Line data Source code
1 : #define _GNU_SOURCE
2 : #include "../../../../disco/tiles.h"
3 : #include "../../../../disco/plugin/fd_bundle_crank.h"
4 :
5 : /* Let's say there was a computer, the "leader" computer, that acted as
6 : a bank. Users could send it messages saying they wanted to deposit
7 : money, or transfer it to someone else.
8 :
9 : That's how, for example, Bank of America works but there are problems
10 : with it. One simple problem is: the bank can set your balance to
11 : zero if they don't like you.
12 :
13 : You could try to fix this by having the bank periodically publish the
14 : list of all account balances and transactions. If the customers add
15 : unforgeable signatures to their deposit slips and transfers, then
16 : the bank cannot zero a balance without it being obvious to everyone.
17 :
18 : There's still problems. The bank can't lie about your balance now or
19 : take your money, but it can just not accept deposits on your behalf
20 : by ignoring you.
21 :
22 : You could fix this by getting a few independent banks together, lets
23 : say Bank of America, Bank of England, and Westpac, and having them
24 : rotate who operates the leader computer periodically. If one bank
25 : ignores your deposits, you can just wait and send them to the next
26 : one.
27 :
28 : This is Solana.
29 :
30 : There's still problems of course but they are largely technical. How
31 : do the banks agree who is leader? How do you recover if a leader
32 : misbehaves? How do customers verify the transactions aren't forged?
33 : How do banks receive and publish and verify each others work quickly?
34 : These are the main technical innovations that enable Solana to work
35 : well.
36 :
37 : What about Proof of History?
38 :
39 : One particular niche problem is about the leader schedule. When the
40 : leader computer is moving from one bank to another, the new bank must
41 : wait for the old bank to say it's done and provide a final list of
42 : balances that it can start working off of. But: what if the computer
43 : at the old bank crashes and never says its done?
44 :
45 : Does the new leader just take over at some point? What if the new
46 : leader is malicious, and says the past thousand leaders crashed, and
47 : there have been no transactions for days? How do you check?
48 :
49 : This is what Proof of History solves. Each bank in the network must
50 : constantly do a lot of busywork (compute hashes), even when it is not
51 : leader.
52 :
53 : If the prior thousand leaders crashed, and no transactions happened
54 : in an hour, the new leader would have to show they did about an hour
55 : of busywork for everyone else to believe them.
56 :
57 : A better name for this is proof of skipping. If a leader is skipping
58 : slots (building off of a slot that is not the direct parent), it must
59 : prove that it waited a good amount of time to do so.
60 :
61 : It's not a perfect solution. For one thing, some banks have really
62 : fast computers and can compute a lot of busywork in a short amount of
63 : time, allowing them to skip prior slot(s) anyway. But: there is a
64 : social component that prevents validators from skipping the prior
65 : leader slot. It is easy to detect when this happens and the network
66 : could respond by ignoring their votes or stake.
67 :
68 : You could come up with other schemes: for example, the network could
69 : just use wall clock time. If a new leader publishes a block without
70 : waiting 400 milliseconds for the prior slot to complete, then there
71 : is no "proof of skipping" and the nodes ignore the slot.
72 :
73 : These schemes have a problem in that they are not deterministic
74 : across the network (different computers have different clocks), and
75 : so they will cause frequent forks which are very expensive to
76 : resolve. Even though the proof of history scheme is not perfect,
77 : it is better than any alternative which is not deterministic.
78 :
79 : With all that background, we can now describe at a high level what
80 : this PoH tile actually does,
81 :
82 : (1) Whenever any other leader in the network finishes a slot, and
83 : the slot is determined to be the best one to build off of, this
84 : tile gets "reset" onto that block, the so called "reset slot".
85 :
86 : (2) The tile is constantly doing busy work, hash(hash(hash(...))) on
87 : top of the last reset slot, even when it is not leader.
88 :
89 : (3) When the tile becomes leader, it continues hashing from where it
90 : was. Typically, the prior leader finishes their slot, so the
91 : reset slot will be the parent one, and this tile only publishes
92 : hashes for its own slot. But if prior slots were skipped, then
93 : there might be a whole chain already waiting.
94 :
95 : That's pretty much it. When we are leader, in addition to doing
96 : busywork, we publish ticks and microblocks to the shred tile. A
97 : microblock is a non-empty group of transactions whose hashes are
98 : mixed-in to the chain, while a tick is a periodic stamp of the
99 : current hash, with no transactions (nothing mixed in). We need
100 : to send both to the shred tile, as ticks are important for other
101 : validators to verify in parallel.
102 :
103 : As well, the tile should never become leader for a slot that it has
104 : published anything for, otherwise it may create a duplicate block.
105 :
106 : Some particularly common misunderstandings:
107 :
108 : - PoH is critical to security.
109 :
110 : This largely isn't true. The target hash rate of the network is
111 : so slow (1 hash per 500 nanoseconds) that a malicious leader can
112 : easily catch up if they start from an old hash, and the only
113 : practical attack prevented is the proof of skipping. Most of the
114 : long range attacks in the Solana whitepaper are not relevant.
115 :
116 : - PoH keeps passage of time.
117 :
118 : This is also not true. The way the network keeps time so it can
119 : decide who is leader is that, each leader uses their operating
120 : system clock to time 400 milliseconds and publishes their block
121 : when this timer expires.
122 :
123 : If a leader just hashed as fast as they could, they could publish
124 : a block in tens of milliseconds, and the rest of the network
125 : would happily accept it. This is why the Solana "clock" as
126 : determined by PoH is not accurate and drifts over time.
127 :
128 : - PoH prevents transaction reordering by the leader.
129 :
130 : The leader can, in theory, wait until the very end of their
131 : leader slot to publish anything at all to the network. They can,
132 : in particular, hold all received transactions for 400
133 : milliseconds and then reorder and publish some right at the end
134 : to advantage certain transactions.
135 :
136 : You might be wondering... if all the PoH chain is helping us do is
137 : prove that slots were skipped correctly, why do we need to "mix in"
138 : transactions to the hash value? Or do anything at all for slots
139 : where we don't skip the prior slot?
140 :
141 : It's a good question, and the answer is that this behavior is not
142 : necessary. An ideal implementation of PoH have no concept of ticks
143 : or mixins, and would not be part of the TPU pipeline at all.
144 : Instead, there would be a simple field "skip_proof" on the last
145 : shred we send for a slot, the hash(hash(...)) value. This field
146 : would only be filled in (and only verified by replayers) in cases
147 : where the slot actually skipped a parent.
148 :
149 : Then what is the "clock? In Solana, time is constructed as follows:
150 :
151 : HASHES
152 :
153 : The base unit of time is a hash. Hereafter, any values whose
154 : units are in hashes are called a "hashcnt" to distinguish them
155 : from actual hashed values.
156 :
157 : Agave generally defines a constant duration for each tick
158 : (see below) and then varies the number of hashcnt per tick, but
159 : as we consider the hashcnt the base unit of time, Firedancer and
160 : this PoH implementation defines everything in terms of hashcnt
161 : duration instead.
162 :
163 : In mainnet-beta, testnet, and devnet the hashcnt ticks over
164 : (increments) every 100 nanoseconds. The hashcnt rate is
165 : specified as 500 nanoseconds according to the genesis, but there
166 : are several features which increase the number of hashes per
167 : tick while keeping tick duration constant, which make the time
168 : per hashcnt lower. These features up to and including the
169 : `update_hashes_per_tick6` feature are activated on mainnet-beta,
170 : devnet, and testnet, and are described in the TICKS section
171 : below.
172 :
173 : Other chains and development environments might have a different
174 : hashcnt rate in the genesis, or they might not have activated
175 : the features which increase the rate yet, which we also support.
176 :
177 : In practice, although each validator follows a hashcnt rate of
178 : 100 nanoseconds, the overall observed hashcnt rate of the
179 : network is a little slower than once every 100 nanoseconds,
180 : mostly because there are gaps and clock synchronization issues
181 : during handoff between leaders. This is referred to as clock
182 : drift.
183 :
184 : TICKS
185 :
186 : The leader needs to periodically checkpoint the hash value
187 : associated with a given hashcnt so that they can publish it to
188 : other nodes for verification.
189 :
190 : On mainnet-beta, testnet, and devnet this occurs once every
191 : 62,500 hashcnts, or approximately once every 6.4 microseconds.
192 : This value is determined at genesis time, and according to the
193 : features below, and could be different in development
194 : environments or on other chains which we support.
195 :
196 : Due to protocol limitations, when mixing in transactions to the
197 : proof-of-history chain, it cannot occur on a tick boundary (but
198 : can occur at any other hashcnt).
199 :
200 : Ticks exist mainly so that verification can happen in parallel.
201 : A verifier computer, rather than needing to do hash(hash(...))
202 : all in sequence to verify a proof-of-history chain, can do,
203 :
204 : Core 0: hash(hash(...))
205 : Core 1: hash(hash(...))
206 : Core 2: hash(hash(...))
207 : Core 3: hash(hash(...))
208 : ...
209 :
210 : Between each pair of tick boundaries.
211 :
212 : Solana sometimes calls the current tick the "tick height",
213 : although it makes more sense to think of it as a counter from
214 : zero, it's just the number of ticks since the genesis hash.
215 :
216 : There is a set of features which increase the number of hashcnts
217 : per tick. These are all deployed on mainnet-beta, devnet, and
218 : testnet.
219 :
220 : name: update_hashes_per_tick
221 : id: 3uFHb9oKdGfgZGJK9EHaAXN4USvnQtAFC13Fh5gGFS5B
222 : hashes per tick: 12,500
223 : hashcnt duration: 500 nanos
224 :
225 : name: update_hashes_per_tick2
226 : id: EWme9uFqfy1ikK1jhJs8fM5hxWnK336QJpbscNtizkTU
227 : hashes per tick: 17,500
228 : hashcnt duration: 357.142857143 nanos
229 :
230 : name: update_hashes_per_tick3
231 : id: 8C8MCtsab5SsfammbzvYz65HHauuUYdbY2DZ4sznH6h5
232 : hashes per tick: 27,500
233 : hashcnt duration: 227.272727273 nanos
234 :
235 : name: update_hashes_per_tick4
236 : id: 8We4E7DPwF2WfAN8tRTtWQNhi98B99Qpuj7JoZ3Aikgg
237 : hashes per tick: 47,500
238 : hashcnt duration: 131.578947368 nanos
239 :
240 : name: update_hashes_per_tick5
241 : id: BsKLKAn1WM4HVhPRDsjosmqSg2J8Tq5xP2s2daDS6Ni4
242 : hashes per tick: 57,500
243 : hashcnt duration: 108.695652174 nanos
244 :
245 : name: update_hashes_per_tick6
246 : id: FKu1qYwLQSiehz644H6Si65U5ZQ2cp9GxsyFUfYcuADv
247 : hashes per tick: 62,500
248 : hashcnt duration: 100 nanos
249 :
250 : In development environments, there is a way to configure the
251 : hashcnt per tick to be "none" during genesis, for a so-called
252 : "low power" tick producer. The idea is not to spin cores during
253 : development. This is equivalent to setting the hashcnt per tick
254 : to be 1, and increasing the hashcnt duration to the desired tick
255 : duration.
