xref: /openbmc/qemu/docs/specs/riscv-iommu.rst (revision b278b60d)
1.. _riscv-iommu:
2
3RISC-V IOMMU support for RISC-V machines
4========================================
5
6QEMU implements a RISC-V IOMMU emulation based on the RISC-V IOMMU spec
7version 1.0 `iommu1.0`_.
8
9The emulation includes a PCI reference device, riscv-iommu-pci, that QEMU
10RISC-V boards can use.  The 'virt' RISC-V machine is compatible with this
11device.
12
13riscv-iommu-pci reference device
14--------------------------------
15
16This device implements the RISC-V IOMMU emulation as recommended by the section
17"Integrating an IOMMU as a PCIe device" of `iommu1.0`_: a PCI device with base
18class 08h, sub-class 06h and programming interface 00h.
19
20As a reference device it doesn't implement anything outside of the specification,
21so it uses a generic default PCI ID given by QEMU: 1b36:0014.
22
23To include the device in the 'virt' machine:
24
25.. code-block:: bash
26
27  $ qemu-system-riscv64 -M virt -device riscv-iommu-pci,[optional_pci_opts] (...)
28
29This will add a RISC-V IOMMU PCI device in the board following any additional
30PCI parameters (like PCI bus address).  The behavior of the RISC-V IOMMU is
31defined by the spec but its operation is OS dependent.
32
33As of this writing the existing Linux kernel support `linux-v8`_, not yet merged,
34does not have support for features like VFIO passthrough.  The IOMMU emulation
35was tested using a public Ventana Micro Systems kernel repository in
36`ventana-linux`_.  This kernel is based on `linux-v8`_ with additional patches that
37enable features like KVM VFIO passthrough with irqbypass.  Until the kernel support
38is feature complete feel free to use the kernel available in the Ventana Micro Systems
39mirror.
40
41The current Linux kernel support will use the IOMMU device to create IOMMU groups
42with any eligible cards available in the system, regardless of factors such as the
43order in which the devices are added in the command line.
44
45This means that these command lines are equivalent as far as the current
46IOMMU kernel driver behaves:
47
48.. code-block:: bash
49
50  $ qemu-system-riscv64 \
51        -M virt,aia=aplic-imsic,aia-guests=5 \
52        -device riscv-iommu-pci,addr=1.0,vendor-id=0x1efd,device-id=0xedf1 \
53        -device e1000e,netdev=net1 -netdev user,id=net1,net=192.168.0.0/24 \
54        -device e1000e,netdev=net2 -netdev user,id=net2,net=192.168.200.0/24 \
55        (...)
56
57  $ qemu-system-riscv64 \
58        -M virt,aia=aplic-imsic,aia-guests=5 \
59        -device e1000e,netdev=net1 -netdev user,id=net1,net=192.168.0.0/24 \
60        -device e1000e,netdev=net2 -netdev user,id=net2,net=192.168.200.0/24 \
61        -device riscv-iommu-pci,addr=1.0,vendor-id=0x1efd,device-id=0xedf1 \
62        (...)
63
64Both will create iommu groups for the two e1000e cards.
65
66Another thing to notice on `linux-v8`_ and `ventana-linux`_ is that the kernel driver
67considers an IOMMU identified as a Rivos device, i.e. it uses Rivos vendor ID.  To
68use the riscv-iommu-pci device with the existing kernel support we need to emulate
69a Rivos PCI IOMMU by setting 'vendor-id' and 'device-id':
70
71.. code-block:: bash
72
73  $ qemu-system-riscv64 -M virt	\
74     -device riscv-iommu-pci,vendor-id=0x1efd,device-id=0xedf1 (...)
75
76Several options are available to control the capabilities of the device, namely:
77
78- "bus": the bus that the IOMMU device uses
79- "ioatc-limit": size of the Address Translation Cache (default to 2Mb)
80- "intremap": enable/disable MSI support
81- "ats": enable ATS support
82- "off" (Out-of-reset translation mode: 'on' for DMA disabled, 'off' for 'BARE' (passthrough))
83- "s-stage": enable s-stage support
84- "g-stage": enable g-stage support
85
86.. _iommu1.0: https://github.com/riscv-non-isa/riscv-iommu/releases/download/v1.0/riscv-iommu.pdf
87
88.. _linux-v8: https://lore.kernel.org/linux-riscv/cover.1718388908.git.tjeznach@rivosinc.com/
89
90.. _ventana-linux: https://github.com/ventanamicro/linux/tree/dev-upstream
91