Line data Source code
1 : #include "fd_poh.h"
2 :
3 : /* The PoH implementation is at its core a state machine ...
4 :
5 : +--------+
6 : | UNINIT |
7 : +--------+
8 : |
9 : +---------+ | +---------+
10 : | v v v |
11 : +-------------------+ +----------+ +------------------+
12 : | WAITING_FOR_SLOT |<----| FOLLOWER |----->| WAITING_FOR_BANK |
13 : +-------------------+ +----------+ +------------------+
14 : | ^ |
15 : | | |
16 : | +----------+ |
17 : |------>| LEADER |<--------+
18 : +----------+
19 :
20 : The state machine starts UNINIT, but once a snapshot is loaded it
21 : will transition to follower.
22 :
23 : The state machine is in a resting the state when FOLLOWER, in this
24 : state it knows a `next_leader_slot` and will continually hash to
25 : advance towards that slot. When it reaches the `next_leader_slot`
26 : it will transition to the WAITING_FOR_BANK state, where it waits for
27 : the replay stage to tell it some information relevant to that leader
28 : slot, so that it can start doing mixins and hashing towards the end
29 : of the block. When the block ends, the state transitions back to
30 : follower, even if the next slot is the leader, as we need the replay
31 : stage to tell us about the new leader slot.
32 :
33 : Sometimes it might happen that we have received the bank from replay
34 : stage before we have reached the `next_leader_slot`, in which case
35 : we transition to the WAITING_FOR_SLOT state, where we wait for the
36 : hash count to reach the leader slot.
37 :
38 : At any time, during any state except UNINIT, we can be suddenly
39 : "reset" by the replay tile. Such reset actions may move the reset
40 : slot backwards or forwards, or set it back to something we have
41 : already seen before. BUT, the `next_leader_slot` must always
42 : advance forward.
43 :
44 : If the PoH machine successfully completes a leader slot, by hashing
45 : it until the end, then the a completion message is sent back to
46 : replay with the final blockhash, after which the state machine enters
47 : the follower state once again, and waits for further instructions
48 : from replay. */
49 :
50 0 : #define STATE_UNINIT (0)
51 0 : #define STATE_FOLLOWER (1)
52 0 : #define STATE_WAITING_FOR_BANK (2)
53 0 : #define STATE_WAITING_FOR_SLOT (3)
54 0 : #define STATE_LEADER (4)
55 0 : #define STATE_WAITING_FOR_RESET (5)
56 :
57 : FD_FN_CONST ulong
58 0 : fd_poh_align( void ) {
59 0 : return FD_POH_ALIGN;
60 0 : }
61 :
62 : FD_FN_CONST ulong
63 0 : fd_poh_footprint( void ) {
64 0 : return FD_POH_FOOTPRINT;
65 0 : }
66 :
67 : void *
68 0 : fd_poh_new( void * shmem ) {
69 0 : fd_poh_t * poh = (fd_poh_t *)shmem;
70 :
71 0 : if( FD_UNLIKELY( !poh ) ) {
72 0 : FD_LOG_WARNING(( "NULL shmem" ));
73 0 : return NULL;
74 0 : }
75 :
76 0 : if( FD_UNLIKELY( !fd_ulong_is_aligned( (ulong)poh, fd_poh_align() ) ) ) {
77 0 : FD_LOG_WARNING(( "misaligned shmem" ));
78 0 : return NULL;
79 0 : }
80 :
81 0 : poh->hashcnt_per_tick = ULONG_MAX;
82 0 : poh->state = STATE_UNINIT;
83 :
84 0 : FD_COMPILER_MFENCE();
85 0 : FD_VOLATILE( poh->magic ) = FD_POH_MAGIC;
86 0 : FD_COMPILER_MFENCE();
87 :
88 0 : return (void *)poh;
89 0 : }
90 :
91 : fd_poh_t *
92 : fd_poh_join( void * shpoh,
93 : fd_poh_out_t * shred_out,
94 0 : fd_poh_out_t * replay_out ) {
95 0 : if( FD_UNLIKELY( !shpoh ) ) {
96 0 : FD_LOG_WARNING(( "NULL shpoh" ));
97 0 : return NULL;
98 0 : }
99 :
100 0 : if( FD_UNLIKELY( !fd_ulong_is_aligned( (ulong)shpoh, fd_poh_align() ) ) ) {
101 0 : FD_LOG_WARNING(( "misaligned shpoh" ));
102 0 : return NULL;
103 0 : }
104 :
105 0 : fd_poh_t * poh = (fd_poh_t *)shpoh;
106 :
107 0 : if( FD_UNLIKELY( poh->magic!=FD_POH_MAGIC ) ) {
108 0 : FD_LOG_WARNING(( "bad magic" ));
109 0 : return NULL;
110 0 : }
111 :
112 0 : *poh->shred_out = *shred_out;
113 0 : *poh->replay_out = *replay_out;
114 :
115 0 : return poh;
116 0 : }
117 :
