Skip to content

Commit

Permalink
GH-2045 Add transition from dpos fork database to instant finality fo…
Browse files Browse the repository at this point in the history 
…rk database. Remove fetch_block_state_* methods.
  • Loading branch information
@heifner
heifner committed Jan 14, 2024
1 parent abdfa68 commit 61a595f
Show file tree
Hide file tree
Showing 3 changed files with 91 additions and 82 deletions.
168 changes: 90 additions & 78 deletions libraries/chain/controller.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -127,7 +127,7 @@ struct block_data_gen_t {
bsp head;
fork_db_t fork_db;

block_data_gen_t(const std::filesystem::path& path) : fork_db(path) {}
explicit block_data_gen_t(const std::filesystem::path& path) : fork_db(path) {}

bsp fork_db_head(bool irreversible_mode) const {
if (irreversible_mode) {
Expand Down Expand Up @@ -177,6 +177,19 @@ struct block_data_t {
uint32_t head_block_num() const { return std::visit([](const auto& bd) { return bd.head->block_num(); }, v); }
block_timestamp_type head_block_time() const { return std::visit([](const auto& bd) { return bd.head->timestamp(); }, v); }
account_name head_block_producer() const { return std::visit([](const auto& bd) { return bd.head->producer(); }, v); }

void transistion_fork_db_to_if(const auto& vbsp) {
std::visit([](auto& bd) { bd.fork_db.close(); }, v);
auto bsp = std::make_shared<block_state>(*std::get<block_state_legacy_ptr>(vbsp));
v.emplace<block_data_new_t>(std::visit([](const auto& bd) { return bd.fork_db.get_data_dir(); }, v));
std::visit(overloaded{
[](block_data_legacy_t&) {},
[&](block_data_new_t& bd) {
bd.head = bsp;
bd.fork_db.reset(*bd.head);
}
}, v);
}

protocol_feature_activation_set_ptr head_activated_protocol_features() const { return std::visit([](const auto& bd) {
return bd.head->get_activated_protocol_features(); }, v);
Expand Down Expand Up @@ -863,8 +876,8 @@ struct building_block {
bb_input, transaction_mroot, action_mroot, std::move(bb.new_proposer_policy), std::move(bb.new_finalizer_policy),
qc_data ? qc_data->qc_info : std::optional<qc_info_t>{} };

assembled_block::assembled_block_if ab{bb.active_producer_authority, bb.parent.next(bhs_input),
bb.pending_trx_metas, bb.pending_trx_receipts,
assembled_block::assembled_block_if ab{std::move(bb.active_producer_authority), bb.parent.next(bhs_input),
std::move(bb.pending_trx_metas), std::move(bb.pending_trx_receipts),
qc_data ? std::move(qc_data->qc) : std::optional<quorum_certificate>{}};

return assembled_block{.v = std::move(ab)};
Expand Down Expand Up @@ -2444,7 +2457,6 @@ struct controller_impl {

// can change during start_block, so use same value throughout
uint32_t hs_lib = hs_irreversible_block_num.load();
const bool hs_active = hs_lib > 0; // the transition from 0 to >0 cannot happen during start_block

emit( self.block_start, head_block_num() + 1 );

Expand Down Expand Up @@ -2548,37 +2560,35 @@ struct controller_impl {

const auto& gpo = self.get_global_properties();

if (!hs_active) {
#warning todo: how do we update the producer_schedule after the switch to IF?
bb.apply_dpos<void>([&](building_block::building_block_dpos& bb_dpos) {
pending_block_header_state_legacy& pbhs = bb_dpos.pending_block_header_state;

if( gpo.proposed_schedule_block_num && // if there is a proposed schedule that was proposed in a block ...
( hs_active || *gpo.proposed_schedule_block_num <= pbhs.dpos_irreversible_blocknum ) && // ... that has now become irreversible or hotstuff activated...
pbhs.prev_pending_schedule.schedule.producers.size() == 0 // ... and there was room for a new pending schedule prior to any possible promotion
)
{
// Promote proposed schedule to pending schedule; happens in next block after hotstuff activated
EOS_ASSERT( gpo.proposed_schedule.version == pbhs.active_schedule_version + 1,
producer_schedule_exception, "wrong producer schedule version specified" );

bb_dpos.new_pending_producer_schedule = producer_authority_schedule::from_shared(gpo.proposed_schedule);