256 :
257 : SLOTS
258 :
259 : Each leader needs to be leader for a fixed amount of time, which
260 : is called a slot. During a slot, a leader has an opportunity to
261 : receive transactions and produce a block for the network,
262 : although they may miss ("skip") the slot if they are offline or
263 : not behaving.
264 :
265 : In mainnet-beta, testnet, and devnet a slot is 64 ticks, or
266 : 4,000,000 hashcnts, or approximately 400 milliseconds.
267 :
268 : Due to the way the leader schedule is constructed, each leader
269 : is always given at least four (4) consecutive slots in the
270 : schedule. This means when becoming leader you will be leader
271 : for at least 4 slots, or 1.6 seconds.
272 :
273 : It is rare, although can happen that a leader gets more than 4
274 : consecutive slots (eg, 8, or 12), if they are lucky with the
275 : leader schedule generation.
276 :
277 : The number of ticks in a slot is fixed at genesis time, and
278 : could be different for development or other chains, which we
279 : support. There is nothing special about 4 leader slots in a
280 : row, and this might be changed in future, and the proof of
281 : history makes no assumptions that this is the case.
282 :
283 : EPOCHS
284 :
285 : Infrequently, the network needs to do certain housekeeping,
286 : mainly things like collecting rent and deciding on the leader
287 : schedule. The length of an epoch is fixed on mainnet-beta,
288 : devnet and testnet at 420,000 slots, or around ~2 (1.94) days.
289 : This value is fixed at genesis time, and could be different for
290 : other chains including development, which we support. Typically
291 : in development, epochs are every 8,192 slots, or around ~1 hour
292 : (54.61 minutes), although it depends on the number of ticks per
293 : slot and the target hashcnt rate of the genesis as well.
294 :
295 : In development, epochs need not be a fixed length either. There
296 : is a "warmup" option, where epochs start short and grow, which
297 : is useful for quickly warming up stake during development.
298 :
299 : The epoch is important because it is the only time the leader
300 : schedule is updated. The leader schedule is a list of which
301 : leader is leader for which slot, and is generated by a special
302 : algorithm that is deterministic and known to all nodes.
303 :
304 : The leader schedule is computed one epoch in advance, so that
305 : at slot T, we always know who will be leader up until the end
306 : of slot T+EPOCH_LENGTH. Specifically, the leader schedule for
307 : epoch N is computed during the epoch boundary crossing from
308 : N-2 to N-1. For mainnet-beta, the slots per epoch is fixed and
309 : will always be 420,000. */
310 :
311 : #include "../../../../disco/pack/fd_pack.h"
312 : #include "../../../../ballet/sha256/fd_sha256.h"
313 : #include "../../../../disco/metrics/fd_metrics.h"
314 : #include "../../../../disco/topo/fd_pod_format.h"
315 : #include "../../../../disco/shred/fd_shredder.h"
316 : #include "../../../../disco/shred/fd_stake_ci.h"
317 : #include "../../../../disco/bank/fd_bank_abi.h"
318 : #include "../../../../disco/keyguard/fd_keyload.h"
319 : #include "../../../../disco/keyguard/fd_keyswitch.h"
320 : #include "../../../../disco/metrics/generated/fd_metrics_poh.h"
321 : #include "../../../../disco/plugin/fd_plugin.h"
322 : #include "../../../../flamenco/leaders/fd_leaders.h"
323 :
324 : #include <string.h>
325 :
326 : /* The maximum number of microblocks that pack is allowed to pack into a
327 : single slot. This is not consensus critical, and pack could, if we
328 : let it, produce as many microblocks as it wants, and the slot would
329 : still be valid.
330 :
331 : We have this here instead so that PoH can estimate slot completion,
332 : and keep the hashcnt up to date as pack progresses through packing
333 : the slot. If this upper bound was not enforced, PoH could tick to
334 : the last hash of the slot and have no hashes left to mixin incoming
335 : microblocks from pack, so this upper bound is a coordination
336 : mechanism so that PoH can progress hashcnts while the slot is active,
337 : and know that pack will not need those hashcnts later to do mixins. */
338 0 : #define MAX_MICROBLOCKS_PER_SLOT (32768UL)
339 :
340 : /* When we are hashing in the background in case a prior leader skips
341 : their slot, we need to store the result of each tick hash so we can
342 : publish them when we become leader. The network requires at least
343 : one leader slot to publish in each epoch for the leader schedule to
344 : generate, so in the worst case we might need two full epochs of slots
345 : to store the hashes. (Eg, if epoch T only had a published slot in
346 : position 0 and epoch T+1 only had a published slot right at the end).
347 :
348 : There is a tighter bound: the block data limit of mainnet-beta is
349 : currently FD_PACK_MAX_DATA_PER_BLOCK, or 27,332,342 bytes per slot.
350 : At 48 bytes per tick, it is not possible to publish a slot that skips
351 : 569,424 or more prior slots. */
352 0 : #define MAX_SKIPPED_TICKS (1UL+(FD_PACK_MAX_DATA_PER_BLOCK/48UL))
353 :
354 0 : #define IN_KIND_BANK (0)
355 0 : #define IN_KIND_PACK (1)
356 0 : #define IN_KIND_STAKE (2)
357 :
358 :
359 : typedef struct {
360 : fd_wksp_t * mem;
361 : ulong chunk0;
362 : ulong wmark;
363 : } fd_poh_in_ctx_t;
364 :
365 : typedef struct {
366 : ulong idx;
367 : fd_wksp_t * mem;
368 : ulong chunk0;
369 : ulong wmark;
370 : ulong chunk;
371 : } fd_poh_out_ctx_t;
372 :
373 : typedef struct {
374 : fd_stem_context_t * stem;
375 :
376 : /* Static configuration determined at genesis creation time. See
377 : long comment above for more information. */
378 : ulong tick_duration_ns;
379 : ulong hashcnt_per_tick;
380 : ulong ticks_per_slot;
381 :
382 : /* Derived from the above configuration, but we precompute it. */
383 : double slot_duration_ns;
384 : double hashcnt_duration_ns;
385 : ulong hashcnt_per_slot;
386 : /* Constant, fixed at initialization. The maximum number of
387 : microblocks that the pack tile can publish in each slot. */
388 : ulong max_microblocks_per_slot;
389 :
390 : /* The current slot and hashcnt within that slot of the proof of
391 : history, including hashes we have been producing in the background
392 : while waiting for our next leader slot. */
393 : ulong slot;
394 : ulong hashcnt;
395 : ulong cus_used;
396 :
397 : /* When we send a microblock on to the shred tile, we need to tell
398 : it how many hashes there have been since the last microblock, so
399 : this tracks the hashcnt of the last published microblock.
400 :
401 : If we are skipping slots prior to our leader slot, the last_slot
402 : will be quite old, and potentially much larger than the number of
403 : hashcnts in one slot. */
404 : ulong last_slot;
405 : ulong last_hashcnt;
406 :
407 : /* If we have published a tick or a microblock for a particular slot
408 : to the shred tile, we should never become leader for that slot
409 : again, otherwise we could publish a duplicate block.
410 :
411 : This value tracks the max slot that we have published a tick or
412 : microblock for so we can prevent this. */
413 : ulong highwater_leader_slot;
414 :
415 : /* See how this field is used below. If we have sequential leader
416 : slots, we don't reset the expected slot end time between the two,
417 : to prevent clock drift. If we didn't do this, our 2nd slot would
418 : end 400ms + `time_for_replay_to_move_slot_and_reset_poh` after
419 : our 1st, rather than just strictly 400ms. */
420 : int lagged_consecutive_leader_start;
421 : ulong expect_sequential_leader_slot;
422 :
423 : /* There's a race condition ... let's say two banks A and B, bank A
424 : processes some transactions, then releases the account locks, and
425 : sends the microblock to PoH to be stamped. Pack now re-packs the
426 : same accounts with a new microblock, sends to bank B, bank B
427 : executes and sends the microblock to PoH, and this all happens fast
428 : enough that PoH picks the 2nd block to stamp before the 1st. The
429 : accounts database changes now are misordered with respect to PoH so
430 : replay could fail.
431 :
432 : To prevent this race, we order all microblocks and only process
433 : them in PoH in the order they are produced by pack. This is a
434 : little bit over-strict, we just need to ensure that microblocks
435 : with conflicting accounts execute in order, but this is easiest to
436 : implement for now. */
437 : ulong expect_microblock_idx;
438 :
439 : /* The PoH tile must never drop microblocks that get committed by the
440 : bank, so it needs to always be able to mixin a microblock hash.
441 : Mixing in requires incrementing the hashcnt, so we need to ensure
442 : at all times that there is enough hascnts left in the slot to
443 : mixin whatever future microblocks pack might produce for it.
444 :
445 : This value tracks that. At any time, max_microblocks_per_slot
446 : - microblocks_lower_bound is an upper bound on the maximum number
447 : of microblocks that might still be received in this slot. */
448 : ulong microblocks_lower_bound;
449 :
450 : uchar __attribute__((aligned(32UL))) reset_hash[ 32 ];
451 : uchar __attribute__((aligned(32UL))) hash[ 32 ];
452 :
453 : /* When we are not leader, we need to save the hashes that were
454 : produced in case the prior leader skips. If they skip, we will
455 : replay these skipped hashes into our next leader bank so that
456 : the slot hashes sysvar can be updated correctly, and also publish
457 : them to peer nodes as part of our outgoing shreds. */
458 : uchar skipped_tick_hashes[ MAX_SKIPPED_TICKS ][ 32 ];
459 :
460 : /* The timestamp in nanoseconds of when the reset slot was received.
461 : This is the timestamp we are building on top of to determine when
462 : our next leader slot starts. */
463 : long reset_slot_start_ns;
464 :
465 : /* The timestamp in nanoseconds of when we got the bank for the
466 : current leader slot. */
467 : long leader_bank_start_ns;
468 :
469 : /* The hashcnt corresponding to the start of the current reset slot. */
470 : ulong reset_slot;
471 :
472 : /* The hashcnt at which our next leader slot begins, or ULONG max if
473 : we have no known next leader slot. */
474 : ulong next_leader_slot;
475 :
476 : /* If an in progress frag should be skipped */
477 : int skip_frag;
478 :
479 : ulong max_active_descendant;
480 :
481 : /* If we currently are the leader according the clock AND we have
482 : received the leader bank for the slot from the replay stage,
483 : this value will be non-NULL.
484 :
485 : Note that we might be inside our leader slot, but not have a bank
486 : yet, in which case this will still be NULL.
487 :
488 : It will be NULL for a brief race period between consecutive leader
489 : slots, as we ping-pong back to replay stage waiting for a new bank.