118 : static void
119 : transition_to_follower( fd_poh_t * poh,
120 : fd_stem_context_t * stem,
121 0 : int completed_leader_slot ) {
122 0 : FD_TEST( poh->state==STATE_LEADER || poh->state==STATE_WAITING_FOR_BANK || poh->state==STATE_WAITING_FOR_SLOT || poh->state==STATE_WAITING_FOR_RESET );
123 :
124 0 : if( FD_LIKELY( completed_leader_slot ) ) FD_TEST( poh->state==STATE_LEADER );
125 :
126 0 : if( FD_LIKELY( poh->state==STATE_LEADER || poh->state==STATE_WAITING_FOR_SLOT ) ) {
127 0 : fd_poh_leader_slot_ended_t * dst = fd_chunk_to_laddr( poh->replay_out->mem, poh->replay_out->chunk );
128 0 : dst->completed = completed_leader_slot;
129 0 : dst->slot = poh->slot-1UL;
130 0 : fd_memcpy( dst->blockhash, poh->hash, 32UL );
131 0 : ulong tspub = (ulong)fd_frag_meta_ts_comp( fd_tickcount() );
132 0 : fd_stem_publish( stem, poh->replay_out->idx, 0UL, poh->replay_out->chunk, sizeof(fd_poh_leader_slot_ended_t), 0UL, 0UL, tspub );
133 0 : poh->replay_out->chunk = fd_dcache_compact_next( poh->replay_out->chunk, sizeof(fd_poh_leader_slot_ended_t), poh->replay_out->chunk0, poh->replay_out->wmark );
134 0 : }
135 :
136 0 : poh->state = STATE_FOLLOWER;
137 0 : }
138 :
139 : static void
140 : update_hashes_per_tick( fd_poh_t * poh,
141 0 : ulong hashcnt_per_tick ) {
142 0 : if( FD_UNLIKELY( poh->hashcnt_per_tick!=hashcnt_per_tick ) ) {
143 0 : if( FD_UNLIKELY( poh->hashcnt_per_tick!=ULONG_MAX ) ) {
144 0 : FD_LOG_WARNING(( "hashes per tick changed from %lu to %lu", poh->hashcnt_per_tick, hashcnt_per_tick ));
145 0 : }
146 :
147 : /* Recompute derived information about the clock. */
148 0 : poh->hashcnt_duration_ns = (double)poh->tick_duration_ns/(double)hashcnt_per_tick;
149 0 : poh->hashcnt_per_slot = poh->ticks_per_slot*hashcnt_per_tick;
150 0 : poh->hashcnt_per_tick = hashcnt_per_tick;
151 :
152 : /* Discard any ticks we might have done in the interim. They will
153 : have the wrong number of hashes per tick. We can just catch back
154 : up quickly if not too many slots were skipped and hopefully
155 : publish on time. Note that tick production and verification of
156 : skipped slots is done for the eventual bank that publishes a
157 : slot, for example:
158 :
159 : Reset Slot: 998
160 : Epoch Transition Slot: 1000
161 : Leader Slot: 1002
162 :
163 : In this case, if a feature changing the hashcnt_per_tick is
164 : activated in slot 1000, and we are publishing empty ticks for
165 : slots 998, 999, 1000, and 1001, they should all have the new
166 : hashes_per_tick number of hashes, rather than the older one, or
167 : some combination. */
168 :
169 0 : FD_TEST( poh->last_slot==poh->reset_slot );
170 0 : FD_TEST( !poh->last_hashcnt );
171 0 : poh->slot = poh->reset_slot;
172 0 : poh->hashcnt = 0UL;
173 0 : }
174 0 : }
175 :
176 : void
177 : fd_poh_reset( fd_poh_t * poh,
178 : fd_stem_context_t * stem,
179 : long timestamp, /* The local timestamp when the reset is occurring */
180 : ulong hashcnt_per_tick, /* The hashcnt per tick of the bank that completed */
181 : ulong ticks_per_slot,
182 : ulong tick_duration_ns,
183 : ulong completed_slot, /* The slot that successfully produced a block */
184 : uchar const * completed_blockhash, /* The hash of the last tick in the produced block */
185 : ulong next_leader_slot, /* The next slot where this node will be leader */
186 : ulong max_microblocks_in_slot, /* The maximum number of microblocks that may appear in a slot */
187 0 : uchar const * completed_block_id /* The block id of the completed block */) {
188 0 : memcpy( poh->reset_hash, completed_blockhash, 32UL );
189 0 : memcpy( poh->hash, completed_blockhash, 32UL );
190 0 : memcpy( poh->completed_block_id, completed_block_id, 32UL );
191 0 : poh->slot = completed_slot+1UL;
192 0 : poh->hashcnt = 0UL;
193 0 : poh->last_slot = poh->slot;
194 0 : poh->last_hashcnt = 0UL;
195 0 : poh->reset_slot = poh->slot;
