Kernel blockstorage pruning #36419

I would like to be able to do pruning in the kernel-node, but right now the kernel can't open a data directory that Bitcoin Core has pruned or prune its own block storage.

This separates block storage from chainstate management first and then adds pruning to the kernel API.

Commits 1 to 6 are refactors, and should not impose any behavior change. Commits 7 to 10 add a block manager and pruning to the kernel API

1. Apply prune locks in BlockManager instead of in validation.
2. Give BlockManager a height range and sizes to prune within, instead of letting it read the chainstate.
3. Move ImportBlocks, which drives validation, out of block storage.
4. Remove the friend declarations, so validation only uses BlockManager's public interface.
5. Create the BlockManager first and pass it to ChainstateManager by reference. Block storage then no longer belongs to chainstate management.
6. Remove ChainstateLoadOptions::prune, which duplicated BlockManager's prune mode.
7. Add btck_BlockManager and read blocks through it. This changes the first parameter of btck_block_read and btck_block_spent_outputs_read.
8. Add pruning up to a height and prune locks. Also lets the kernel open a data directory that Bitcoin Core has pruned.
9. Add automatic pruning to a target size.
10. Add btck_chainstate_manager_get_first_available_entry

<!--e57a25ab6845829454e8d69fc972939a-->

The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

<!--006a51241073e994b41acfe9ec718e94-->

External sites

• Code coverage & benchmarks: https://corecheck.dev/bitcoin/bitcoin/pulls/36419
• Ralph review: https://git.fish.foo/bitcoin/bitcoin/pulls/36419#author-ralph

<!--021abf342d371248e50ceaed478a90ca-->

Reviews

See the guideline and AI policy for information on the review process. A summary of reviews will appear here.

<!--5faf32d7da4f0f540f40219e4f7537a3-->

LLM Linter (✨ experimental)

Possible typos and grammar issues:

• // make sure we don't prune above any of the prune locks bestblocks -> // make sure we don't prune above any of the prune locks' best blocks [“bestblocks” is a typo; the intended phrase is unclear without splitting it into two words]

Possible places where comparison-specific test macros should replace generic comparisons:

• [src/test/kernel/test_kernel.cpp] BOOST_CHECK_THROW(chain.GetByHeight(1000), std::runtime_error); -> Prefer BOOST_CHECK_EXCEPTION with a predicate that checks the exception text, rather than only the generic exception type.
• [src/test/kernel/test_kernel.cpp] BOOST_CHECK_THROW(blockman.ReadBlockSpentOutputs(chain.GetByHeight(1)), std::runtime_error); -> Prefer BOOST_CHECK_EXCEPTION with a message matcher to validate the specific failure.
• [src/test/kernel/test_kernel.cpp] BOOST_CHECK_THROW(create_chainman(test_directory, /*reindex=*/false, /*wipe_chainstate=*/false, /*block_tree_db_in_memory=*/false, /*chainstate_db_in_memory=*/false, context), std::runtime_error); -> Prefer BOOST_CHECK_EXCEPTION with a predicate on the error message.
• [src/test/kernel/test_kernel.cpp] BOOST_CHECK_THROW(blockman.ReadBlockSpentOutputs(tip), std::runtime_error); -> Prefer BOOST_CHECK_EXCEPTION with a message matcher instead of only asserting the exception type.

^{2026-10-02 15:21:02}