GR740-UM-DS, Nov 2017, Version 1.7
124
www.cobham.com/gaisler
GR740
transitioned into its idle state, pending the completion of an earlier access, the delay suffered by an
access be longer than what is shown in the tables in this section. Accesses may also suffer increased
delays during collisions when the core has been instantiated to form a bi-directional bridge. Locked
accesses that abort on-going read operations will also mean additional delays.
Note that since the core has support for read and/or write combining, the number of cycles required
for the master will change depending on the access size and length of the incoming burst access.
12.3
General access protection and address translation
12.3.1 Overview
The core provides two types of access protection. The first option is to use a bit vector to implement
access restriction on a memory page basis. The second option is to use a page-table to provide access
restriction and address translation. Regardless of the protection strategy, the core provides means to
assign masters on the Master I/O AHB bus in groups where each group can be associated with a data
structure (access restriction vector or page table) in memory. The core supports a dynamically config-
urable page size from 4 to 512 KiB.
When a master on the Master I/O AHB bus initiates an access to be propagated, the bridge will first
look at the incoming master’s group assignment setting to determine to which group the master
belongs. When the group is known, the bridge can propagate or inhibit the access based on the group’s
attributes, or determine the address of the in-memory data structures to use for access checks (and
possibly address translation). The in-memory data structure may be cached by the bridge, otherwise
the information will be fetched from main memory.
Once the bridge has the necessary information to process the incoming access, the access will be
either allowed to propagate through the core or, in case the access is to a restricted memory location,
be inhibited. If the access is inhibited, the bridge will issue an AMBA ERROR response to the master
if the incoming access is a read access. The bridge implements posted writes, therefore write opera-
tions will not receive an AMBA ERROR response. An interrupt can, optionally, be asserted when an
access is inhibited. The AHB failing access register can be configured to log the first or most recent
access that was inhibited.
It is possible for masters to access the bridge’s register interface through the bridge. In this case the
bridge will perform an access to itself over the Processor and Slave I/O AHB buses.
Table 117.
Access latencies
Access
Master acc. cycles Slave cycles
Delay incurred by performing access over core
Single read
3
1
4 * clk
mst
Burst read with prefetch
2 + (burst length)
x
2
2 * clk
slv
+ (2 + burst length)* clk
mst
Single write
xx
(2)
0
0
Burst write
xx
(2 + (burst length)) 0
0
x
A prefetch operation ends at the address boundary defined by the prefetch buffer’s size
xx
The core implements posted writes, the number of cycles taken by the master side can only affect the next access.