196 0 : poh->next_leader_slot = next_leader_slot;
197 0 : poh->max_microblocks_per_slot = max_microblocks_in_slot;
198 0 : poh->reset_slot_start_ns = timestamp;
199 :
200 0 : if( FD_UNLIKELY( poh->state==STATE_UNINIT ) ) {
201 0 : poh->tick_duration_ns = tick_duration_ns;
202 0 : poh->ticks_per_slot = ticks_per_slot;
203 0 : poh->state = STATE_FOLLOWER;
204 0 : } else {
205 0 : FD_TEST( tick_duration_ns==poh->tick_duration_ns );
206 0 : FD_TEST( ticks_per_slot==poh->ticks_per_slot );
207 0 : }
208 0 : update_hashes_per_tick( poh, hashcnt_per_tick );
209 :
210 : /* When we reset, we need to allow PoH to tick freely again rather
211 : than being constrained. If we are leader after the reset, this
212 : is OK because we won't tick until we get a bank, and the lower
213 : bound will be reset with the value from the bank. */
214 0 : poh->microblocks_lower_bound = poh->max_microblocks_per_slot;
215 :
216 0 : if( FD_UNLIKELY( poh->state!=STATE_FOLLOWER ) ) transition_to_follower( poh, stem, 0 );
217 0 : if( FD_UNLIKELY( poh->slot==poh->next_leader_slot ) ) poh->state = STATE_WAITING_FOR_BANK;
218 0 : }
219 :
220 : void
221 : fd_poh_begin_leader( fd_poh_t * poh,
222 : ulong slot,
223 : ulong hashcnt_per_tick,
224 : ulong ticks_per_slot,
225 : ulong tick_duration_ns,
226 0 : ulong max_microblocks_in_slot ) {
227 0 : FD_TEST( poh->state==STATE_FOLLOWER || poh->state==STATE_WAITING_FOR_BANK );
228 0 : FD_TEST( slot==poh->next_leader_slot );
229 :
230 : /* PoH ends the slot once it "ticks" through all of the hashes, but we
231 : only want that to happen if we have received a done packing message
232 : from pack, so we always reserve an empty microblock at the end so
233 : the tick advance will not end the slot without being told. */
234 0 : poh->max_microblocks_per_slot = max_microblocks_in_slot+1UL;
235 :
236 0 : FD_TEST( tick_duration_ns==poh->tick_duration_ns );
237 0 : FD_TEST( ticks_per_slot==poh->ticks_per_slot );
238 0 : update_hashes_per_tick( poh, hashcnt_per_tick );
239 :
240 0 : if( FD_LIKELY( poh->state==STATE_FOLLOWER ) ) poh->state = STATE_WAITING_FOR_SLOT;
241 0 : else poh->state = STATE_LEADER;
242 :
243 0 : poh->microblocks_lower_bound = 0UL;
244 :
245 0 : FD_LOG_INFO(( "begin_leader(slot=%lu, last_slot=%lu, last_hashcnt=%lu)", slot, poh->last_slot, poh->last_hashcnt ));
246 0 : }
247 :
248 : int
249 0 : fd_poh_have_leader_bank( fd_poh_t const * poh ) {
250 0 : return poh->state==STATE_WAITING_FOR_SLOT || poh->state==STATE_LEADER;
251 0 : }
252 :
253 : int
254 0 : fd_poh_hashing_to_leader_slot( fd_poh_t const * poh ) {
255 0 : int hashing = poh->state==STATE_WAITING_FOR_SLOT || poh->state==STATE_LEADER;
256 0 : return hashing && poh->slot<poh->next_leader_slot;
257 0 : }
258 :
259 : int
260 0 : fd_poh_must_tick( fd_poh_t const * poh ) {
261 0 : return poh->state==STATE_LEADER && (poh->hashcnt%poh->hashcnt_per_tick)==(poh->hashcnt_per_tick-1UL);
262 0 : }
263 :
264 : void
265 : fd_poh_done_packing( fd_poh_t * poh,
266 0 : ulong microblocks_in_slot ) {
267 0 : FD_TEST( poh->state==STATE_LEADER );
268 0 : FD_LOG_INFO(( "done_packing(slot=%lu,seen_microblocks=%lu,microblocks_in_slot=%lu)",
269 0 : poh->slot,
270 0 : poh->microblocks_lower_bound,
271 0 : microblocks_in_slot ));
272 0 : FD_TEST( poh->microblocks_lower_bound==microblocks_in_slot );
273 0 : FD_TEST( poh->microblocks_lower_bound<=poh->max_microblocks_per_slot );
274 0 : poh->microblocks_lower_bound += poh->max_microblocks_per_slot - microblocks_in_slot;
275 0 : FD_TEST( poh->microblocks_lower_bound==poh->max_microblocks_per_slot );
276 0 : }
277 :
278 : static void
279 : publish_tick( fd_poh_t * poh,
280 : fd_stem_context_t * stem,
281 : uchar hash[ static 32 ],
282 0 : int is_skipped ) {
283 0 : ulong hashcnt = poh->hashcnt_per_tick*(1UL+(poh->last_hashcnt/poh->hashcnt_per_tick));
284 :
