The cosimulation environment does not know if a memory access from spike
is due to an instruction fetch or a data memory access. It uses a
heuristic to differentiate the two. Any access between the PC and the PC
+ 8 is considered an instruction fetch.
This heuristic did not correctly handle addresses at the top of the
range where the PC + 8 calculation overflows. This commit fixes the top
of range handling.
The 'pre_val' MIP addresses the scenario where MIP changes as an
instruction is excuting, this means a CSR instruction can observe a
different MIP from the one that decides whether or not that instruction
will be interrupted.
Where an access is unaligned Ibex splits it into two transactions, each
of which undergoes a PMP check. It is possible for the first half to
fail a PMP check and the second to succeed and hence produce a request
on the memory interface.
In Spike it accesses memory byte by byte and if it encounters a PMP
error for a particular byte it won't try any further bytes.
This results in a mis-match between Ibex and spike when an unaligned
transaction is split across two PMP regions, one of which allows the
access and the other doesn't. Ibex generates a transaction and spike
doesn't producing an error.
This adds a fixup into the co-simulation environment. It detects when we
have an access that fails PMP that is misaligned. Where this has
resulted in Ibex producing a memory request that spike would not we
remove it from the list of memory requests to check after checking that
the request passes PMP within spike.
When SpikeCosim is getting destructed a SIGSEGV was observed on CentOS
7. The root cause hasn't been identified other than it relates to the
deletion of `isa_parser_t`, potentially some kind of use after free
error.
This is an (optional) hacky workaround that simply never deletes the
`isa_parser_t` pointer in SpikeCosim. As in practise this occurs at the
end of simulation when the process is terminating the memory leak is of
little consequence.
Longer term this issue should be investigated and properly fixed.
Previously any changes in interrupt state or debug requests were
strictly associated with retired instructions. This causes cosim
mismatches where a lower priority interrupt occurs in time before a
higher priority interrupt or debug request but between instruction
fetches/retirements so both the low and high priority interrupts are
signalled with the instruction retirement.
This introduces a way for the RVFI to signal an interrupt has occurred
that isn't associated with an instruction retirement to allow the cosim
to see the seperation in time between different interrupts and debug
requests and hence model behaviour correctly.
When Ibex does a load that receives data with bad integrity it
suppresses the write to the destination register. The implements
matching functionality for cosim.
This commit is mainly an extension to cosim environment to drive the newly
introduced state variable `nmi_int` in Spike.
This commit
- Extends RVFI interface by a single bit (ext_nmi_int)
- Configures cosim to set nmi_int inside Spike
Signed-off-by: Canberk Topal <ctopal@lowrisc.org>
When DCSR is set such that ebreak will enter debug mode we were getting
cosim mismatches. This was because Ibex produces the ebreak on the RVFI
interface and spike effectively skips right over it and executes the
first instruction of the debug handler immediately. Traps have similar
but not identical behaviour so we need a special case in the step
function to handle this.
This commit adds another field in SpikeCosim class so that Spike
can hardcode the correct event registers.
Signed-off-by: Canberk Topal <ctopal@lowrisc.org>
The ic_scr_key_valid field indicates whether the ICache scrambling key
is valid.
CPUCTRL is also renamed CPUCTRLSTS as it contains both control and
status bits.
Spike has a hack that decrements mcycle/minstret when written.
Previously this was not handled correctly by cosim. As we can only write
32 bits at a time two writes must be used and incremented to counteract
the decrementing handled carefully.
Interrupts are disabled in Debug Mode (Sdext 4.1.2), and simultaneously
registers, including MCAUSE, are not updated by exceptions (Sdext 4.1.3),
so reading MCAUSE[31] after an exception (eg. invalid instruction) in
debug_mode may still report the previous cause (which could be an interrupt).
Seperate concerns so the flow of stepping spike and checking against
the ibex RVFI data is clearer. One 'step' of either system produces
different amounts of progress, and this conditional-checking-and-stepping
is needed to tie up the flows.
Extends RVFI connections further to include 30 mhpmcounterX registers.
Sets them up before every cosim step to let Spike know their real values.
Signed-off-by: Canberk Topal <ctopal@lowrisc.org>
This works with versions ibex-cosim-v0.2 and ibex-cosim-v0.3. The latter
version is required to support the mseccfg CSR added with ePMP.
Signed-off-by: Pirmin Vogel <vogelpi@lowrisc.org>
