Depending on the device and other considerations, you may use the mmap() function to map device registers into the address space of your process (see the mmap(2) reference page). When the kernel maps a device address into process space, it does it using the TLB mechanism. From mmap() you receive a valid address in process space. This address is mapped through a TLB entry to an address in segment that accesses uncached physical memory. When your program refers to this address, the reference is directed to the system bus and the device.
Portions of kernel virtual memory (kseg0 or xkseg) can be accessed from a user process. Access is based on the use of device special files (see the mem(7) reference page). Access is done using two models, a device model and a memory map model.
The device special file /dev/kmem represents kernel virtual memory (kseg0 or xkseg). It can be opened, read and written similarly to /dev/mem. Clearly both of these devices should have file permissions that restrict their use even for input.
The /dev/kmem device, representing kernel virtual memory, cannot be used with mmap(). However, a third device special, /dev/mmem (note the double "m"), represents access to only those addresses that are configured in the file /var/sysgen/master.d/mem. As distributed, this file is configured to allow access to the free-running timer device and, in some systems, to graphics hardware.
For an example of mapped access to physical memory, see the example code in the syssgi(2) reference page related to the SGI_QUERY_CYCLECNTR option. In this operation, the address of the timer (a device register) is mapped into the process's address space using a TLB entry. When the user process accesses the mapped address, the TLB entry converts it to an address in kseg1/xkphys, which then bypasses the cache.