285 0 : uchar * dst = (uchar *)fd_chunk_to_laddr( poh->shred_out->mem, poh->shred_out->chunk );
286 :
287 0 : FD_TEST( poh->last_slot>=poh->reset_slot );
288 0 : fd_entry_batch_meta_t * meta = (fd_entry_batch_meta_t *)dst;
289 0 : if( FD_UNLIKELY( is_skipped ) ) {
290 : /* We are publishing ticks for a skipped slot, the reference tick
291 : and block complete flags should always be zero. */
292 0 : meta->reference_tick = 0UL;
293 0 : meta->block_complete = 0;
294 0 : } else {
295 0 : meta->reference_tick = hashcnt/poh->hashcnt_per_tick;
296 0 : meta->block_complete = hashcnt==poh->hashcnt_per_slot;
297 0 : }
298 :
299 0 : meta->parent_block_id_valid = 1;
300 0 : fd_memcpy( meta->parent_block_id, poh->completed_block_id, 32UL );
301 :
302 0 : ulong slot = fd_ulong_if( meta->block_complete, poh->slot-1UL, poh->slot );
303 0 : meta->parent_offset = 1UL+slot-poh->reset_slot;
304 :
305 0 : FD_TEST( hashcnt>poh->last_hashcnt );
306 0 : ulong hash_delta = hashcnt-poh->last_hashcnt;
307 :
308 0 : dst += sizeof(fd_entry_batch_meta_t);
309 0 : fd_entry_batch_header_t * tick = (fd_entry_batch_header_t *)dst;
310 0 : tick->hashcnt_delta = hash_delta;
311 0 : fd_memcpy( tick->hash, hash, 32UL );
312 0 : tick->txn_cnt = 0UL;
313 :
314 0 : ulong tspub = (ulong)fd_frag_meta_ts_comp( fd_tickcount() );
315 0 : ulong sz = sizeof(fd_entry_batch_meta_t)+sizeof(fd_entry_batch_header_t);
316 0 : ulong sig = fd_disco_poh_sig( slot, POH_PKT_TYPE_MICROBLOCK, 0UL );
317 0 : fd_stem_publish( stem, poh->shred_out->idx, sig, poh->shred_out->chunk, sz, 0UL, 0UL, tspub );
318 0 : poh->shred_out->chunk = fd_dcache_compact_next( poh->shred_out->chunk, sz, poh->shred_out->chunk0, poh->shred_out->wmark );
319 :
320 0 : if( FD_UNLIKELY( hashcnt==poh->hashcnt_per_slot ) ) {
321 0 : poh->last_slot++;
322 0 : poh->last_hashcnt = 0UL;
323 0 : } else {
324 0 : poh->last_hashcnt = hashcnt;
325 0 : }
326 0 : }
327 :
328 : void
329 : fd_poh_advance( fd_poh_t * poh,
330 : fd_stem_context_t * stem,
331 : int * opt_poll_in,
332 0 : int * charge_busy ) {
333 0 : if( FD_UNLIKELY( poh->state==STATE_UNINIT || poh->state==STATE_WAITING_FOR_RESET ) ) return;
334 0 : if( FD_UNLIKELY( poh->state==STATE_WAITING_FOR_BANK ) ) {
335 : /* If we are the leader, but we didn't yet learn what the leader
336 : bank object is from the replay tile, do not do any hashing. */
337 0 : return;
338 0 : }
339 :
340 : /* If we have skipped ticks pending because we skipped some slots to
341 : become leader, register them now one at a time. */
342 0 : if( FD_UNLIKELY( poh->state==STATE_LEADER && poh->last_slot<poh->slot ) ) {
343 0 : ulong publish_hashcnt = poh->last_hashcnt+poh->hashcnt_per_tick;
344 0 : ulong tick_idx = (poh->last_slot*poh->ticks_per_slot+publish_hashcnt/poh->hashcnt_per_tick)%MAX_SKIPPED_TICKS;
345 :
346 0 : publish_tick( poh, stem, poh->skipped_tick_hashes[ tick_idx ], 1 );
347 :
348 : /* If we are catching up now and publishing a bunch of skipped
349 : ticks, we do not want to process any incoming microblocks until
350 : all the skipped ticks have been published out; otherwise we would
351 : intersperse skipped tick messages with microblocks. */
352 0 : *opt_poll_in = 0;
353 0 : *charge_busy = 1;
354 0 : return;
355 0 : }
356 :
357 0 : int low_power_mode = poh->hashcnt_per_tick==1UL;
358 :
359 : /* If we are the leader, always leave enough capacity in the slot so
360 : that we can mixin any potential microblocks still coming from the
361 : pack tile for this slot. */
362 0 : ulong max_remaining_microblocks = poh->max_microblocks_per_slot - poh->microblocks_lower_bound;
363 :
364 : /* With hashcnt_per_tick hashes per tick, we actually get
365 : hashcnt_per_tick-1 chances to mixin a microblock. For each tick
366 : span that we need to reserve, we also need to reserve the hashcnt
367 : for the tick, hence the +
368 : max_remaining_microblocks/(hashcnt_per_tick-1) rounded up.
