mirror of
https://github.com/openhwgroup/cvw
synced 2025-02-11 06:05:49 +00:00
8e1476cb8c
This bug occurs when there are back to back HPTW requests and the first generates an access fault during the walk. The old implementation uses a delayed version of the fault to prevent the HTPW fsm from transitioning out of the IDLE state. Because the first request generates the fault and the second request is pipelined the second request appears as if it also faults so the FSM does not perform the walk. The new implementation adds a FAULT state. When the HPTW generates an access fault it transitions to this state removes the HPTWStall and then transitions to IDLE. There may still be a remaining bug here if the pipeline is stalled for another reason. However I don't think it is possible by construction. The only possible sources of stalls at this point would be IFU and LSU stalls and both are required to make this condition happen. |
||
|---|---|---|
| .. | ||
| cache | ||
| ebu | ||
| fpu | ||
| generic | ||
| hazard | ||
| ieu | ||
| ifu | ||
| lsu | ||
| mdu | ||
| mmu | ||
| privileged | ||
| uncore | ||
| wally | ||