- NetBSD Manual Pages
SHMIF(4) NetBSD Kernel Interfaces Manual SHMIF(4)
Powered by man-cgi (2021-06-01).
Maintained for NetBSD
by Kimmo Suominen.
Based on man-cgi by Panagiotis Christias.
shmif -- rump kernel shared memory network interface
rump_pub_shmif_create(const char *path, int *ifnum);
The shmif interface uses a memory mapped regular file as a virtual Ether-
net bus. All interfaces connected to the same bus see each others' traf-
Using a memory mapped regular file as a bus has two implications:
1. The bus identifier is not in flat global namespace.
2. Configuring and using the interface is possible without superuser
privileges on the host (normal host file access permissions for the
It is not possible to directly access the host networking facilities from
a rump kernel using purely shmif. However, traffic can be routed to
another rump kernel instance which provides both shmif and virt(4) net-
An shmif interface can be created in two ways:
· Programmatically by calling rump_pub_shmif_create(). The bus path-
name is passed in path. The number of the newly created interface is
available after a successful call by dereferencing ifnum.
· Dynamically at runtime with ifconfig(8) or equivalent using the
create command. In this case the bus path must be configured with
ifconfig(8) linkstr before the interface address can be configured.
Destroying an shmif interface is possible only via ifconfig(8) destroy.
An shmif interface emulates TX/RX offload options in software. They are
specified by ifconfig(8). Alternatively, its if_capenable flag is
directly specified by environment variable RUMP_SHMIF_CAPENABLE, for
0x7ff80 for all TX/RX offload
0x6aa80 for all TX offload
0x15500 for all RX offload
See /usr/include/net/if.h for more details.
rump(3), virt(4), ifconfig(8)
NetBSD 9.2 December 12, 2018 NetBSD 9.2