369 :
370 : However, if hashcnt_per_tick is 1 because we're in low power mode,
371 : this should probably just be max_remaining_microblocks. */
372 0 : ulong max_remaining_ticks_or_microblocks = max_remaining_microblocks;
373 0 : if( FD_LIKELY( !low_power_mode ) ) max_remaining_ticks_or_microblocks += (max_remaining_microblocks+poh->hashcnt_per_tick-2UL)/(poh->hashcnt_per_tick-1UL);
374 :
375 0 : ulong restricted_hashcnt = fd_ulong_if( poh->hashcnt_per_slot>=max_remaining_ticks_or_microblocks, poh->hashcnt_per_slot-max_remaining_ticks_or_microblocks, 0UL );
376 :
377 0 : ulong min_hashcnt = poh->hashcnt;
378 :
379 0 : if( FD_LIKELY( !low_power_mode ) ) {
380 : /* Recall that there are two kinds of events that will get published
381 : to the shredder,
382 :
383 : (a) Ticks. These occur every 62,500 (hashcnt_per_tick) hashcnts,
384 : and there will be 64 (ticks_per_slot) of them in each slot.
385 :
386 : Ticks must not have any transactions mixed into the hash.
387 : This is not strictly needed in theory, but is required by the
388 : current consensus protocol. They get published here in
389 : after_credit.
390 :
391 : (b) Microblocks. These can occur at any other hashcnt, as long
392 : as it is not a tick. Microblocks cannot be empty, and must
393 : have at least one transactions mixed in. These get
394 : published in after_frag.
395 :
396 : If hashcnt_per_tick is 1, then we are in low power mode and the
397 : following does not apply, since we can mix in transactions at any
398 : time.
399 :
400 : In the normal, non-low-power mode, though, we have to be careful
401 : to make sure that we do not publish microblocks on tick
402 : boundaries. To do that, we need to obey two rules:
403 : (i) after_credit must not leave hashcnt one before a tick
404 : boundary
405 : (ii) if after_credit begins one before a tick boundary, it must
406 : advance hashcnt and publish the tick
407 :
408 : There's some interplay between min_hashcnt and restricted_hashcnt
409 : here, and we need to show that there's always a value of
410 : target_hashcnt we can pick such that
411 : min_hashcnt <= target_hashcnt <= restricted_hashcnt.
412 : We'll prove this by induction for current_slot==0 and
413 : is_leader==true, since all other slots should be the same.
414 :
415 : Let m_j and r_j be the min_hashcnt and restricted_hashcnt
416 : (respectively) for the jth call to after_credit in a slot. We
417 : want to show that for all values of j, it's possible to pick a
418 : value h_j, the value of target_hashcnt for the jth call to
419 : after_credit (which is also the value of hashcnt after
420 : after_credit has completed) such that m_j<=h_j<=r_j.
421 :
422 : Additionally, let T be hashcnt_per_tick and N be ticks_per_slot.
423 :
424 : Starting with the base case, j==0. m_j=0, and
425 : r_0 = N*T - max_microblocks_per_slot
426 : - ceil(max_microblocks_per_slot/(T-1)).
427 :
428 : This is monotonic decreasing in max_microblocks_per_slot, so it
429 : achieves its minimum when max_microblocks_per_slot is its
430 : maximum.
431 : r_0 >= N*T - N*(T-1) - ceil( (N*(T-1))/(T-1))
432 : = N*T - N*(T-1)-N = 0.
433 : Thus, m_0 <= r_0, as desired.
434 :
435 :
436 :
437 : Then, for the inductive step, assume there exists h_j such that
438 : m_j<=h_j<=r_j, and we want to show that there exists h_{j+1},
439 : which is the same as showing m_{j+1}<=r_{j+1}.
440 :
441 : Let a_j be 1 if we had a microblock immediately following the jth
442 : call to after_credit, and 0 otherwise. Then hashcnt at the start
443 : of the (j+1)th call to after_frag is h_j+a_j.
444 : Also, set b_{j+1}=1 if we are in the case covered by rule (ii)
445 : above during the (j+1)th call to after_credit, i.e. if
446 : (h_j+a_j)%T==T-1. Thus, m_{j+1} = h_j + a_j + b_{j+1}.
447 :
448 : If we received an additional microblock, then
449 : max_remaining_microblocks goes down by 1, and
450 : max_remaining_ticks_or_microblocks goes down by either 1 or 2,
451 : which means restricted_hashcnt goes up by either 1 or 2. In
452 : particular, it goes up by 2 if the new value of
453 : max_remaining_microblocks (at the start of the (j+1)th call to
454 : after_credit) is congruent to 0 mod T-1. Let b'_{j+1} be 1 if
455 : this condition is met and 0 otherwise. If we receive a
456 : done_packing message, restricted_hashcnt can go up by more, but
457 : we can ignore that case, since it is less restrictive.
458 : Thus, r_{j+1}=r_j+a_j+b'_{j+1}.