490 :
491 : Agave refers to this as the "working bank". */
492 : void const * current_leader_bank;
493 :
494 : fd_sha256_t * sha256;
495 :
496 : fd_stake_ci_t * stake_ci;
497 :
498 : /* The last sequence number of an outgoing fragment to the shred tile,
499 : or ULONG max if no such fragment. See fd_keyswitch.h for details
500 : of how this is used. */
501 : ulong shred_seq;
502 :
503 : int halted_switching_key;
504 :
505 : fd_keyswitch_t * keyswitch;
506 : fd_pubkey_t identity_key;
507 :
508 : /* We need a few pieces of information to compute the right addresses
509 : for bundle crank information that we need to send to pack. */
510 : struct {
511 : int enabled;
512 : fd_pubkey_t vote_account;
513 : fd_bundle_crank_gen_t gen[1];
514 : } bundle;
515 :
516 :
517 : /* The Agave client needs to be notified when the leader changes,
518 : so that they can resume the replay stage if it was suspended waiting. */
519 : void * signal_leader_change;
520 :
521 : /* These are temporarily set in during_frag so they can be used in
522 : after_frag once the frag has been validated as not overrun. */
523 : uchar _txns[ USHORT_MAX ];
524 : fd_microblock_trailer_t _microblock_trailer[ 1 ];
525 :
526 : int in_kind[ 64 ];
527 : fd_poh_in_ctx_t in[ 64 ];
528 :
529 : fd_poh_out_ctx_t shred_out[ 1 ];
530 : fd_poh_out_ctx_t pack_out[ 1 ];
531 : fd_poh_out_ctx_t plugin_out[ 1 ];
532 :
533 : fd_histf_t begin_leader_delay[ 1 ];
534 : fd_histf_t first_microblock_delay[ 1 ];
535 : fd_histf_t slot_done_delay[ 1 ];
536 : fd_histf_t bundle_init_delay[ 1 ];
537 : } fd_poh_ctx_t;
538 :
539 : /* The PoH recorder is implemented in Firedancer but for now needs to
540 : work with Agave, so we have a locking scheme for them to
541 : co-operate.
542 :
543 : This is because the PoH tile lives in the Agave memory address
544 : space and their version of concurrency is locking the PoH recorder
545 : and reading arbitrary fields.
546 :
547 : So we allow them to lock the PoH tile, although with a very bad (for
548 : them) locking scheme. By default, the tile has full and exclusive
549 : access to the data. If part of Agave wishes to read/write they
550 : can either,
551 :
552 : 1. Rewrite their concurrency to message passing based on mcache
553 : (preferred, but not feasible).
554 : 2. Signal to the tile they wish to acquire the lock, by setting
555 : fd_poh_waiting_lock to 1.
556 :
557 : During after_credit, the tile will check if the waiting lock is set
558 : to 1, and if so, set the returned lock to 1, indicating to the waiter
559 : that they may now proceed.
560 :
561 : When the waiter is done reading and writing, they restore the
562 : returned lock value back to zero, and the POH tile continues with its
563 : day. */
564 :
565 : static fd_poh_ctx_t * fd_poh_global_ctx;
566 :
567 : static volatile ulong fd_poh_waiting_lock __attribute__((aligned(128UL)));
568 : static volatile ulong fd_poh_returned_lock __attribute__((aligned(128UL)));
569 :
570 : /* Agave also needs to write to some mcaches, so we trampoline
571 : that via. the PoH tile as well. */
572 :
573 : struct poh_link {
574 : fd_frag_meta_t * mcache;
575 : ulong depth;
576 : ulong tx_seq;
577 :
578 : void * mem;
579 : void * dcache;
580 : ulong chunk0;
581 : ulong wmark;
582 : ulong chunk;
583 :
584 : ulong cr_avail;
585 : ulong rx_cnt;
586 : ulong * rx_fseqs[ 32UL ];
587 : };
588 :
589 : typedef struct poh_link poh_link_t;
590 :
591 : poh_link_t gossip_dedup;
592 : poh_link_t stake_out;
593 : poh_link_t crds_shred;
594 : poh_link_t replay_resolv;
595 :
596 : poh_link_t replay_plugin;
597 : poh_link_t gossip_plugin;
598 : poh_link_t start_progress_plugin;
599 : poh_link_t vote_listener_plugin;
600 : poh_link_t validator_info_plugin;
601 :
602 : static void
603 0 : poh_link_wait_credit( poh_link_t * link ) {
604 0 : if( FD_LIKELY( link->cr_avail ) ) return;
605 :
606 0 : while( 1 ) {
607 0 : ulong cr_query = ULONG_MAX;
608 0 : for( ulong i=0UL; i<link->rx_cnt; i++ ) {
609 0 : ulong const * _rx_seq = link->rx_fseqs[ i ];
610 0 : ulong rx_seq = FD_VOLATILE_CONST( *_rx_seq );
611 0 : ulong rx_cr_query = (ulong)fd_long_max( (long)link->depth - fd_long_max( fd_seq_diff( link->tx_seq, rx_seq ), 0L ), 0L );
612 0 : cr_query = fd_ulong_min( rx_cr_query, cr_query );
613 0 : }
614 0 : if( FD_LIKELY( cr_query>0UL ) ) {
615 0 : link->cr_avail = cr_query;
616 0 : break;
617 0 : }
618 0 : FD_SPIN_PAUSE();
619 0 : }
620 0 : }
621 :
622 : static void
623 : poh_link_publish( poh_link_t * link,
624 : ulong sig,
625 : uchar const * data,
626 0 : ulong data_sz ) {
627 0 : while( FD_UNLIKELY( !FD_VOLATILE_CONST( link->mcache ) ) ) FD_SPIN_PAUSE();
628 0 : if( FD_UNLIKELY( !link->mem ) ) return; /* link not enabled, don't publish */
629 0 : poh_link_wait_credit( link );
630 :
631 0 : uchar * dst = (uchar *)fd_chunk_to_laddr( link->mem, link->chunk );
632 0 : fd_memcpy( dst, data, data_sz );
633 0 : ulong tspub = (ulong)fd_frag_meta_ts_comp( fd_tickcount() );
634 0 : fd_mcache_publish( link->mcache, link->depth, link->tx_seq, sig, link->chunk, data_sz, 0UL, 0UL, tspub );
635 0 : link->chunk = fd_dcache_compact_next( link->chunk, data_sz, link->chunk0, link->wmark );
636 0 : link->cr_avail--;
637 0 : link->tx_seq++;
638 0 : }
639 :
640 : static void
641 : poh_link_init( poh_link_t * link,
642 : fd_topo_t * topo,
643 : fd_topo_tile_t * tile,
644 0 : ulong out_idx ) {
645 0 : fd_topo_link_t * topo_link = &topo->links[ tile->out_link_id[ out_idx ] ];
646 0 : fd_topo_wksp_t * wksp = &topo->workspaces[ topo->objs[ topo_link->dcache_obj_id ].wksp_id ];
647 :
648 0 : link->mem = wksp->wksp;
649 0 : link->depth = fd_mcache_depth( topo_link->mcache );
650 0 : link->tx_seq = 0UL;
651 0 : link->dcache = topo_link->dcache;
652 0 : link->chunk0 = fd_dcache_compact_chunk0( wksp->wksp, topo_link->dcache );
653 0 : link->wmark = fd_dcache_compact_wmark ( wksp->wksp, topo_link->dcache, topo_link->mtu );
654 0 : link->chunk = link->chunk0;
655 0 : link->cr_avail = 0UL;
656 0 : link->rx_cnt = 0UL;
657 0 : for( ulong i=0UL; i<topo->tile_cnt; i++ ) {
658 0 : fd_topo_tile_t * _tile = &topo->tiles[ i ];
659 0 : for( ulong j=0UL; j<_tile->in_cnt; j++ ) {
660 0 : if( _tile->in_link_id[ j ]==topo_link->id && _tile->in_link_reliable[ j ] ) {
661 0 : FD_TEST( link->rx_cnt<32UL );
662 0 : link->rx_fseqs[ link->rx_cnt++ ] = _tile->in_link_fseq[ j ];
663 0 : break;
664 0 : }
665 0 : }
666 0 : }
667 0 : FD_COMPILER_MFENCE();
668 0 : link->mcache = topo_link->mcache;
669 0 : FD_COMPILER_MFENCE();
670 0 : FD_TEST( link->mcache );
671 0 : }
672 :
673 : /* To help show correctness, functions that might be called from
674 : Rust, either directly or indirectly, have this fake "attribute"
675 : CALLED_FROM_RUST, which is actually nothing. Calls from Rust
676 : typically execute on threads did not call fd_boot, so they do not
677 : have the typical FD_TL variables. In particular, they cannot use
678 : normal metrics, and their log messages don't have full context.
679 : Additionally, Rust functions marked CALLED_FROM_RUST cannot call back
680 : into a C fd_ext function without causing a deadlock (although the
681 : other Rust fd_ext functions have a similar problem).
682 :
683 : To prevent annotation from polluting the whole codebase, calls to
684 : functions outside this file are manually checked and marked as being
685 : safe at each call rather than annotated. */
686 : #define CALLED_FROM_RUST
687 :
688 : static CALLED_FROM_RUST fd_poh_ctx_t *
689 0 : fd_ext_poh_write_lock( void ) {
690 0 : for(;;) {
691 : /* Acquire the waiter lock to make sure we are the first writer in the queue. */
692 0 : if( FD_LIKELY( !FD_ATOMIC_CAS( &fd_poh_waiting_lock, 0UL, 1UL) ) ) break;
693 0 : FD_SPIN_PAUSE();
694 0 : }
695 0 : FD_COMPILER_MFENCE();
696 0 : for(;;) {
697 : /* Now wait for the tile to tell us we can proceed. */
698 0 : if( FD_LIKELY( FD_VOLATILE_CONST( fd_poh_returned_lock ) ) ) break;
699 0 : FD_SPIN_PAUSE();
700 0 : }
701 0 : FD_COMPILER_MFENCE();
702 0 : return fd_poh_global_ctx;
703 0 : }
704 :
705 : static CALLED_FROM_RUST void
706 0 : fd_ext_poh_write_unlock( void ) {
707 0 : FD_COMPILER_MFENCE();
708 0 : FD_VOLATILE( fd_poh_returned_lock ) = 0UL;
709 0 : }
710 :
711 : /* The PoH tile needs to interact with the Agave address space to
712 : do certain operations that Firedancer hasn't reimplemented yet, a.k.a
713 : transaction execution. We have Agave export some wrapper
714 : functions that we call into during regular tile execution. These do
715 : not need any locking, since they are called serially from the single
716 : PoH tile. */
717 :
718 : extern CALLED_FROM_RUST void fd_ext_bank_acquire( void const * bank );
719 : extern CALLED_FROM_RUST void fd_ext_bank_release( void const * bank );
720 : extern CALLED_FROM_RUST void fd_ext_poh_signal_leader_change( void * sender );
721 : extern void fd_ext_poh_register_tick( void const * bank, uchar const * hash );
722 :
723 : /* fd_ext_poh_initialize is called by Agave on startup to
724 : initialize the PoH tile with some static configuration, and the
725 : initial reset slot and hash which it retrieves from a snapshot.
726 :
727 : This function is called by some random Agave thread, but
728 : it blocks booting of the PoH tile. The tile will spin until it
729 : determines that this initialization has happened.
730 :
731 : signal_leader_change is an opaque Rust object that is used to
732 : tell the replay stage that the leader has changed. It is a
733 : Box::into_raw(Arc::increment_strong(crossbeam::Sender)), so it
734 : has infinite lifetime unless this C code releases the refcnt.
