Check for known block when processing "header first" #2834
Conversation
| /// recently. Keeps duplicates from the network in check. | ||
| /// ctx here is specific to the header_head (tip of the header chain) | ||
| fn check_header_known(header: &BlockHeader, ctx: &mut BlockContext<'_>) -> Result<(), Error> { | ||
| let header_head = ctx.batch.header_head()?; |
There was a problem hiding this comment.
It said ctx here is specific to the header_head (tip of the header chain).
but in check_know(), the check_known_head(header, ctx) is using ctx.batch.head(), not header_head.
There was a problem hiding this comment.
I mean your new code has very different behavior from the old check_header_known which use header_head. now you're using ctx.batch.head().
There was a problem hiding this comment.
We used to check that the header was known based on the header chain.
The only time we would know about headers and not yet know about the corresponding full block is during initial (header) sync.
So it's safe to check if we know about the header based on the full chain - we're processing "header first" in the context of processing full blocks.
When processing a header in "header first" propagation we currently perform a quick check to see if we have just processed this header.
This does not cover the case where a peer sends (intentionally or otherwise) an old header.
We should do the same check here that we already do when processing full blocks -
If this is a header for a "known" block then we should halt processing of the header.