459 :
460 : If h_j < r_j (strictly less), then h_j+a_j < r_j+a_j. And thus,
461 : since b_{j+1}<=b'_{j+1}+1, just by virtue of them both being
462 : binary,
463 : h_j + a_j + b_{j+1} < r_j + a_j + b'_{j+1} + 1,
464 : which is the same (for integers) as
465 : h_j + a_j + b_{j+1} <= r_j + a_j + b'_{j+1},
466 : m_{j+1} <= r_{j+1}
467 :
468 : On the other hand, if h_j==r_j, this is easy unless b_{j+1}==1,
469 : which can also only happen if a_j==1. Then (h_j+a_j)%T==T-1,
470 : which means there's an integer k such that
471 :
472 : h_j+a_j==(ticks_per_slot-k)*T-1
473 : h_j ==ticks_per_slot*T - k*(T-1)-1 - k-1
474 : ==ticks_per_slot*T - (k*(T-1)+1) - ceil( (k*(T-1)+1)/(T-1) )
475 :
476 : Since h_j==r_j in this case, and
477 : r_j==(ticks_per_slot*T) - max_remaining_microblocks_j - ceil(max_remaining_microblocks_j/(T-1)),
478 : we can see that the value of max_remaining_microblocks at the
479 : start of the jth call to after_credit is k*(T-1)+1. Again, since
480 : a_j==1, then the value of max_remaining_microblocks at the start
481 : of the j+1th call to after_credit decreases by 1 to k*(T-1),
482 : which means b'_{j+1}=1.
483 :
484 : Thus, h_j + a_j + b_{j+1} == r_j + a_j + b'_{j+1}, so, in
485 : particular, h_{j+1}<=r_{j+1} as desired. */
486 0 : min_hashcnt += (ulong)(min_hashcnt%poh->hashcnt_per_tick == (poh->hashcnt_per_tick-1UL)); /* add b_{j+1}, enforcing rule (ii) */
487 0 : }
488 : /* Now figure out how many hashes are needed to "catch up" the hash
489 : count to the current system clock, and clamp it to the allowed
490 : range. */
491 0 : long now = fd_log_wallclock();
492 0 : ulong target_hashcnt;
493 0 : if( FD_LIKELY( poh->state==STATE_FOLLOWER ||poh->state==STATE_WAITING_FOR_SLOT ) ) {
494 0 : target_hashcnt = (ulong)((double)(now - poh->reset_slot_start_ns) / poh->hashcnt_duration_ns) - (poh->slot-poh->reset_slot)*poh->hashcnt_per_slot;
495 0 : } else {
496 0 : FD_TEST( poh->state==STATE_LEADER );
497 0 : target_hashcnt = (ulong)((double)(now - poh->leader_slot_start_ns) / poh->hashcnt_duration_ns);
498 0 : }
499 : /* Clamp to [min_hashcnt, restricted_hashcnt] as above */
500 0 : target_hashcnt = fd_ulong_max( fd_ulong_min( target_hashcnt, restricted_hashcnt ), min_hashcnt );
501 :
502 : /* The above proof showed that it was always possible to pick a value
503 : of target_hashcnt, but we still have a lot of freedom in how to
504 : pick it. It simplifies the code a lot if we don't keep going after
505 : a tick in this function. In particular, we want to publish at most
506 : 1 tick in this call, since otherwise we could consume infinite
507 : credits to publish here. The credits are set so that we should
508 : only ever publish one tick during this loop. Also, all the extra
509 : stuff (leader transitions, publishing ticks, etc.) we have to do
510 : happens at tick boundaries, so this lets us consolidate all those
511 : cases.
512 :
513 : Mathematically, since the current value of hashcnt is h_j+a_j, the
514 : next tick (advancing a full tick if we're currently at a tick) is
515 : t_{j+1} = T*(floor( (h_j+a_j)/T )+1). We need to show that if we set
516 : h'_{j+1} = min( h_{j+1}, t_{j+1} ), it is still valid.
517 :
518 : First, h'_{j+1} <= h_{j+1} <= r_{j+1}, so we're okay in that
519 : direction.
520 :
521 : Next, observe that t_{j+1}>=h_j + a_j + 1, and recall that b_{j+1}
522 : is 0 or 1. So then,
523 : t_{j+1} >= h_j+a_j+b_{j+1} = m_{j+1}.
