c_emulator: validate DTB range against configured PMA regions #1437
Conversation
3dd89e0 to
0e5e735
Compare
model/postlude/validate_config.sail
Outdated
|
|
||
| // Return true iff [addr, addr+size) is fully contained in a single configured PMA memory region. | ||
| function dtb_within_configured_pma_memory(addr : bits(64), size : bits(64)) -> bool = { | ||
| if size == zeros() then true |
There was a problem hiding this comment.
I think this is unnecessary too?
There was a problem hiding this comment.
Thanks! I dropped the explicit size == 0 case and simplified it to is_some.
model/postlude/validate_config.sail
Outdated
| // Return true iff [addr, addr+size) is fully contained in a single configured PMA memory region. | ||
| function dtb_within_configured_pma_memory(addr : bits(64), size : bits(64)) -> bool = { | ||
| if size == zeros() then true | ||
| else match matching_pma_bits_range(pma_regions, addr, size) { |
There was a problem hiding this comment.
I believe there's an is_some function you can use, so
function is_within_configured_pma_memory(addr : bits(64), size : bits(64)) -> bool =
is_some(matching_pma_bits_range(pma_regions, addr, size))
Also this isn't really dtb specific so we may as well make the name generic.
0e5e735 to
c7d5d7b
Compare
|
How did this get through the merge queue even though CI is failing? It looks like the CI passed in the merge queue but failed once it was actually merged into master. Pre-commit is the part that failed, so it wasn't just a transient issue or anything like that. |
|
It looks like CI for PR #1437 failed both on merge queue (Job 769, https://github.com/riscv/sail-riscv/actions/runs/20458313718) as well as once it was on master (Job 770, https://github.com/riscv/sail-riscv/actions/runs/20458784253). So this looks like a Github actions bug. |
Add a pre-write check that the DTB range [dtb_address, dtb_address + dtb_size) is fully contained in a single configured memory.regions PMA memory region; otherwise print an error and exit.
Fixes #1308.