735 :
736 : It can be used with `fd_ext_poh_signal_leader_change` which
737 : will just issue a nonblocking send on the channel. */
738 :
739 : CALLED_FROM_RUST void
740 : fd_ext_poh_initialize( ulong tick_duration_ns, /* See clock comments above, will be 6.4 microseconds for mainnet-beta. */
741 : ulong hashcnt_per_tick, /* See clock comments above, will be 62,500 for mainnet-beta. */
742 : ulong ticks_per_slot, /* See clock comments above, will almost always be 64. */
743 : ulong tick_height, /* The counter (height) of the tick to start hashing on top of. */
744 : uchar const * last_entry_hash, /* Points to start of a 32 byte region of memory, the hash itself at the tick height. */
745 0 : void * signal_leader_change /* See comment above. */ ) {
746 0 : FD_COMPILER_MFENCE();
747 0 : for(;;) {
748 : /* Make sure the ctx is initialized before trying to take the lock. */
749 0 : if( FD_LIKELY( FD_VOLATILE_CONST( fd_poh_global_ctx ) ) ) break;
750 0 : FD_SPIN_PAUSE();
751 0 : }
752 0 : fd_poh_ctx_t * ctx = fd_ext_poh_write_lock();
753 :
754 0 : ctx->slot = tick_height/ticks_per_slot;
755 0 : ctx->hashcnt = 0UL;
756 0 : ctx->cus_used = 0UL;
757 0 : ctx->last_slot = ctx->slot;
758 0 : ctx->last_hashcnt = 0UL;
759 0 : ctx->reset_slot = ctx->slot;
760 0 : ctx->reset_slot_start_ns = fd_log_wallclock(); /* safe to call from Rust */
761 :
762 0 : memcpy( ctx->reset_hash, last_entry_hash, 32UL );
763 0 : memcpy( ctx->hash, last_entry_hash, 32UL );
764 :
765 0 : ctx->signal_leader_change = signal_leader_change;
766 :
767 : /* Static configuration about the clock. */
768 0 : ctx->tick_duration_ns = tick_duration_ns;
769 0 : ctx->hashcnt_per_tick = hashcnt_per_tick;
770 0 : ctx->ticks_per_slot = ticks_per_slot;
771 :
772 : /* Recompute derived information about the clock. */
773 0 : ctx->slot_duration_ns = (double)ticks_per_slot*(double)tick_duration_ns;
774 0 : ctx->hashcnt_duration_ns = (double)tick_duration_ns/(double)hashcnt_per_tick;
775 0 : ctx->hashcnt_per_slot = ticks_per_slot*hashcnt_per_tick;
776 :
777 0 : if( FD_UNLIKELY( ctx->hashcnt_per_tick==1UL ) ) {
778 : /* Low power producer, maximum of one microblock per tick in the slot */
779 0 : ctx->max_microblocks_per_slot = ctx->ticks_per_slot;
780 0 : } else {
781 : /* See the long comment in after_credit for this limit */
782 0 : ctx->max_microblocks_per_slot = fd_ulong_min( MAX_MICROBLOCKS_PER_SLOT, ctx->ticks_per_slot*(ctx->hashcnt_per_tick-1UL) );
783 0 : }
784 :
785 0 : fd_ext_poh_write_unlock();
786 0 : }
787 :
788 : /* fd_ext_poh_acquire_bank gets the current leader bank if there is one
789 : currently active. PoH might think we are leader without having a
790 : leader bank if the replay stage has not yet noticed we are leader.
791 :
792 : The bank that is returned is owned the caller, and must be converted
793 : to an Arc<Bank> by calling Arc::from_raw() on it. PoH increments the
794 : reference count before returning the bank, so that it can also keep
795 : its internal copy.
796 :
797 : If there is no leader bank, NULL is returned. In this case, the
798 : caller should not call `Arc::from_raw()`. */
799 :
800 : CALLED_FROM_RUST void const *
801 0 : fd_ext_poh_acquire_leader_bank( void ) {
802 0 : fd_poh_ctx_t * ctx = fd_ext_poh_write_lock();
803 0 : void const * bank = NULL;
804 0 : if( FD_LIKELY( ctx->current_leader_bank ) ) {
805 : /* Clone refcount before we release the lock. */
806 0 : fd_ext_bank_acquire( ctx->current_leader_bank );
807 0 : bank = ctx->current_leader_bank;
808 0 : }
809 0 : fd_ext_poh_write_unlock();
810 0 : return bank;
811 0 : }
812 :
813 : /* fd_ext_poh_reset_slot returns the slot height one above the last good
814 : (unskipped) slot we are building on top of. This is always a good
815 : known value, and will not be ULONG_MAX. */
816 :
817 : CALLED_FROM_RUST ulong
818 0 : fd_ext_poh_reset_slot( void ) {
819 0 : fd_poh_ctx_t * ctx = fd_ext_poh_write_lock();
820 0 : ulong reset_slot = ctx->reset_slot;
821 0 : fd_ext_poh_write_unlock();
822 0 : return reset_slot;
823 0 : }
824 :
825 : CALLED_FROM_RUST void
826 0 : fd_ext_poh_update_active_descendant( ulong max_active_descendant ) {
827 0 : fd_poh_ctx_t * ctx = fd_ext_poh_write_lock();
828 0 : ctx->max_active_descendant = max_active_descendant;
829 0 : fd_ext_poh_write_unlock();
830 0 : }
831 :
832 : /* fd_ext_poh_reached_leader_slot returns 1 if we have reached a slot
833 : where we are leader. This is used by the replay stage to determine
834 : if it should create a new leader bank descendant of the prior reset
835 : slot block.
836 :
837 : Sometimes, even when we reach our slot we do not return 1, as we are
838 : giving a grace period to the prior leader to finish publishing their
839 : block.
840 :
841 : out_leader_slot is the slot height of the leader slot we reached, and
842 : reset_slot is the slot height of the last good (unskipped) slot we
843 : are building on top of. */
844 :
845 : CALLED_FROM_RUST int
846 : fd_ext_poh_reached_leader_slot( ulong * out_leader_slot,
847 0 : ulong * out_reset_slot ) {
848 0 : fd_poh_ctx_t * ctx = fd_ext_poh_write_lock();
849 :
850 0 : *out_leader_slot = ctx->next_leader_slot;
851 0 : *out_reset_slot = ctx->reset_slot;
852 :
853 0 : if( FD_UNLIKELY( ctx->next_leader_slot==ULONG_MAX ||
854 0 : ctx->slot<ctx->next_leader_slot ) ) {
855 : /* Didn't reach our leader slot yet. */
856 0 : fd_ext_poh_write_unlock();
857 0 : return 0;
858 0 : }
859 :
860 0 : if( FD_UNLIKELY( ctx->halted_switching_key ) ) {
861 : /* Reached our leader slot, but the leader pipeline is halted
862 : because we are switching identity key. */
863 0 : fd_ext_poh_write_unlock();
864 0 : return 0;
865 0 : }
866 :
867 0 : if( FD_LIKELY( ctx->reset_slot==ctx->next_leader_slot ) ) {
868 : /* We were reset onto our leader slot, because the prior leader
869 : completed theirs, so we should start immediately, no need for a
870 : grace period. */
871 0 : fd_ext_poh_write_unlock();
872 0 : return 1;
873 0 : }
874 :
875 0 : long now_ns = fd_log_wallclock();
876 0 : long expected_start_time_ns = ctx->reset_slot_start_ns + (long)((double)(ctx->next_leader_slot-ctx->reset_slot)*ctx->slot_duration_ns);
877 :
878 : /* If a prior leader is still in the process of publishing their slot,
879 : delay ours to let them finish ... unless they are so delayed that
880 : we risk getting skipped by the leader following us. 1.2 seconds
881 : is a reasonable default here, although any value between 0 and 1.6
882 : seconds could be considered reasonable. This is arbitrary and
883 : chosen due to intuition. */
884 :
885 0 : if( FD_UNLIKELY( now_ns<expected_start_time_ns+(long)(3.0*ctx->slot_duration_ns) ) ) {
886 : /* If the max_active_descendant is >= next_leader_slot, we waited
887 : too long and a leader after us started publishing to try and skip
888 : us. Just start our leader slot immediately, we might win ... */
889 :
890 0 : if( FD_LIKELY( ctx->max_active_descendant>=ctx->reset_slot && ctx->max_active_descendant<ctx->next_leader_slot ) ) {
891 : /* If one of the leaders between the reset slot and our leader
892 : slot is in the process of publishing (they have a descendant
893 : bank that is in progress of being replayed), then keep waiting.
894 : We probably wouldn't get a leader slot out before they
895 : finished.
896 :
897 : Unless... we are past the deadline to start our slot by more
898 : than 1.2 seconds, in which case we should probably start it to
899 : avoid getting skipped by the leader behind us. */
900 0 : fd_ext_poh_write_unlock();
901 0 : return 0;
902 0 : }
903 0 : }
904 :
905 0 : fd_ext_poh_write_unlock();
906 0 : return 1;
907 0 : }
908 :
909 : CALLED_FROM_RUST static inline void
910 : publish_plugin_slot_start( fd_poh_ctx_t * ctx,
911 : ulong slot,
912 0 : ulong parent_slot ) {
913 0 : if( FD_UNLIKELY( !ctx->plugin_out->mem ) ) return;
914 :
915 0 : fd_plugin_msg_slot_start_t * slot_start = (fd_plugin_msg_slot_start_t *)fd_chunk_to_laddr( ctx->plugin_out->mem, ctx->plugin_out->chunk );
916 0 : *slot_start = (fd_plugin_msg_slot_start_t){ .slot = slot, .parent_slot = parent_slot };
917 0 : fd_stem_publish( ctx->stem, ctx->plugin_out->idx, FD_PLUGIN_MSG_SLOT_START, ctx->plugin_out->chunk, sizeof(fd_plugin_msg_slot_start_t), 0UL, 0UL, 0UL );
918 0 : ctx->plugin_out->chunk = fd_dcache_compact_next( ctx->plugin_out->chunk, sizeof(fd_plugin_msg_slot_start_t), ctx->plugin_out->chunk0, ctx->plugin_out->wmark );
919 0 : }
920 :
921 : CALLED_FROM_RUST static inline void
922 : publish_plugin_slot_end( fd_poh_ctx_t * ctx,
923 : ulong slot,
924 0 : ulong cus_used ) {
925 0 : if( FD_UNLIKELY( !ctx->plugin_out->mem ) ) return;
926 :
927 0 : fd_plugin_msg_slot_end_t * slot_end = (fd_plugin_msg_slot_end_t *)fd_chunk_to_laddr( ctx->plugin_out->mem, ctx->plugin_out->chunk );
928 0 : *slot_end = (fd_plugin_msg_slot_end_t){ .slot = slot, .cus_used = cus_used };
929 0 : fd_stem_publish( ctx->stem, ctx->plugin_out->idx, FD_PLUGIN_MSG_SLOT_END, ctx->plugin_out->chunk, sizeof(fd_plugin_msg_slot_end_t), 0UL, 0UL, 0UL );
930 0 : ctx->plugin_out->chunk = fd_dcache_compact_next( ctx->plugin_out->chunk, sizeof(fd_plugin_msg_slot_end_t), ctx->plugin_out->chunk0, ctx->plugin_out->wmark );
931 0 : }
932 :
933 : extern int
934 : fd_ext_bank_load_account( void const * bank,
935 : int fixed_root,
936 : uchar const * addr,
937 : uchar * owner,
938 : uchar * data,
939 : ulong * data_sz );
940 :
941 : CALLED_FROM_RUST static void
942 : publish_became_leader( fd_poh_ctx_t * ctx,
943 : ulong slot,
944 0 : ulong epoch ) {
945 0 : double tick_per_ns = fd_tempo_tick_per_ns( NULL );
946 0 : fd_histf_sample( ctx->begin_leader_delay, (ulong)((double)(fd_log_wallclock()-ctx->reset_slot_start_ns)/tick_per_ns) );
947 :
948 0 : if( FD_UNLIKELY( ctx->lagged_consecutive_leader_start ) ) {
949 : /* If we are mirroring Agave behavior, the wall clock gets reset
950 : here so we don't count time spent waiting for a bank to freeze
951 : or replay stage to actually start the slot towards our 400ms.