524 :
525 : We know h_{j+1) >= m_{j+1} from before, so then h'_{j+1} >=
526 : m_{j+1}, as desired. */
527 :
528 0 : ulong next_tick_hashcnt = poh->hashcnt_per_tick * (1UL+(poh->hashcnt/poh->hashcnt_per_tick));
529 0 : target_hashcnt = fd_ulong_min( target_hashcnt, next_tick_hashcnt );
530 :
531 : /* We still need to enforce rule (i). We know that min_hashcnt%T !=
532 : T-1 because of rule (ii). That means that if target_hashcnt%T ==
533 : T-1 at this point, target_hashcnt > min_hashcnt (notice the
534 : strict), so target_hashcnt-1 >= min_hashcnt and is thus still a
535 : valid choice for target_hashcnt. */
536 0 : target_hashcnt -= (ulong)( (!low_power_mode) & ((target_hashcnt%poh->hashcnt_per_tick)==(poh->hashcnt_per_tick-1UL)) );
537 :
538 0 : FD_TEST( target_hashcnt >= poh->hashcnt );
539 0 : FD_TEST( target_hashcnt >= min_hashcnt );
540 0 : FD_TEST( target_hashcnt <= restricted_hashcnt );
541 :
542 0 : if( FD_UNLIKELY( poh->hashcnt==target_hashcnt ) ) return; /* Nothing to do, don't publish a tick twice */
543 :
544 0 : *charge_busy = 1;
545 :
546 0 : if( FD_LIKELY( poh->hashcnt<target_hashcnt ) ) {
547 0 : fd_sha256_hash_32_repeated( poh->hash, poh->hash, target_hashcnt-poh->hashcnt );
548 0 : poh->hashcnt = target_hashcnt;
549 0 : }
550 :
551 0 : if( FD_UNLIKELY( poh->hashcnt==poh->hashcnt_per_slot ) ) {
552 0 : poh->slot++;
553 0 : poh->hashcnt = 0UL;
554 0 : }
555 :
556 0 : switch( poh->state ) {
557 0 : case STATE_LEADER: {
558 0 : if( FD_UNLIKELY( !(poh->hashcnt%poh->hashcnt_per_tick) ) ) {
559 : /* We ticked while leader... send an empty microblock (a tick)
560 : to the shred tile. */
561 0 : publish_tick( poh, stem, poh->hash, 0 );
562 0 : }
563 0 : if( FD_UNLIKELY( poh->slot>poh->next_leader_slot ) ) {
564 : /* We ticked while leader and are no longer leader... transition
565 : the state machine. */
566 0 : FD_TEST( !max_remaining_microblocks );
567 0 : FD_LOG_INFO(( "fd_poh_ticked_outof_leader(slot=%lu)", poh->slot-1UL ));
568 0 : transition_to_follower( poh, stem, 1 );
569 0 : poh->state = STATE_WAITING_FOR_RESET;
570 0 : }
571 0 : break;
572 0 : }
573 0 : case STATE_WAITING_FOR_SLOT:
574 0 : case STATE_FOLLOWER: {
575 0 : if( FD_UNLIKELY( !(poh->hashcnt%poh->hashcnt_per_tick ) ) ) {
576 : /* We finished a tick while not leader... save the current hash
577 : so it can be played back into the bank when we become the
578 : leader. */
579 0 : ulong tick_idx = (poh->slot*poh->ticks_per_slot+poh->hashcnt/poh->hashcnt_per_tick)%MAX_SKIPPED_TICKS;
580 0 : fd_memcpy( poh->skipped_tick_hashes[ tick_idx ], poh->hash, 32UL );
581 :
582 0 : ulong initial_tick_idx = (poh->last_slot*poh->ticks_per_slot+poh->last_hashcnt/poh->hashcnt_per_tick)%MAX_SKIPPED_TICKS;
583 0 : if( FD_UNLIKELY( tick_idx==initial_tick_idx ) ) FD_LOG_ERR(( "Too many skipped ticks from slot %lu to slot %lu, chain must halt", poh->last_slot, poh->slot ));
584 0 : }
585 :
586 0 : FD_TEST( poh->slot<=poh->next_leader_slot );
587 0 : if( FD_UNLIKELY( poh->slot==poh->next_leader_slot ) ) {
588 : /* We ticked while not leader and are now leader... transition
589 : the state machine. */
590 0 : if( FD_LIKELY( poh->state==STATE_FOLLOWER ) ) poh->state = STATE_WAITING_FOR_BANK;
591 0 : else poh->state = STATE_LEADER;
592 0 : }
593 0 : break;
594 0 : }
595 0 : default: {
596 0 : break;
597 0 : }
598 0 : }
599 0 : }
600 :
601 : static void
602 : publish_microblock( fd_poh_t * poh,
603 : fd_stem_context_t * stem,
604 : ulong slot,
605 : ulong hashcnt_delta,
606 : ulong txn_cnt,
607 0 : fd_txn_p_t const * txns ) {
608 0 : uchar * dst = (uchar *)fd_chunk_to_laddr( poh->shred_out->mem, poh->shred_out->chunk );
609 0 : FD_TEST( slot>=poh->reset_slot );
610 0 : fd_entry_batch_meta_t * meta = (fd_entry_batch_meta_t *)dst;
611 0 : meta->parent_offset = 1UL+slot-poh->reset_slot;
612 0 : meta->reference_tick = (poh->hashcnt/poh->hashcnt_per_tick) % poh->ticks_per_slot;
613 0 : meta->block_complete = !poh->hashcnt;
614 :
615 0 : meta->parent_block_id_valid = 1;
616 0 : fd_memcpy( meta->parent_block_id, poh->completed_block_id, 32UL );
617 :
618 0 : dst += sizeof(fd_entry_batch_meta_t);
619 0 : fd_entry_batch_header_t * header = (fd_entry_batch_header_t *)dst;
620 0 : header->hashcnt_delta = hashcnt_delta;
