[otbn,doc] Document what happens if URND LFSR gets stuck
As well as adding to the documentation, this adds a new generic
"internal state is corrupt" bit to ERR_BITS and FATAL_ALERT_CAUSE and
wires everything up in the RTL. This is driven to zero for ERR_BITS in
the controller and for FATAL_ALERT_CAUSE in otbn.sv.
Signed-off-by: Rupert Swarbrick <rswarbrick@lowrisc.org>
diff --git a/hw/ip/otbn/doc/_index.md b/hw/ip/otbn/doc/_index.md
index 737a4a8..5c91d1e 100644
--- a/hw/ip/otbn/doc/_index.md
+++ b/hw/ip/otbn/doc/_index.md
@@ -530,6 +530,10 @@
Each new execution of OTBN will reseed the `URND` PRNG.
The PRNG state is advanced every cycle when OTBN is running.
+The PRNG has a long cycle length but has a fixed point: the sequence of numbers will get stuck if the state ever happens to become zero.
+This will never happen in normal operation.
+If a fault causes the state to become zero, OTBN raises a `BAD_INTERNAL_STATE` fatal error.
+
### Operational States {#design-details-operational-states}
<!--