952 :
953 : See extended comments in the config file on this option. */
954 0 : ctx->reset_slot_start_ns = fd_log_wallclock() - (long)((double)(slot-ctx->reset_slot)*ctx->slot_duration_ns);
955 0 : }
956 :
957 0 : fd_bundle_crank_tip_payment_config_t config[1] = { 0 };
958 0 : fd_acct_addr_t tip_receiver_owner[1] = { 0 };
959 :
960 0 : if( FD_UNLIKELY( ctx->bundle.enabled ) ) {
961 0 : long bundle_time = -fd_tickcount();
962 0 : fd_acct_addr_t tip_payment_config[1];
963 0 : fd_acct_addr_t tip_receiver[1];
964 0 : fd_bundle_crank_get_addresses( ctx->bundle.gen, epoch, tip_payment_config, tip_receiver );
965 :
966 0 : fd_acct_addr_t _dummy[1];
967 0 : uchar dummy[1];
968 :
969 0 : void const * bank = ctx->current_leader_bank;
970 :
971 : /* Calling rust from a C function that is CALLED_FROM_RUST risks
972 : deadlock. In this case, I checked the load_account function and
973 : ensured it never calls any C functions that acquire the lock. */
974 0 : ulong sz1 = sizeof(config), sz2 = 1UL;
975 0 : int found1 = fd_ext_bank_load_account( bank, 0, tip_payment_config->b, _dummy->b, (uchar *)config, &sz1 );
976 0 : int found2 = fd_ext_bank_load_account( bank, 0, tip_receiver->b, tip_receiver_owner->b, dummy, &sz2 );
977 : /* The bundle crank code detects whether the accounts were found by
978 : whether they have non-zero values (since found and uninitialized
979 : should be treated the same), so we actually don't really care
980 : about the value of found{1,2}. */
981 0 : (void)found1; (void)found2;
982 0 : bundle_time += fd_tickcount();
983 0 : fd_histf_sample( ctx->bundle_init_delay, (ulong)bundle_time );
984 0 : }
985 :
986 0 : long slot_start_ns = ctx->reset_slot_start_ns + (long)((double)(slot-ctx->reset_slot)*ctx->slot_duration_ns);
987 :
988 : /* No need to check flow control, there are always credits became when we
989 : are leader, we will not "become" leader again until we are done, so at
990 : most one frag in flight at a time. */
991 :
992 0 : uchar * dst = (uchar *)fd_chunk_to_laddr( ctx->pack_out->mem, ctx->pack_out->chunk );
993 :
994 0 : fd_became_leader_t * leader = (fd_became_leader_t *)dst;
995 0 : leader->slot_start_ns = slot_start_ns;
996 0 : leader->slot_end_ns = (long)((double)slot_start_ns + ctx->slot_duration_ns);
997 0 : leader->bank = ctx->current_leader_bank;
998 0 : leader->max_microblocks_in_slot = ctx->max_microblocks_per_slot;
999 0 : leader->ticks_per_slot = ctx->ticks_per_slot;
1000 0 : leader->total_skipped_ticks = ctx->ticks_per_slot*(slot-ctx->reset_slot);
1001 0 : leader->epoch = epoch;
1002 0 : leader->bundle->config[0] = config[0];
1003 :
1004 0 : memcpy( leader->bundle->last_blockhash, ctx->reset_hash, 32UL );
1005 0 : memcpy( leader->bundle->tip_receiver_owner, tip_receiver_owner, 32UL );
1006 :
1007 0 : if( FD_UNLIKELY( leader->ticks_per_slot+leader->total_skipped_ticks>=MAX_SKIPPED_TICKS ) )
1008 0 : FD_LOG_ERR(( "Too many skipped ticks %lu for slot %lu, chain must halt", leader->ticks_per_slot+leader->total_skipped_ticks, slot ));
1009 :
1010 0 : ulong sig = fd_disco_poh_sig( slot, POH_PKT_TYPE_BECAME_LEADER, 0UL );
1011 0 : fd_stem_publish( ctx->stem, ctx->pack_out->idx, sig, ctx->pack_out->chunk, sizeof(fd_became_leader_t), 0UL, 0UL, 0UL );
1012 0 : ctx->pack_out->chunk = fd_dcache_compact_next( ctx->pack_out->chunk, sizeof(fd_became_leader_t), ctx->pack_out->chunk0, ctx->pack_out->wmark );
1013 0 : }
1014 :
1015 : /* The PoH tile knows when it should become leader by waiting for its
1016 : leader slot (with the operating system clock). This function is so
1017 : that when it becomes the leader, it can be told what the leader bank
1018 : is by the replay stage. See the notes in the long comment above for
1019 : more on how this works. */
1020 :
1021 : CALLED_FROM_RUST void
1022 : fd_ext_poh_begin_leader( void const * bank,
1023 : ulong slot,
1024 : ulong epoch,
1025 0 : ulong hashcnt_per_tick ) {
1026 0 : fd_poh_ctx_t * ctx = fd_ext_poh_write_lock();
1027 :
1028 0 : FD_TEST( !ctx->current_leader_bank );
1029 :
1030 0 : if( FD_UNLIKELY( slot!=ctx->slot ) ) FD_LOG_ERR(( "Trying to begin leader slot %lu but we are now on slot %lu", slot, ctx->slot ));
1031 0 : if( FD_UNLIKELY( slot!=ctx->next_leader_slot ) ) FD_LOG_ERR(( "Trying to begin leader slot %lu but next leader slot is %lu", slot, ctx->next_leader_slot ));
1032 :
1033 0 : if( FD_UNLIKELY( ctx->hashcnt_per_tick!=hashcnt_per_tick ) ) {
1034 0 : FD_LOG_WARNING(( "hashes per tick changed from %lu to %lu", ctx->hashcnt_per_tick, hashcnt_per_tick ));
1035 :
1036 : /* Recompute derived information about the clock. */
1037 0 : ctx->hashcnt_duration_ns = (double)ctx->tick_duration_ns/(double)hashcnt_per_tick;
1038 0 : ctx->hashcnt_per_slot = ctx->ticks_per_slot*hashcnt_per_tick;
1039 0 : ctx->hashcnt_per_tick = hashcnt_per_tick;
1040 :
1041 0 : if( FD_UNLIKELY( ctx->hashcnt_per_tick==1UL ) ) {
1042 : /* Low power producer, maximum of one microblock per tick in the slot */
1043 0 : ctx->max_microblocks_per_slot = ctx->ticks_per_slot;
1044 0 : } else {
1045 : /* See the long comment in after_credit for this limit */
1046 0 : ctx->max_microblocks_per_slot = fd_ulong_min( MAX_MICROBLOCKS_PER_SLOT, ctx->ticks_per_slot*(ctx->hashcnt_per_tick-1UL) );
1047 0 : }
1048 :
1049 : /* Discard any ticks we might have done in the interim. They will
1050 : have the wrong number of hashes per tick. We can just catch back
1051 : up quickly if not too many slots were skipped and hopefully
1052 : publish on time. Note that tick production and verification of
1053 : skipped slots is done for the eventual bank that publishes a
1054 : slot, for example:
1055 :
1056 : Reset Slot: 998
1057 : Epoch Transition Slot: 1000
1058 : Leader Slot: 1002
1059 :
1060 : In this case, if a feature changing the hashcnt_per_tick is
1061 : activated in slot 1000, and we are publishing empty ticks for
1062 : slots 998, 999, 1000, and 1001, they should all have the new
1063 : hashes_per_tick number of hashes, rather than the older one, or
1064 : some combination. */
1065 :
1066 0 : FD_TEST( ctx->last_slot==ctx->reset_slot );
1067 0 : FD_TEST( !ctx->last_hashcnt );
1068 0 : ctx->slot = ctx->reset_slot;
1069 0 : ctx->hashcnt = 0UL;
1070 0 : }
1071 :
1072 0 : ctx->current_leader_bank = bank;
1073 0 : ctx->microblocks_lower_bound = 0UL;
1074 0 : ctx->cus_used = 0UL;
1075 0 : ctx->expect_microblock_idx = 0UL;
1076 :
1077 : /* We are about to start publishing to the shred tile for this slot
1078 : so update the highwater mark so we never republish in this slot
1079 : again. Also check that the leader slot is greater than the
1080 : highwater, which should have been ensured earlier. */
1081 :
1082 0 : FD_TEST( ctx->highwater_leader_slot==ULONG_MAX || slot>=ctx->highwater_leader_slot );
1083 0 : ctx->highwater_leader_slot = fd_ulong_max( fd_ulong_if( ctx->highwater_leader_slot==ULONG_MAX, 0UL, ctx->highwater_leader_slot ), slot );
1084 :
1085 0 : publish_became_leader( ctx, slot, epoch );
1086 0 : FD_LOG_INFO(( "fd_ext_poh_begin_leader(slot=%lu, highwater_leader_slot=%lu, last_slot=%lu, last_hashcnt=%lu)", slot, ctx->highwater_leader_slot, ctx->last_slot, ctx->last_hashcnt ));
1087 :
1088 0 : fd_ext_poh_write_unlock();
1089 0 : }
1090 :
1091 : /* Determine what the next slot is in the leader schedule is that we are
1092 : leader. Includes the current slot. If we are not leader in what
1093 : remains of the current and next epoch, return ULONG_MAX. */
1094 :
1095 : static inline CALLED_FROM_RUST ulong
1096 0 : next_leader_slot( fd_poh_ctx_t * ctx ) {
1097 : /* If we have published anything in a particular slot, then we
1098 : should never become leader for that slot again. */
1099 0 : ulong min_leader_slot = fd_ulong_max( ctx->slot, fd_ulong_if( ctx->highwater_leader_slot==ULONG_MAX, 0UL, ctx->highwater_leader_slot ) );
1100 :
1101 0 : for(;;) {
1102 0 : fd_epoch_leaders_t * leaders = fd_stake_ci_get_lsched_for_slot( ctx->stake_ci, min_leader_slot ); /* Safe to call from Rust */
1103 0 : if( FD_UNLIKELY( !leaders ) ) break;
1104 :
1105 0 : while( min_leader_slot<(leaders->slot0+leaders->slot_cnt) ) {
1106 0 : fd_pubkey_t const * leader = fd_epoch_leaders_get( leaders, min_leader_slot ); /* Safe to call from Rust */
1107 0 : if( FD_UNLIKELY( !memcmp( leader->key, ctx->identity_key.key, 32UL ) ) ) return min_leader_slot;
1108 0 : min_leader_slot++;
1109 0 : }
1110 0 : }
1111 :
1112 0 : return ULONG_MAX;
1113 0 : }
1114 :
1115 : extern int
1116 : fd_ext_admin_rpc_set_identity( uchar const * identity_keypair,
1117 : int require_tower );
1118 :
1119 : static inline int FD_FN_SENSITIVE
1120 : maybe_change_identity( fd_poh_ctx_t * ctx,
1121 0 : int definitely_not_leader ) {
1122 0 : if( FD_UNLIKELY( ctx->halted_switching_key && fd_keyswitch_state_query( ctx->keyswitch )==FD_KEYSWITCH_STATE_UNHALT_PENDING ) ) {
1123 0 : ctx->halted_switching_key = 0;
1124 0 : fd_keyswitch_state( ctx->keyswitch, FD_KEYSWITCH_STATE_COMPLETED );
1125 0 : return 1;
1126 0 : }
1127 :
1128 : /* Cannot change identity while in the middle of a leader slot, else
1129 : poh state machine would become corrupt. */
1130 :
1131 0 : int is_leader = !definitely_not_leader && ctx->next_leader_slot!=ULONG_MAX && ctx->slot>=ctx->next_leader_slot;
1132 0 : if( FD_UNLIKELY( is_leader ) ) return 0;
1133 :
1134 0 : if( FD_UNLIKELY( fd_keyswitch_state_query( ctx->keyswitch )==FD_KEYSWITCH_STATE_SWITCH_PENDING ) ) {
1135 0 : int failed = fd_ext_admin_rpc_set_identity( ctx->keyswitch->bytes, fd_keyswitch_param_query( ctx->keyswitch )==1 );
1136 0 : explicit_bzero( ctx->keyswitch->bytes, 32UL );
1137 0 : FD_COMPILER_MFENCE();
1138 0 : if( FD_UNLIKELY( failed==-1 ) ) {
1139 0 : fd_keyswitch_state( ctx->keyswitch, FD_KEYSWITCH_STATE_FAILED );
1140 0 : return 0;
1141 0 : }
1142 :
1143 0 : memcpy( ctx->identity_key.uc, ctx->keyswitch->bytes+32UL, 32UL );
1144 0 : fd_stake_ci_set_identity( ctx->stake_ci, &ctx->identity_key );
1145 :
1146 : /* When we switch key, we might have ticked part way through a slot
1147 : that we are now leader in. This violates the contract of the
1148 : tile, that when we become leader, we have not ticked in that slot
1149 : at all. To see why this would be bad, consider the case where we
1150 : have ticked almost to the end, and there isn't enough space left
1151 : to reserve the minimum amount of microblocks needed by pack.