621 0 : fd_memcpy( header->hash, poh->hash, 32UL );
622 :
623 0 : dst += sizeof(fd_entry_batch_header_t);
624 0 : ulong payload_sz = 0UL;
625 0 : ulong included_txn_cnt = 0UL;
626 0 : for( ulong i=0UL; i<txn_cnt; i++ ) {
627 0 : fd_txn_p_t const * txn = txns + i;
628 0 : if( FD_UNLIKELY( !(txn->flags & FD_TXN_P_FLAGS_EXECUTE_SUCCESS) ) ) continue;
629 :
630 0 : fd_memcpy( dst, txn->payload, txn->payload_sz );
631 0 : payload_sz += txn->payload_sz;
632 0 : dst += txn->payload_sz;
633 0 : included_txn_cnt++;
634 0 : }
635 0 : header->txn_cnt = included_txn_cnt;
636 :
637 : /* We always have credits to publish here, because we have a burst
638 : value of 3 credits, and at most we will publish_tick() once and
639 : then publish_became_leader() once, leaving one credit here to
640 : publish the microblock. */
641 0 : ulong tspub = (ulong)fd_frag_meta_ts_comp( fd_tickcount() );
642 0 : ulong sz = sizeof(fd_entry_batch_meta_t)+sizeof(fd_entry_batch_header_t)+payload_sz;
643 0 : ulong new_sig = fd_disco_poh_sig( slot, POH_PKT_TYPE_MICROBLOCK, 0UL );
644 0 : fd_stem_publish( stem, poh->shred_out->idx, new_sig, poh->shred_out->chunk, sz, 0UL, 0UL, tspub );
645 0 : poh->shred_out->chunk = fd_dcache_compact_next( poh->shred_out->chunk, sz, poh->shred_out->chunk0, poh->shred_out->wmark );
646 0 : }
647 :
648 : void
649 : fd_poh1_mixin( fd_poh_t * poh,
650 : fd_stem_context_t * stem,
651 : ulong slot,
652 : uchar const * hash,
653 : ulong txn_cnt,
654 0 : fd_txn_p_t const * txns ) {
655 0 : if( FD_UNLIKELY( slot!=poh->next_leader_slot || slot!=poh->slot ) ) {
656 0 : FD_LOG_ERR(( "packed too early or late slot=%lu, current_slot=%lu", slot, poh->slot ));
657 0 : }
658 0 : if( FD_UNLIKELY( (poh->hashcnt%poh->hashcnt_per_tick)==(poh->hashcnt_per_tick-1UL) ) ) FD_LOG_CRIT(( "a tick will be skipped due to hashcnt %lu hashcnt_per_tick %lu", poh->hashcnt, poh->hashcnt_per_tick ));
659 :
660 0 : FD_TEST( poh->state==STATE_LEADER );
661 0 : FD_TEST( poh->microblocks_lower_bound<poh->max_microblocks_per_slot );
662 0 : poh->microblocks_lower_bound += 1UL;
663 :
664 0 : ulong executed_txn_cnt = 0UL;
665 0 : for( ulong i=0UL; i<txn_cnt; i++ ) {
666 : /* It's important that we check if a transaction is included in the
667 : block with FD_TXN_P_FLAGS_EXECUTE_SUCCESS since
668 : actual_consumed_cus may have a nonzero value for excluded
669 : transactions used for monitoring purposes */
670 0 : if( FD_LIKELY( txns[ i ].flags & FD_TXN_P_FLAGS_EXECUTE_SUCCESS ) ) {
671 0 : executed_txn_cnt++;
672 0 : }
673 0 : }
674 :
675 : /* We don't publish transactions that fail to execute. If all the
676 : transactions failed to execute, the microblock would be empty,
677 : causing agave to think it's a tick and complain. Instead, we just
678 : skip the microblock and don't hash or update the hashcnt. */
679 0 : if( FD_UNLIKELY( !executed_txn_cnt ) ) return;
680 :
681 0 : uchar data[ 64 ];
682 0 : fd_memcpy( data, poh->hash, 32UL );
683 0 : fd_memcpy( data+32UL, hash, 32UL );
684 0 : fd_sha256_hash( data, 64UL, poh->hash );
685 :
686 0 : poh->hashcnt++;
687 0 : FD_TEST( poh->hashcnt>poh->last_hashcnt );
688 0 : ulong hashcnt_delta = poh->hashcnt - poh->last_hashcnt;
689 :
690 : /* The hashing loop above will never leave us exactly one away from
691 : crossing a tick boundary, so this increment will never cause the
692 : current tick (or the slot) to change, except in low power mode
693 : for development, in which case we do need to register the tick
694 : with the leader bank. We don't need to publish the tick since
695 : sending the microblock below is the publishing action. */
696 0 : if( FD_UNLIKELY( !(poh->hashcnt%poh->hashcnt_per_slot ) ) ) {
697 0 : poh->slot++;
698 0 : poh->hashcnt = 0UL;
699 0 : }
700 :
701 0 : poh->last_slot = poh->slot;
702 0 : poh->last_hashcnt = poh->hashcnt;
703 :
704 0 : if( FD_UNLIKELY( !(poh->hashcnt%poh->hashcnt_per_tick ) ) ) {
705 0 : if( FD_UNLIKELY( poh->slot>poh->next_leader_slot ) ) {
706 : /* We ticked while leader and are no longer leader... transition
707 : the state machine. */
708 0 : transition_to_follower( poh, stem, 1 );
709 0 : poh->state = STATE_WAITING_FOR_RESET;
710 0 : }
711 0 : }
712 :
713 0 : publish_microblock( poh, stem, slot, hashcnt_delta, txn_cnt, txns );
714 0 : }
|
