[otbn,doc] Document the bus-accessible part of DMEM #8890
Conversation
rswarbrick
added
Component:Doc
|
I really dislike the "lower" vs "upper" naming, as it always confuses me: when I think of a memory layout, I start at the top with address 0, and it then grows downwards. So the "upper half" is address 0 to 2k-1 for me, and the "lower half" is address 2k to 4k-1. I'd just make it explicit which addresses we mean. Felix, Chris: Can you have a look? This feature is intended to give OTBN software a scratchpad/private memory, which cannot easily be read out from the bus. Intermediate values, keys, etc. should be stored there to give them a bit of additional protection from read-out attacks. |
|
@imphil: I'd forgotten you disliked that terminology. Sorry! Will re-word (and similarly in the tooling / DV patches I've just pushed). |
rswarbrick
force-pushed
the
otbn-2kb-dmem-doc
branch
from
886e8f6
to
d9c832c
Compare
|
Done. Honestly, I think it's more clunky, but apparently not everyone has the same mental model of memory as me :-) |
rswarbrick
force-pushed
the
otbn-2kb-dmem-doc
branch
from
d9c832c
to
8919391
Compare
hw/ip/otbn/doc/_index.md
Outdated
| @@ -500,6 +500,9 @@ These accesses are ignored if OTBN is busy. | |||
| A host processor can check whether OTBN is busy by reading the {{< regref "STATUS">}} register. | |||
| All memory accesses through the register interface must be word-aligned 32b word accesses. | |||
|
|
|||
| While DMEM is 4kiB, only first 2kiB (at addresses `0x0` to `0x7ff`) is visible through the register interface. | |||
| This is to allow OTBN applications to store sensitive information in the upper half, making it harder for that information to leak back to Ibex. | |||
There was a problem hiding this comment.
Did you miss the "upper" here when switching to the first/second terminology.
There was a problem hiding this comment.
Gah! Yes, I did: thanks. For other window sizes: we could do that easily enough. The only awkward bit is that there are some (rather stupid) bits of tooling that would need to know both sizes: at the moment, my pending patches hardwire the "divide by 2" but there's no real reason we couldn't do better.
rswarbrick
force-pushed
the
otbn-2kb-dmem-doc
branch
2 times, most recently
from
218b252
to
6cd0338
Compare
This doesn't yet change the implementation, nor does it change DV or tooling code. Signed-off-by: Rupert Swarbrick <rswarbrick@lowrisc.org>
rswarbrick
force-pushed
the
otbn-2kb-dmem-doc
branch
from
6cd0338
to
a8e595c
Compare
This doesn't yet change the implementation, nor does it change DV or
tooling code.
Fixes #8720.
(@GregAC, @imphil: Any ideas on other good reviewers for doc/spec changes like this once Philipp has left?)