1152 :
1153 : To resolve this, we just reset PoH back to the reset slot, and
1154 : let it try to catch back up quickly. This is OK since the network
1155 : rarely skips. */
1156 0 : ctx->slot = ctx->reset_slot;
1157 0 : ctx->hashcnt = 0UL;
1158 0 : memcpy( ctx->hash, ctx->reset_hash, 32UL );
1159 :
1160 0 : ctx->halted_switching_key = 1;
1161 0 : ctx->keyswitch->result = ctx->shred_seq;
1162 0 : fd_keyswitch_state( ctx->keyswitch, FD_KEYSWITCH_STATE_COMPLETED );
1163 0 : }
1164 :
1165 0 : return 0;
1166 0 : }
1167 :
1168 : static CALLED_FROM_RUST void
1169 0 : no_longer_leader( fd_poh_ctx_t * ctx ) {
1170 0 : if( FD_UNLIKELY( ctx->current_leader_bank ) ) fd_ext_bank_release( ctx->current_leader_bank );
1171 : /* If we stop being leader in a slot, we can never become leader in
1172 : that slot again, and all in-flight microblocks for that slot
1173 : should be dropped. */
1174 0 : ctx->highwater_leader_slot = fd_ulong_max( fd_ulong_if( ctx->highwater_leader_slot==ULONG_MAX, 0UL, ctx->highwater_leader_slot ), ctx->slot );
1175 0 : ctx->current_leader_bank = NULL;
1176 0 : int identity_changed = maybe_change_identity( ctx, 1 );
1177 0 : ctx->next_leader_slot = next_leader_slot( ctx );
1178 0 : if( FD_UNLIKELY( identity_changed ) ) {
1179 0 : FD_LOG_INFO(( "fd_poh_identity_changed(next_leader_slot=%lu)", ctx->next_leader_slot ));
1180 0 : }
1181 :
1182 0 : FD_COMPILER_MFENCE();
1183 0 : fd_ext_poh_signal_leader_change( ctx->signal_leader_change );
1184 0 : FD_LOG_INFO(( "no_longer_leader(next_leader_slot=%lu)", ctx->next_leader_slot ));
1185 0 : }
1186 :
1187 : /* fd_ext_poh_reset is called by the Agave client when a slot on
1188 : the active fork has finished a block and we need to reset our PoH to
1189 : be ticking on top of the block it produced. */
1190 :
1191 : CALLED_FROM_RUST void
1192 : fd_ext_poh_reset( ulong completed_bank_slot, /* The slot that successfully produced a block */
1193 : uchar const * reset_blockhash, /* The hash of the last tick in the produced block */
1194 0 : ulong hashcnt_per_tick /* The hashcnt per tick of the bank that completed */ ) {
1195 0 : fd_poh_ctx_t * ctx = fd_ext_poh_write_lock();
1196 :
1197 0 : ulong slot_before_reset = ctx->slot;
1198 0 : int leader_before_reset = ctx->slot>=ctx->next_leader_slot;
1199 0 : if( FD_UNLIKELY( leader_before_reset && ctx->current_leader_bank ) ) {
1200 : /* If we were in the middle of a leader slot that we notified pack
1201 : pack to start packing for we can never publish into that slot
1202 : again, mark all in-flight microblocks to be dropped. */
1203 0 : ctx->highwater_leader_slot = fd_ulong_max( fd_ulong_if( ctx->highwater_leader_slot==ULONG_MAX, 0UL, ctx->highwater_leader_slot ), 1UL+ctx->slot );
1204 0 : }
1205 :
1206 0 : ctx->leader_bank_start_ns = fd_log_wallclock(); /* safe to call from Rust */
1207 0 : if( FD_UNLIKELY( ctx->expect_sequential_leader_slot==(completed_bank_slot+1UL) ) ) {
1208 : /* If we are being reset onto a slot, it means some block was fully
1209 : processed, so we reset to build on top of it. Typically we want
1210 : to update the reset_slot_start_ns to the current time, because
1211 : the network will give the next leader 400ms to publish,
1212 : regardless of how long the prior leader took.
1213 :
1214 : But: if we were leader in the prior slot, and the block was our
1215 : own we can do better. We know that the next slot should start
1216 : exactly 400ms after the prior one started, so we can use that as
1217 : the reset slot start time instead. */
1218 0 : ctx->reset_slot_start_ns = ctx->reset_slot_start_ns + (long)((double)((completed_bank_slot+1UL)-ctx->reset_slot)*ctx->slot_duration_ns);
1219 0 : } else {
1220 0 : ctx->reset_slot_start_ns = ctx->leader_bank_start_ns;
1221 0 : }
1222 0 : ctx->expect_sequential_leader_slot = ULONG_MAX;
1223 :
1224 0 : memcpy( ctx->reset_hash, reset_blockhash, 32UL );
1225 0 : memcpy( ctx->hash, reset_blockhash, 32UL );
1226 0 : ctx->slot = completed_bank_slot+1UL;
1227 0 : ctx->hashcnt = 0UL;
1228 0 : ctx->last_slot = ctx->slot;
1229 0 : ctx->last_hashcnt = 0UL;
1230 0 : ctx->reset_slot = ctx->slot;
1231 :
1232 0 : if( FD_UNLIKELY( ctx->hashcnt_per_tick!=hashcnt_per_tick ) ) {
1233 0 : FD_LOG_WARNING(( "hashes per tick changed from %lu to %lu", ctx->hashcnt_per_tick, hashcnt_per_tick ));
1234 :
1235 : /* Recompute derived information about the clock. */
1236 0 : ctx->hashcnt_duration_ns = (double)ctx->tick_duration_ns/(double)hashcnt_per_tick;
1237 0 : ctx->hashcnt_per_slot = ctx->ticks_per_slot*hashcnt_per_tick;
1238 0 : ctx->hashcnt_per_tick = hashcnt_per_tick;
1239 :
1240 0 : if( FD_UNLIKELY( ctx->hashcnt_per_tick==1UL ) ) {
1241 : /* Low power producer, maximum of one microblock per tick in the slot */
1242 0 : ctx->max_microblocks_per_slot = ctx->ticks_per_slot;
1243 0 : } else {
1244 : /* See the long comment in after_credit for this limit */
1245 0 : ctx->max_microblocks_per_slot = fd_ulong_min( MAX_MICROBLOCKS_PER_SLOT, ctx->ticks_per_slot*(ctx->hashcnt_per_tick-1UL) );
1246 0 : }
1247 0 : }
1248 :
1249 : /* When we reset, we need to allow PoH to tick freely again rather
1250 : than being constrained. If we are leader after the reset, this
1251 : is OK because we won't tick until we get a bank, and the lower
1252 : bound will be reset with the value from the bank. */
1253 0 : ctx->microblocks_lower_bound = ctx->max_microblocks_per_slot;
1254 :
1255 0 : if( FD_UNLIKELY( leader_before_reset ) ) {
1256 : /* No longer have a leader bank if we are reset. Replay stage will
1257 : call back again to give us a new one if we should become leader
1258 : for the reset slot.
1259 :
1260 : The order is important here, ctx->hashcnt must be updated before
1261 : calling no_longer_leader. */
1262 0 : no_longer_leader( ctx );
1263 0 : }
1264 0 : ctx->next_leader_slot = next_leader_slot( ctx );
1265 0 : FD_LOG_INFO(( "fd_ext_poh_reset(slot=%lu,next_leader_slot=%lu)", ctx->reset_slot, ctx->next_leader_slot ));
1266 :
1267 0 : if( FD_UNLIKELY( ctx->slot>=ctx->next_leader_slot ) ) {
1268 : /* We are leader after the reset... two cases: */
1269 0 : if( FD_LIKELY( ctx->slot==slot_before_reset ) ) {
1270 : /* 1. We are reset onto the same slot we are already leader on.
