GR740-UM-DS, Nov 2017, Version 1.7
485
www.cobham.com/gaisler
GR740
In GR740 revision 1, the following registers are available
•
Error mode status register
•
Processor boot address registers for processors 0 - 3
The revision 1 register interface allows software to force a processor into debug or error mode. This
means that the interface can be used to stop (and restart) a processor while the interface in silicon revi-
sion 0 requires that a processor is idle before the processor can be restarted. See section 21.2.10 for
more information.
43.2.13 Ethernet mode selection register clocked by TXCLK
The Ethernet controllers had an internal register used to pipeline a signal that selects between 10/100
Mbit and 1000 Mbit (Gigabit) operation. This internal register was clocked by the external transmit
clock used for 10/100 Mbit operation. External transceivers remove the TXCLK when entering Giga-
bit mode and this may lead the internal control signal to be stuck in 10/100 Mbit mode. At least one
clock edge on TXCLK is required to switch from 10/100 Mbit mode to Gigabit mode.
Workaround:
No workaround is required to use the device in 10/100 Mbit networks. To support
Gigabit operation, a running clock (such as the Gigabit transmit clock) can be permanently connected
to the ETH0_TXCLK and ETH1_TXCLK inputs.
Applicable to:
This issue is only present in silicon revision 0.
43.2.14 GRETH_GBIT1 MDIO access starved by GRETH_GBIT0 EDCL infinite retries
The Ethernet controllers in the design share the same MDIO bus. If the Ethernet debug communica-
tions link (EDCL) is enabled then the Ethernet controller will try to configure the external transceiver
after reset. In case the transceiver could not be configured then the first Ethernet controller would
retry the configuration sequence infinitely and prevent the second Ethernet controller from accessing
the MDIO bus.
Workaround:
Always connect a transceiver to the first Ethernet interface if the second Ethernet
interface will be used.
Applicable to:
This issue is only present in silicon revision 0.
43.2.15 General purpose I/O port interrupt flag register immediately cleared
The general purpose I/O port interrupt flag register was cleared the clock cycle after the interrupt
source had been deasserted. This made the flag register only useful for level interrupts. See section
22.3.10.
Workaround:
None
Applicable to:
This issue is only present in silicon revision 0.
43.2.16 AHB status register multiple error logging
Silicon revision 1 adds functionality to the AHB status registers to support logging of multiple errors
and to allow filtering of correctable and uncorrectable errors. See section 27.
43.2.17 PCI DMA controller enable and retry behaviour
Checking of the PCI DMA controller’s channel enable bit was not implemented correctly, requiring
that silicon revision 0 devices always had all DMA channel enable bits set to ’1’ for transfers to be
handled correctly. The controller did not handle RETRY replies to on the last word of the descriptor
correctly.
Silicon revision 1 also updates the behaviour of the DMA controller, see section 15.6.3.