if( !replaying ) {
ilog( "promoting proposed schedule (set in block ${proposed_num}) to pending; current block: ${n} lib: ${lib} schedule: ${schedule} ",
("proposed_num", *gpo.proposed_schedule_block_num)("n", pbhs.block_num)
("lib", hs_active ? hs_lib : pbhs.dpos_irreversible_blocknum)
("schedule", bb_dpos.new_pending_producer_schedule ) );
}

db.modify( gpo, [&]( auto& gp ) {
gp.proposed_schedule_block_num = std::optional<block_num_type>();
gp.proposed_schedule.version=0;
gp.proposed_schedule.producers.clear();
});
// instant finality uses alternative method for chaning producer schedule
bb.apply_dpos<void>([&](building_block::building_block_dpos& bb_dpos) {
pending_block_header_state_legacy& pbhs = bb_dpos.pending_block_header_state;

if( gpo.proposed_schedule_block_num && // if there is a proposed schedule that was proposed in a block ...
( *gpo.proposed_schedule_block_num <= pbhs.dpos_irreversible_blocknum ) && // ... that has now become irreversible ...
pbhs.prev_pending_schedule.schedule.producers.size() == 0 // ... and there was room for a new pending schedule prior to any possible promotion
)
{
// Promote proposed schedule to pending schedule; happens in next block after hotstuff activated
EOS_ASSERT( gpo.proposed_schedule.version == pbhs.active_schedule_version + 1,
producer_schedule_exception, "wrong producer schedule version specified" );

bb_dpos.new_pending_producer_schedule = producer_authority_schedule::from_shared(gpo.proposed_schedule);

if( !replaying ) {
ilog( "promoting proposed schedule (set in block ${proposed_num}) to pending; current block: ${n} lib: ${lib} schedule: ${schedule} ",
("proposed_num", *gpo.proposed_schedule_block_num)("n", pbhs.block_num)
("lib", pbhs.dpos_irreversible_blocknum)
("schedule", bb_dpos.new_pending_producer_schedule ) );
}
});
}

db.modify( gpo, [&]( auto& gp ) {
gp.proposed_schedule_block_num = std::optional<block_num_type>();
gp.proposed_schedule.version=0;
gp.proposed_schedule.producers.clear();
});
}
});

try {
transaction_metadata_ptr onbtrx =
Expand Down Expand Up @@ -2619,7 +2629,7 @@ struct controller_impl {
void finish_block()
{
EOS_ASSERT( pending, block_validate_exception, "it is not valid to finalize when there is no pending block");
EOS_ASSERT( std::holds_alternative<building_block>(pending->_block_stage), block_validate_exception, "already called finalize_block");
EOS_ASSERT( std::holds_alternative<building_block>(pending->_block_stage), block_validate_exception, "already called finish_block");

try {
auto& bb = std::get<building_block>(pending->_block_stage);
Expand All @@ -2646,24 +2656,9 @@ struct controller_impl {
create_block_summary( assembled_block.id() );

pending->_block_stage = std::move(assembled_block);

/*
ilog( "finalized block ${n} (${id}) at ${t} by ${p} (${signing_key}); schedule_version: ${v} lib: ${lib} #dtrxs: ${ndtrxs} ${np}",
("n",pbhs.block_num())
("id",id)
("t",pbhs.timestamp)
("p",pbhs.producer)
("signing_key", pbhs.block_signing_key)
("v",pbhs.active_schedule_version)
("lib",pbhs.dpos_irreversible_blocknum)
("ndtrxs",db.get_index<generated_transaction_multi_index,by_trx_id>().size())
("np",block_ptr->new_producers)
);
*/

}
FC_CAPTURE_AND_RETHROW()
} /// finalize_block
}

/**
* @post regardless of the success of commit block there is no active pending block
Expand Down Expand Up @@ -2707,8 +2702,28 @@ struct controller_impl {

if( s == controller::block_status::incomplete ) {
log_irreversible();
pacemaker->beat();
}

// TODO: temp transistion to instant-finality, happens immediately after block with new_finalizer_policy
auto transition = [&](auto& fork_db, auto& head) -> bool {
const auto& bsp = std::get<std::decay_t<decltype(head)>>(cb.bsp);
std::optional<block_header_extension> ext = bsp->block->extract_header_extension(instant_finality_extension::extension_id());
if (ext) {
const auto& if_extension = std::get<instant_finality_extension>(*ext);
if (if_extension.new_finalizer_policy) {
ilog("Transition to instant finality happening after block ${b}", ("b", bsp->block_num()));
hs_irreversible_block_num = bsp->block_num();

log_irreversible();
return true;
}
}
return false;
};
if (block_data.apply_dpos<bool>(transition)) {
block_data.transistion_fork_db_to_if(cb.bsp);
}