1271 : This is a common case when we have two leader slots in a
1272 : row, replay stage will reset us to our own slot. No need to
1273 : do anything here, we already sent a SLOT_START. */
1274 0 : FD_TEST( leader_before_reset );
1275 0 : } else {
1276 : /* 2. We are reset onto a different slot. If we were leader
1277 : before, we should first end that slot, then begin the new
1278 : one if we are newly leader now. */
1279 0 : if( FD_LIKELY( leader_before_reset ) ) publish_plugin_slot_end( ctx, slot_before_reset, ctx->cus_used );
1280 0 : else publish_plugin_slot_start( ctx, ctx->next_leader_slot, ctx->reset_slot );
1281 0 : }
1282 0 : } else {
1283 0 : if( FD_UNLIKELY( leader_before_reset ) ) publish_plugin_slot_end( ctx, slot_before_reset, ctx->cus_used );
1284 0 : }
1285 :
1286 0 : fd_ext_poh_write_unlock();
1287 0 : }
1288 :
1289 : /* Since it can't easily return an Option<Pubkey>, return 1 for Some and
1290 : 0 for None. */
1291 : CALLED_FROM_RUST int
1292 : fd_ext_poh_get_leader_after_n_slots( ulong n,
1293 0 : uchar out_pubkey[ static 32 ] ) {
1294 0 : fd_poh_ctx_t * ctx = fd_ext_poh_write_lock();
1295 0 : ulong slot = ctx->slot + n;
1296 0 : fd_epoch_leaders_t * leaders = fd_stake_ci_get_lsched_for_slot( ctx->stake_ci, slot ); /* Safe to call from Rust */
1297 :
1298 0 : int copied = 0;
1299 0 : if( FD_LIKELY( leaders ) ) {
1300 0 : fd_pubkey_t const * leader = fd_epoch_leaders_get( leaders, slot ); /* Safe to call from Rust */
1301 0 : if( FD_LIKELY( leader ) ) {
1302 0 : memcpy( out_pubkey, leader, 32UL );
1303 0 : copied = 1;
1304 0 : }
1305 0 : }
1306 0 : fd_ext_poh_write_unlock();
1307 0 : return copied;
1308 0 : }
1309 :
1310 : FD_FN_CONST static inline ulong
1311 3 : scratch_align( void ) {
1312 3 : return 128UL;
1313 3 : }
1314 :
1315 : FD_FN_PURE static inline ulong
1316 3 : scratch_footprint( fd_topo_tile_t const * tile ) {
1317 3 : (void)tile;
1318 3 : ulong l = FD_LAYOUT_INIT;
1319 3 : l = FD_LAYOUT_APPEND( l, alignof( fd_poh_ctx_t ), sizeof( fd_poh_ctx_t ) );
1320 3 : l = FD_LAYOUT_APPEND( l, fd_stake_ci_align(), fd_stake_ci_footprint() );
1321 3 : l = FD_LAYOUT_APPEND( l, FD_SHA256_ALIGN, FD_SHA256_FOOTPRINT );
1322 3 : return FD_LAYOUT_FINI( l, scratch_align() );
1323 3 : }
1324 :
1325 : static void
1326 : publish_tick( fd_poh_ctx_t * ctx,
1327 : fd_stem_context_t * stem,
1328 : uchar hash[ static 32 ],
1329 0 : int is_skipped ) {
1330 0 : ulong hashcnt = ctx->hashcnt_per_tick*(1UL+(ctx->last_hashcnt/ctx->hashcnt_per_tick));
1331 :
1332 0 : uchar * dst = (uchar *)fd_chunk_to_laddr( ctx->shred_out->mem, ctx->shred_out->chunk );
1333 :
1334 0 : FD_TEST( ctx->last_slot>=ctx->reset_slot );
1335 0 : fd_entry_batch_meta_t * meta = (fd_entry_batch_meta_t *)dst;
1336 0 : if( FD_UNLIKELY( is_skipped ) ) {
1337 : /* We are publishing ticks for a skipped slot, the reference tick
1338 : and block complete flags should always be zero. */
1339 0 : meta->reference_tick = 0UL;
1340 0 : meta->block_complete = 0;
1341 0 : } else {
1342 0 : meta->reference_tick = hashcnt/ctx->hashcnt_per_tick;
1343 0 : meta->block_complete = hashcnt==ctx->hashcnt_per_slot;
1344 0 : }
1345 :
1346 0 : ulong slot = fd_ulong_if( meta->block_complete, ctx->slot-1UL, ctx->slot );
1347 0 : meta->parent_offset = 1UL+slot-ctx->reset_slot;
1348 :
1349 0 : FD_TEST( hashcnt>ctx->last_hashcnt );
1350 0 : ulong hash_delta = hashcnt-ctx->last_hashcnt;
1351 :
1352 0 : dst += sizeof(fd_entry_batch_meta_t);
1353 0 : fd_entry_batch_header_t * tick = (fd_entry_batch_header_t *)dst;
1354 0 : tick->hashcnt_delta = hash_delta;
1355 0 : fd_memcpy( tick->hash, hash, 32UL );
1356 0 : tick->txn_cnt = 0UL;
1357 :
1358 0 : ulong tspub = (ulong)fd_frag_meta_ts_comp( fd_tickcount() );
1359 0 : ulong sz = sizeof(fd_entry_batch_meta_t)+sizeof(fd_entry_batch_header_t);
1360 0 : ulong sig = fd_disco_poh_sig( slot, POH_PKT_TYPE_MICROBLOCK, 0UL );
1361 0 : fd_stem_publish( stem, ctx->shred_out->idx, sig, ctx->shred_out->chunk, sz, 0UL, 0UL, tspub );
1362 0 : ctx->shred_seq = stem->seqs[ ctx->shred_out->idx ];
1363 0 : ctx->shred_out->chunk = fd_dcache_compact_next( ctx->shred_out->chunk, sz, ctx->shred_out->chunk0, ctx->shred_out->wmark );
1364 :
1365 0 : if( FD_UNLIKELY( hashcnt==ctx->hashcnt_per_slot ) ) {
1366 0 : ctx->last_slot++;
1367 0 : ctx->last_hashcnt = 0UL;
1368 0 : } else {
1369 0 : ctx->last_hashcnt = hashcnt;
1370 0 : }
1371 0 : }
1372 :
1373 : static inline void
1374 : after_credit( fd_poh_ctx_t * ctx,
1375 : fd_stem_context_t * stem,
1376 : int * opt_poll_in,
1377 0 : int * charge_busy ) {
1378 0 : ctx->stem = stem;
1379 :
1380 0 : FD_COMPILER_MFENCE();
1381 0 : if( FD_UNLIKELY( fd_poh_waiting_lock ) ) {
1382 0 : FD_VOLATILE( fd_poh_returned_lock ) = 1UL;
1383 0 : FD_COMPILER_MFENCE();
1384 0 : for(;;) {
1385 0 : if( FD_UNLIKELY( !FD_VOLATILE_CONST( fd_poh_returned_lock ) ) ) break;
1386 0 : FD_SPIN_PAUSE();
1387 0 : }
1388 0 : FD_COMPILER_MFENCE();
1389 0 : FD_VOLATILE( fd_poh_waiting_lock ) = 0UL;
1390 0 : *opt_poll_in = 0;
1391 0 : *charge_busy = 1;
1392 0 : return;
1393 0 : }
1394 0 : FD_COMPILER_MFENCE();
1395 :
1396 0 : int is_leader = ctx->next_leader_slot!=ULONG_MAX && ctx->slot>=ctx->next_leader_slot;
1397 0 : if( FD_UNLIKELY( is_leader && !ctx->current_leader_bank ) ) {
1398 : /* If we are the leader, but we didn't yet learn what the leader
1399 : bank object is from the replay stage, do not do any hashing.
1400 :
1401 : This is not ideal, but greatly simplifies the control flow. */
1402 0 : return;
1403 0 : }
1404 :
1405 : /* If we have skipped ticks pending because we skipped some slots to
1406 : become leader, register them now one at a time. */
1407 0 : if( FD_UNLIKELY( is_leader && ctx->last_slot<ctx->slot ) ) {
1408 0 : ulong publish_hashcnt = ctx->last_hashcnt+ctx->hashcnt_per_tick;
1409 0 : ulong tick_idx = (ctx->last_slot*ctx->ticks_per_slot+publish_hashcnt/ctx->hashcnt_per_tick)%MAX_SKIPPED_TICKS;
1410 :
1411 0 : fd_ext_poh_register_tick( ctx->current_leader_bank, ctx->skipped_tick_hashes[ tick_idx ] );
1412 0 : publish_tick( ctx, stem, ctx->skipped_tick_hashes[ tick_idx ], 1 );
1413 :
1414 : /* If we are catching up now and publishing a bunch of skipped
1415 : ticks, we do not want to process any incoming microblocks until
1416 : all the skipped ticks have been published out; otherwise we would
1417 : intersperse skipped tick messages with microblocks. */
1418 0 : *opt_poll_in = 0;
1419 0 : *charge_busy = 1;
1420 0 : return;
1421 0 : }
1422 :
1423 0 : int low_power_mode = ctx->hashcnt_per_tick==1UL;
1424 :
1425 : /* If we are the leader, always leave enough capacity in the slot so
1426 : that we can mixin any potential microblocks still coming from the
1427 : pack tile for this slot. */
1428 0 : ulong max_remaining_microblocks = ctx->max_microblocks_per_slot - ctx->microblocks_lower_bound;
1429 : /* With hashcnt_per_tick hashes per tick, we actually get
1430 : hashcnt_per_tick-1 chances to mixin a microblock. For each tick
1431 : span that we need to reserve, we also need to reserve the hashcnt
1432 : for the tick, hence the +
1433 : max_remaining_microblocks/(hashcnt_per_tick-1) rounded up.
1434 :
1435 : However, if hashcnt_per_tick is 1 because we're in low power mode,
1436 : this should probably just be max_remaining_microblocks. */
1437 0 : ulong max_remaining_ticks_or_microblocks = max_remaining_microblocks;
1438 0 : if( FD_LIKELY( !low_power_mode ) ) max_remaining_ticks_or_microblocks += (max_remaining_microblocks+ctx->hashcnt_per_tick-2UL)/(ctx->hashcnt_per_tick-1UL);
1439 :
1440 0 : ulong restricted_hashcnt = fd_ulong_if( ctx->hashcnt_per_slot>=max_remaining_ticks_or_microblocks, ctx->hashcnt_per_slot-max_remaining_ticks_or_microblocks, 0UL );
1441 :
1442 0 : ulong min_hashcnt = ctx->hashcnt;
1443 :
1444 0 : if( FD_LIKELY( !low_power_mode ) ) {
1445 : /* Recall that there are two kinds of events that will get published
1446 : to the shredder,
1447 :
1448 : (a) Ticks. These occur every 62,500 (hashcnt_per_tick) hashcnts,
1449 : and there will be 64 (ticks_per_slot) of them in each slot.
1450 :
1451 : Ticks must not have any transactions mixed into the hash.
1452 : This is not strictly needed in theory, but is required by the
1453 : current consensus protocol. They get published here in
1454 : after_credit.
1455 :
1456 : (b) Microblocks. These can occur at any other hashcnt, as long
1457 : as it is not a tick. Microblocks cannot be empty, and must
1458 : have at least one transactions mixed in. These get
1459 : published in after_frag.
1460 :
1461 : If hashcnt_per_tick is 1, then we are in low power mode and the
1462 : following does not apply, since we can mix in transactions at any
1463 : time.
1464 :
1465 : In the normal, non-low-power mode, though, we have to be careful
1466 : to make sure that we do not publish microblocks on tick
1467 : boundaries. To do that, we need to obey two rules:
1468 : (i) after_credit must not leave hashcnt one before a tick
1469 : boundary
1470 : (ii) if after_credit begins one before a tick boundary, it must
1471 : advance hashcnt and publish the tick
1472 :
1473 : There's some interplay between min_hashcnt and restricted_hashcnt
1474 : here, and we need to show that there's always a value of
1475 : target_hashcnt we can pick such that
1476 : min_hashcnt <= target_hashcnt <= restricted_hashcnt.
1477 : We'll prove this by induction for current_slot==0 and
1478 : is_leader==true, since all other slots should be the same.
1479 :
1480 : Let m_j and r_j be the min_hashcnt and restricted_hashcnt
1481 : (respectively) for the jth call to after_credit in a slot. We
1482 : want to show that for all values of j, it's possible to pick a
1483 : value h_j, the value of target_hashcnt for the jth call to
1484 : after_credit (which is also the value of hashcnt after
1485 : after_credit has completed) such that m_j<=h_j<=r_j.
