xref: /qemu/docs/system/devices/ivshmem.rst (revision 5db05230)
1Inter-VM Shared Memory device
2-----------------------------
3
4On Linux hosts, a shared memory device is available. The basic syntax
5is:
6
7.. parsed-literal::
8
9   |qemu_system_x86| -device ivshmem-plain,memdev=hostmem
10
11where hostmem names a host memory backend. For a POSIX shared memory
12backend, use something like
13
14::
15
16   -object memory-backend-file,size=1M,share,mem-path=/dev/shm/ivshmem,id=hostmem
17
18If desired, interrupts can be sent between guest VMs accessing the same
19shared memory region. Interrupt support requires using a shared memory
20server and using a chardev socket to connect to it. The code for the
21shared memory server is qemu.git/contrib/ivshmem-server. An example
22syntax when using the shared memory server is:
23
24.. parsed-literal::
25
26   # First start the ivshmem server once and for all
27   ivshmem-server -p pidfile -S path -m shm-name -l shm-size -n vectors
28
29   # Then start your qemu instances with matching arguments
30   |qemu_system_x86| -device ivshmem-doorbell,vectors=vectors,chardev=id
31                    -chardev socket,path=path,id=id
32
33When using the server, the guest will be assigned a VM ID (>=0) that
34allows guests using the same server to communicate via interrupts.
35Guests can read their VM ID from a device register (see
36:doc:`../../specs/ivshmem-spec`).
37
38Migration with ivshmem
39~~~~~~~~~~~~~~~~~~~~~~
40
41With device property ``master=on``, the guest will copy the shared
42memory on migration to the destination host. With ``master=off``, the
43guest will not be able to migrate with the device attached. In the
44latter case, the device should be detached and then reattached after
45migration using the PCI hotplug support.
46
47At most one of the devices sharing the same memory can be master. The
48master must complete migration before you plug back the other devices.
49
50ivshmem and hugepages
51~~~~~~~~~~~~~~~~~~~~~
52
53Instead of specifying the <shm size> using POSIX shm, you may specify a
54memory backend that has hugepage support:
55
56.. parsed-literal::
57
58   |qemu_system_x86| -object memory-backend-file,size=1G,mem-path=/dev/hugepages/my-shmem-file,share,id=mb1
59                    -device ivshmem-plain,memdev=mb1
60
61ivshmem-server also supports hugepages mount points with the ``-m``
62memory path argument.
63