edtlib: add hash attribute to nodes
#83748
Conversation
Add a new "hash" attribute to all Devicetree EDT nodes. The hash is calculated on the full path of the node; this means that its value remains stable across rebuilds. The hash is checked for uniqueness among nodes in the same EDT. This computed token is then added to `devicetree_generated.h` and made accessible to Zephyr code via a new DT_HASH(node_id) macro. Signed-off-by: Luca Burelli <[email protected]>
| # Calculates the hash associated with the node's full path. | ||
| hasher = hashlib.sha256() | ||
| hasher.update(path.encode()) | ||
| hash = base64.b64encode(hasher.digest()).decode() |
There was a problem hiding this comment.
Why encode it with base64? Maybe use hexdigest instead?
There was a problem hiding this comment.
My first version was with hexdigest(), but that is a 64 char string and thus the final identifiers ended up well over 80. Using base64 as was suggested in this thread trims ~20 chars off and makes them easier to distinguish, at basically the same vanishingly small collision probability.
There was a problem hiding this comment.
Understood, b64encode can be passed alternatives for the + and / chars already:
return base64.b64encode(hasher.digest(), altchars="__").decode().rstrip("=")
|
Does this PR have an associated open issue? I would like to review the discussion. Could you provide any links to discussions related to this proposal? |
@rruuaanng the rationale is described in the linked PR, but there is no issue, and by your comment it looks like there should have been. I'm afraid I'm still new to many of the Zephyr processes... 😰 |
| * @param node_id node identifier | ||
| * @return hash value as a preprocessor token | ||
| */ | ||
| #define DT_HASH(node_id) DT_CAT(node_id, _HASH) |
There was a problem hiding this comment.
| #define DT_HASH(node_id) DT_CAT(node_id, _HASH) | |
| #define DT_NODE_HASH(node_id) DT_CAT(node_id, _HASH) |
There was a problem hiding this comment.
I believe this is closely related to the node hierarchy, so HASH should not be used as a secondary naming convention. So, we should clearly label it as the function for obtaining the node's hash.
There was a problem hiding this comment.
Could you add a test for macro DT_HASH(or DT_NODE_HASH) in tests/lib/devicetree/api/src/main.c? I think this would be better, as the ultimate audience is C, not edtlib.
I just want to know more context, you're doing great. However, I'd like to check if this PR has opened an issue, as I don’t want to miss anything. |
Add a new
hashattribute to all Devicetree EDT nodes. The hash is calculated on the full path of the node; this means that its value remains stable across rebuilds. The hash is checked for uniqueness among nodes in the same EDT.This computed token is then added to
devicetree_generated.hand made accessible to Zephyr code via a newDT_HASH(node_id)macro.A proposed usage of this new attribute is in RFC #77799, to allow stable device lookups from extensions.