function.
50 ENQCMD is a new instruction on Intel platforms that atomically submits a
51 work descriptor to a device. The descriptor includes the operation to be
52 performed, virtual addresses of all parameters, virtual address of a completion
53 record, and the PASID (process address space ID) of the current process.
55 ENQCMD works with non-posted semantics and carries a status back if the
56 command was accepted by hardware. This allows the submitter to know if the
57 submission needs to be retried or other device specific mechanisms to
58 implement fairness or ensure forward progress should be provided.
60 ENQCMD is the glue that ensures applications can directly submit commands
61 to the hardware and also permits hardware to be aware of application context
62 to perform I/O operations via use of PASID.
64 Process Address Space Tagging
65 =============================
67 A new thread-scoped MSR (IA32_PASID) provides the connection between
68 user processes and the rest of the hardware. When an application first
69 accesses an SVA-capable device, this MSR is initialized with a newly
70 allocated PASID. The driver for the device calls an IOMMU-specific API
71 that sets up the routing for DMA and page-requests.
73 For example, the Intel Data Streaming Accelerator (DSA) uses
74 iommu_sva_bind_device(), which will do the following:
76 - Allocate the PASID, and program the process page-table (%cr3 register) in the
77 PASID context entries.
78 - Register for mmu_notifier() to track any page-table invalidations to keep
79 the device TLB in sync. For example, when a page-table entry is invalidated,
80 the IOMMU propagates the invalidation to the device TLB. This will force any
81 future access by the device to this virtual address to participate in
82 ATS. If the IOMMU responds with proper response that a page is not
83 present, the device would request the page to be paged in via the PCIe PRI
84 protocol before performing I/O.
86 This MSR is managed with the XSAVE feature set as "supervisor state" to
87 ensure the MSR is updated during context switch.
92 The kernel must allocate a PASID on behalf of each process which will use
93 ENQCMD and program it into the new MSR to communicate the process identity to
94 platform hardware. ENQCMD uses the PASID stored in this MSR to tag requests
95 from this process. When a user submits a work descriptor to a device using the
96 ENQCMD instruction, the PASID field in the descriptor is auto-filled with the
97 value from MSR_IA32_PASID. Requests for DMA from the device are also tagged
98 with the same PASID. The platform IOMMU uses the PASID in the transaction to
99 perform address translation. The IOMMU APIs setup the corresponding PASID
100 entry in IOMMU with the process address used by the CPU (e.g. %cr3 register in
103 The MSR must be configured on each logical CPU before any application
104 thread can interact with a device. Threads that belong to the same
105 process share the same page tables, thus the same MSR value.
107 PASID is cleared when a process is created. The PASID allocation and MSR
108 programming may occur long after a process and its threads have been created.
109 One thread must call iommu_sva_bind_device() to allocate the PASID for the
110 process. If a thread uses ENQCMD without the MSR first being populated, a #GP
111 will be raised. The kernel will update the PASID MSR with the PASID for all
112 threads in the process. A single process PASID can be used simultaneously
113 with multiple devices since they all share the same address space.
115 One thread can call iommu_sva_unbind_device() to free the allocated PASID.
116 The kernel will clear the PASID MSR for all threads belonging to the process.
118 New threads inherit the MSR value from the parent.
123 * Each process has many threads, but only one PASID.
124 * Devices have a limited number (~10's to 1000's) of hardware workqueues.
125 The device driver manages allocating hardware workqueues.
126 * A single mmap() maps a single hardware workqueue as a "portal" and
127 each portal maps down to a single workqueue.
128 * For each device with which a process interacts, there must be
129 one or more mmap()'d portals.
130 * Many threads within a process can share a single portal to access
132 * Multiple processes can separately mmap() the same portal, in
133 which case they still share one device hardware workqueue.
134 * The single process-wide PASID is used by all threads to interact
135 with all devices. There is not, for instance, a PASID for each
136 thread or each thread<->device pair.
143 Shared Virtual Addressing (SVA) permits I/O hardware and the processor to
144 work in the same address space, i.e., to share it. Some call it Shared
145 Virtual Memory (SVM), but Linux community wanted to avoid confusing it with
146 POSIX Shared Memory and Secure Virtual Machines which were terms already in
151 A Process Address Space ID (PASID) is a PCIe-defined Transaction Layer Packet
152 (TLP) prefix. A PASID is a 20-bit number allocated and managed by the OS.
