Solana Core BPF Stake Program.
| Information | Account Address |
|---|---|
| Stake Program | Stake11111111111111111111111111111111111111 |
This repository contains the BPF port of the native Agave Stake Program, along with the official Rust interface and Codama-generated Rust and Javascript clients for interacting with it. The Stake Program is what allows stake accounts to be created, delegated, split, merged, and so forth.
Historically, the Stake Program was built into the Agave validator client as a pseudo-program. This would necessitate that any future validator client reimplement it exactly. To ease the development of new validator clients, the Stake Program has been converted to an ordinary BPF program which can be invoked like any normal program. This BPF Stake Program is now live on all clusters.
The BPF Stake Program has received one external audit:
- Zellic (2025-03-12)
- Review commit hash
5ec49ccb - Final report (todo)
- Review commit hash
To build the BPF Stake Program, you must pass the bpf-entrypoint flag:
cargo build-sbf --features bpf-entrypoint
The BPF program deployed on all clusters is built with solana-verify. It may be verified independently by comparing the output of:
solana-verify get-program-hash -um Stake11111111111111111111111111111111111111
with:
solana-verify build --library-name program -- --features bpf-entrypoint
solana-verify get-executable-hash target/deploy/solana_stake_program.so
It is possible for a solana-verify version mismatch to affect the hash; BPF Stake Program 1.0.0 was built with solana-verify 0.4.6.
Instructions, errors, and assorted structs related to the stake program, which used to live in the Solana SDK repo, now live here. For more, see docs.rs.
Previously, the Stake Program was essentially free, costing 1500 Compute Units irrespective of any work the program actually did. As a normal BPF program, the Stake Program now pays Compute Units as any other program.
For the initial 1.0.0 release, we followed the existing Agave code as closely as possible, to minimize the possibility of introducing any bugs or changing any behaviors in the port because of the total rewrite of the calling interface the port necessitated. This means the existing program is expected to be non-optimal. With the ability to test more thoroughly against this 1.0.0 version, we expect to be able to optimize these costs in the future.
Based on the sample invocations in program/tests/interface.rs, approximate costs as of 2025-07-28 are as follows. These should be treated as baselines and are rounded to thousands; instructions may do less or more work depending on arguments and account states.
| Instruction | Estimated Cost |
|---|---|
Initialize |
9000 |
Authorize |
12000 |
DelegateStake |
15000 |
Split |
20000 |
Withdraw |
8000 |
Deactivate |
13000 |
SetLockup |
11000 |
Merge |
21000 |
AuthorizeWithSeed |
14000 |
InitializeChecked |
6000 |
AuthorizeChecked |
12000 |
AuthorizeCheckedWithSeed |
14000 |
SetLockupChecked |
11000 |
GetMinimumDelegation |
(negligible) |
DeactivateDelinquent |
18000 |
MoveStake |
25000 |
MoveLamports |
15000 |