} catch (...) {
// dont bother resetting pending, instead abort the block
reset_pending_on_exit.cancel();
Expand Down Expand Up @@ -3994,48 +4009,45 @@ std::optional<signed_block_header> controller::fetch_block_header_by_id( const b
}

signed_block_ptr controller::fetch_block_by_number( uint32_t block_num )const { try {
auto blk_state = fetch_block_state_by_number( block_num );
if( blk_state ) {
return blk_state->block;
}
auto fetch_block = [&](auto& fork_db, auto& head) -> signed_block_ptr {
auto bsp = fork_db.search_on_branch( fork_db.head()->id(), block_num);
if (bsp) return bsp->block;
return {};
};
auto b = my->block_data.apply<signed_block_ptr>(fetch_block);
if (b)
return b;

return my->blog.read_block_by_num(block_num);
} FC_CAPTURE_AND_RETHROW( (block_num) ) }

std::optional<signed_block_header> controller::fetch_block_header_by_number( uint32_t block_num )const { try {
auto blk_state = fetch_block_state_by_number( block_num );
if( blk_state ) {
return blk_state->header;
}
auto fetch_block = [&](auto& fork_db, auto& head) -> signed_block_ptr {
auto bsp = fork_db.search_on_branch( fork_db.head()->id(), block_num);
if (bsp) return bsp->block;
return {};
};
auto b = my->block_data.apply<signed_block_ptr>(fetch_block);
if (b)
return *b;

return my->blog.read_block_header_by_num(block_num);
} FC_CAPTURE_AND_RETHROW( (block_num) ) }


block_state_legacy_ptr controller::fetch_block_state_by_id( block_id_type id )const {
// returns nullptr when in IF mode
auto get_block_state = [&](auto& fork_db, auto& head) -> block_state_legacy_ptr { return fork_db.get_block(id); };
return my->block_data.apply_dpos<block_state_legacy_ptr>(get_block_state);
}

block_state_legacy_ptr controller::fetch_block_state_by_number( uint32_t block_num )const {
try {
// returns nullptr when in IF mode
auto fetch_block_state = [&](auto& fork_db, auto& head) -> block_state_legacy_ptr {
return fork_db.search_on_branch( fork_db.head()->id(), block_num);
};
return my->block_data.apply_dpos<block_state_legacy_ptr>(fetch_block_state);
} FC_CAPTURE_AND_RETHROW( (block_num) )
}

block_id_type controller::get_block_id_for_num( uint32_t block_num )const { try {
const auto& blog_head = my->blog.head();

bool find_in_blog = (blog_head && block_num <= blog_head->block_num());

if( !find_in_blog ) {
auto bsp = fetch_block_state_by_number( block_num );
if( bsp ) return bsp->id();
auto fetch_block_id = [&](auto& fork_db, auto& head) -> std::optional<block_id_type> {
auto bsp = fork_db.search_on_branch( fork_db.head()->id(), block_num);
if (bsp) return bsp->id();
return {};
};
auto id = my->block_data.apply<std::optional<block_id_type>>(fetch_block_id);
if (id) return *id;
}

auto id = my->blog.read_block_id_by_num(block_num);
Expand Down
1 change: 1 addition & 0 deletions libraries/chain/include/eosio/chain/block_state_legacy.hpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -52,6 +52,7 @@ namespace eosio::chain {
friend struct fc::reflector<block_state_legacy>;
friend struct controller_impl;
friend struct completed_block;
friend struct block_state;

bool is_pub_keys_recovered()const { return _pub_keys_recovered; }

Expand Down
4 changes: 0 additions & 4 deletions libraries/chain/include/eosio/chain/controller.hpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -287,10 +287,6 @@ namespace eosio::chain {
std::optional<signed_block_header> fetch_block_header_by_number( uint32_t block_num )const;
// thread-safe
std::optional<signed_block_header> fetch_block_header_by_id( const block_id_type& id )const;
// return block_state_legacy from forkdb, thread-safe
block_state_legacy_ptr fetch_block_state_by_number( uint32_t block_num )const;
// return block_state_legacy from forkdb, thread-safe
block_state_legacy_ptr fetch_block_state_by_id( block_id_type id )const;
// thread-safe
block_id_type get_block_id_for_num( uint32_t block_num )const;

Expand Down

0 comments on commit 61a595f

Please sign in to comment.