153 PASID is included in all transactions between the platform and the device.
155 * How are shared workqueues different?
157 Traditionally, in order for userspace applications to interact with hardware,
158 there is a separate hardware instance required per process. For example,
159 consider doorbells as a mechanism of informing hardware about work to process.
160 Each doorbell is required to be spaced 4k (or page-size) apart for process
161 isolation. This requires hardware to provision that space and reserve it in
162 MMIO. This doesn't scale as the number of threads becomes quite large. The
163 hardware also manages the queue depth for Shared Work Queues (SWQ), and
164 consumers don't need to track queue depth. If there is no space to accept
165 a command, the device will return an error indicating retry.
167 A user should check Deferrable Memory Write (DMWr) capability on the device
168 and only submits ENQCMD when the device supports it. In the new DMWr PCIe
169 terminology, devices need to support DMWr completer capability. In addition,
170 it requires all switch ports to support DMWr routing and must be enabled by
171 the PCIe subsystem, much like how PCIe atomic operations are managed for
174 SWQ allows hardware to provision just a single address in the device. When
175 used with ENQCMD to submit work, the device can distinguish the process
176 submitting the work since it will include the PASID assigned to that
177 process. This helps the device scale to a large number of processes.
179 * Is this the same as a user space device driver?
181 Communicating with the device via the shared workqueue is much simpler
182 than a full blown user space driver. The kernel driver does all the
183 initialization of the hardware. User space only needs to worry about
184 submitting work and processing completions.
186 * Is this the same as SR-IOV?
188 Single Root I/O Virtualization (SR-IOV) focuses on providing independent
189 hardware interfaces for virtualizing hardware. Hence, it's required to be
190 almost fully functional interface to software supporting the traditional
191 BARs, space for interrupts via MSI-X, its own register layout.
192 Virtual Functions (VFs) are assisted by the Physical Function (PF)
195 Scalable I/O Virtualization builds on the PASID concept to create device
196 instances for virtualization. SIOV requires host software to assist in
197 creating virtual devices; each virtual device is represented by a PASID
198 along with the bus/device/function of the device. This allows device
199 hardware to optimize device resource creation and can grow dynamically on
200 demand. SR-IOV creation and management is very static in nature. Consult
201 references below for more details.
203 * Why not just create a virtual function for each app?
205 Creating PCIe SR-IOV type Virtual Functions (VF) is expensive. VFs require
206 duplicated hardware for PCI config space and interrupts such as MSI-X.
207 Resources such as interrupts have to be hard partitioned between VFs at
208 creation time, and cannot scale dynamically on demand. The VFs are not
209 completely independent from the Physical Function (PF). Most VFs require
210 some communication and assistance from the PF driver. SIOV, in contrast,
211 creates a software-defined device where all the configuration and control
212 aspects are mediated via the slow path. The work submission and completion
213 happen without any mediation.
215 * Does this support virtualization?
217 ENQCMD can be used from within a guest VM. In these cases, the VMM helps
218 with setting up a translation table to translate from Guest PASID to Host
219 PASID. Please consult the ENQCMD instruction set reference for more
222 * Does memory need to be pinned?
224 When devices support SVA along with platform hardware such as IOMMU
225 supporting such devices, there is no need to pin memory for DMA purposes.
226 Devices that support SVA also support other PCIe features that remove the
227 pinning requirement for memory.
229 Device TLB support - Device requests the IOMMU to lookup an address before
230 use via Address Translation Service (ATS) requests. If the mapping exists
231 but there is no page allocated by the OS, IOMMU hardware returns that no
234 Device requests the virtual address to be mapped via Page Request
235 Interface (PRI). Once the OS has successfully completed the mapping, it
236 returns the response back to the device. The device requests again for
237 a translation and continues.
239 IOMMU works with the OS in managing consistency of page-tables with the
240 device. When removing pages, it interacts with the device to remove any
241 device TLB entry that might have been cached before removing the mappings from
248 https://01.org/blogs/ashokraj/2018/recent-enhancements-intel-virtualization-technology-directed-i/o-intel-vt-d
251 https://01.org/blogs/2019/assignable-interfaces-intel-scalable-i/o-virtualization-linux
254 https://software.intel.com/sites/default/files/managed/c5/15/architecture-instruction-set-extensions-programming-reference.pdf
257 https://software.intel.com/sites/default/files/341204-intel-data-streaming-accelerator-spec.pdf