1486 :
1487 : Additionally, let T be hashcnt_per_tick and N be ticks_per_slot.
1488 :
1489 : Starting with the base case, j==0. m_j=0, and
1490 : r_0 = N*T - max_microblocks_per_slot
1491 : - ceil(max_microblocks_per_slot/(T-1)).
1492 :
1493 : This is monotonic decreasing in max_microblocks_per_slot, so it
1494 : achieves its minimum when max_microblocks_per_slot is its
1495 : maximum.
1496 : r_0 >= N*T - N*(T-1) - ceil( (N*(T-1))/(T-1))
1497 : = N*T - N*(T-1)-N = 0.
1498 : Thus, m_0 <= r_0, as desired.
1499 :
1500 :
1501 :
1502 : Then, for the inductive step, assume there exists h_j such that
1503 : m_j<=h_j<=r_j, and we want to show that there exists h_{j+1},
1504 : which is the same as showing m_{j+1}<=r_{j+1}.
1505 :
1506 : Let a_j be 1 if we had a microblock immediately following the jth
1507 : call to after_credit, and 0 otherwise. Then hashcnt at the start
1508 : of the (j+1)th call to after_frag is h_j+a_j.
1509 : Also, set b_{j+1}=1 if we are in the case covered by rule (ii)
1510 : above during the (j+1)th call to after_credit, i.e. if
1511 : (h_j+a_j)%T==T-1. Thus, m_{j+1} = h_j + a_j + b_{j+1}.
1512 :
1513 : If we received an additional microblock, then
1514 : max_remaining_microblocks goes down by 1, and
1515 : max_remaining_ticks_or_microblocks goes down by either 1 or 2,
1516 : which means restricted_hashcnt goes up by either 1 or 2. In
1517 : particular, it goes up by 2 if the new value of
1518 : max_remaining_microblocks (at the start of the (j+1)th call to
1519 : after_credit) is congruent to 0 mod T-1. Let b'_{j+1} be 1 if
1520 : this condition is met and 0 otherwise. If we receive a
1521 : done_packing message, restricted_hashcnt can go up by more, but
1522 : we can ignore that case, since it is less restrictive.
1523 : Thus, r_{j+1}=r_j+a_j+b'_{j+1}.
1524 :
1525 : If h_j < r_j (strictly less), then h_j+a_j < r_j+a_j. And thus,
1526 : since b_{j+1}<=b'_{j+1}+1, just by virtue of them both being
1527 : binary,
1528 : h_j + a_j + b_{j+1} < r_j + a_j + b'_{j+1} + 1,
1529 : which is the same (for integers) as
1530 : h_j + a_j + b_{j+1} <= r_j + a_j + b'_{j+1},
1531 : m_{j+1} <= r_{j+1}
1532 :
1533 : On the other hand, if h_j==r_j, this is easy unless b_{j+1}==1,
1534 : which can also only happen if a_j==1. Then (h_j+a_j)%T==T-1,
1535 : which means there's an integer k such that
1536 :
1537 : h_j+a_j==(ticks_per_slot-k)*T-1
1538 : h_j ==ticks_per_slot*T - k*(T-1)-1 - k-1
1539 : ==ticks_per_slot*T - (k*(T-1)+1) - ceil( (k*(T-1)+1)/(T-1) )
1540 :
1541 : Since h_j==r_j in this case, and
1542 : r_j==(ticks_per_slot*T) - max_remaining_microblocks_j - ceil(max_remaining_microblocks_j/(T-1)),
1543 : we can see that the value of max_remaining_microblocks at the
1544 : start of the jth call to after_credit is k*(T-1)+1. Again, since
1545 : a_j==1, then the value of max_remaining_microblocks at the start
1546 : of the j+1th call to after_credit decreases by 1 to k*(T-1),
1547 : which means b'_{j+1}=1.
1548 :
1549 : Thus, h_j + a_j + b_{j+1} == r_j + a_j + b'_{j+1}, so, in
1550 : particular, h_{j+1}<=r_{j+1} as desired. */
1551 0 : min_hashcnt += (ulong)(min_hashcnt%ctx->hashcnt_per_tick == (ctx->hashcnt_per_tick-1UL)); /* add b_{j+1}, enforcing rule (ii) */
1552 0 : }
1553 : /* Now figure out how many hashes are needed to "catch up" the hash
1554 : count to the current system clock, and clamp it to the allowed
1555 : range. */
1556 0 : long now = fd_log_wallclock();
1557 0 : ulong target_hashcnt;
1558 0 : if( FD_LIKELY( !is_leader ) ) {
1559 0 : target_hashcnt = (ulong)((double)(now - ctx->reset_slot_start_ns) / ctx->hashcnt_duration_ns) - (ctx->slot-ctx->reset_slot)*ctx->hashcnt_per_slot;
1560 0 : } else {
1561 : /* We might have gotten very behind on hashes, but if we are leader
1562 : we want to catch up gradually over the remainder of our leader
1563 : slot, not all at once right now. This helps keep the tile from
1564 : being oversubscribed and taking a long time to process incoming
1565 : microblocks. */
1566 0 : long expected_slot_start_ns = ctx->reset_slot_start_ns + (long)((double)(ctx->slot-ctx->reset_slot)*ctx->slot_duration_ns);
1567 0 : double actual_slot_duration_ns = ctx->slot_duration_ns<(double)(ctx->leader_bank_start_ns - expected_slot_start_ns) ? 0.0 : ctx->slot_duration_ns - (double)(ctx->leader_bank_start_ns - expected_slot_start_ns);
1568 0 : double actual_hashcnt_duration_ns = actual_slot_duration_ns / (double)ctx->hashcnt_per_slot;
1569 0 : target_hashcnt = fd_ulong_if( actual_hashcnt_duration_ns==0.0, restricted_hashcnt, (ulong)((double)(now - ctx->leader_bank_start_ns) / actual_hashcnt_duration_ns) );
1570 0 : }
1571 : /* Clamp to [min_hashcnt, restricted_hashcnt] as above */
1572 0 : target_hashcnt = fd_ulong_max( fd_ulong_min( target_hashcnt, restricted_hashcnt ), min_hashcnt );
1573 :
1574 : /* The above proof showed that it was always possible to pick a value
1575 : of target_hashcnt, but we still have a lot of freedom in how to
1576 : pick it. It simplifies the code a lot if we don't keep going after
1577 : a tick in this function. In particular, we want to publish at most
1578 : 1 tick in this call, since otherwise we could consume infinite
1579 : credits to publish here. The credits are set so that we should
1580 : only ever publish one tick during this loop. Also, all the extra
1581 : stuff (leader transitions, publishing ticks, etc.) we have to do
1582 : happens at tick boundaries, so this lets us consolidate all those
1583 : cases.
1584 :
1585 : Mathematically, since the current value of hashcnt is h_j+a_j, the
1586 : next tick (advancing a full tick if we're currently at a tick) is
1587 : t_{j+1} = T*(floor( (h_j+a_j)/T )+1). We need to show that if we set
1588 : h'_{j+1} = min( h_{j+1}, t_{j+1} ), it is still valid.
1589 :
1590 : First, h'_{j+1} <= h_{j+1} <= r_{j+1}, so we're okay in that
1591 : direction.
1592 :
1593 : Next, observe that t_{j+1}>=h_j + a_j + 1, and recall that b_{j+1}
1594 : is 0 or 1. So then,
1595 : t_{j+1} >= h_j+a_j+b_{j+1} = m_{j+1}.
1596 :
1597 : We know h_{j+1) >= m_{j+1} from before, so then h'_{j+1} >=
1598 : m_{j+1}, as desired. */
1599 :
1600 0 : ulong next_tick_hashcnt = ctx->hashcnt_per_tick * (1UL+(ctx->hashcnt/ctx->hashcnt_per_tick));
1601 0 : target_hashcnt = fd_ulong_min( target_hashcnt, next_tick_hashcnt );
1602 :
1603 : /* We still need to enforce rule (i). We know that min_hashcnt%T !=
1604 : T-1 because of rule (ii). That means that if target_hashcnt%T ==
1605 : T-1 at this point, target_hashcnt > min_hashcnt (notice the
1606 : strict), so target_hashcnt-1 >= min_hashcnt and is thus still a
1607 : valid choice for target_hashcnt. */
1608 0 : target_hashcnt -= (ulong)( (!low_power_mode) & ((target_hashcnt%ctx->hashcnt_per_tick)==(ctx->hashcnt_per_tick-1UL)) );
1609 :
1610 0 : FD_TEST( target_hashcnt >= ctx->hashcnt );
1611 0 : FD_TEST( target_hashcnt >= min_hashcnt );
1612 0 : FD_TEST( target_hashcnt <= restricted_hashcnt );
1613 :
1614 0 : if( FD_UNLIKELY( ctx->hashcnt==target_hashcnt ) ) return; /* Nothing to do, don't publish a tick twice */
1615 :
1616 0 : *charge_busy = 1;
1617 :
1618 0 : while( ctx->hashcnt<target_hashcnt ) {
1619 0 : fd_sha256_hash( ctx->hash, 32UL, ctx->hash );
1620 0 : ctx->hashcnt++;
1621 0 : }
1622 :
1623 0 : if( FD_UNLIKELY( ctx->hashcnt==ctx->hashcnt_per_slot ) ) {
1624 0 : ctx->slot++;
1625 0 : ctx->hashcnt = 0UL;
1626 0 : }
1627 :
1628 0 : if( FD_UNLIKELY( !is_leader && !(ctx->hashcnt%ctx->hashcnt_per_tick ) ) ) {
1629 : /* We finished a tick while not leader... save the current hash so
1630 : it can be played back into the bank when we become the leader. */
1631 0 : ulong tick_idx = (ctx->slot*ctx->ticks_per_slot+ctx->hashcnt/ctx->hashcnt_per_tick)%MAX_SKIPPED_TICKS;
1632 0 : fd_memcpy( ctx->skipped_tick_hashes[ tick_idx ], ctx->hash, 32UL );
1633 :
1634 0 : ulong initial_tick_idx = (ctx->last_slot*ctx->ticks_per_slot+ctx->last_hashcnt/ctx->hashcnt_per_tick)%MAX_SKIPPED_TICKS;
1635 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", ctx->last_slot, ctx->slot ));
1636 0 : }
1637 :
1638 0 : if( FD_UNLIKELY( is_leader && !(ctx->hashcnt%ctx->hashcnt_per_tick) ) ) {
1639 : /* We ticked while leader... tell the leader bank. */
1640 0 : fd_ext_poh_register_tick( ctx->current_leader_bank, ctx->hash );
1641 :
1642 : /* And send an empty microblock (a tick) to the shred tile. */
1643 0 : publish_tick( ctx, stem, ctx->hash, 0 );
1644 0 : }
1645 :
1646 0 : if( FD_UNLIKELY( !is_leader && ctx->slot>=ctx->next_leader_slot ) ) {
1647 : /* We ticked while not leader and are now leader... transition
1648 : the state machine. */
1649 0 : publish_plugin_slot_start( ctx, ctx->next_leader_slot, ctx->reset_slot );
1650 0 : FD_LOG_INFO(( "fd_poh_ticked_into_leader(slot=%lu, reset_slot=%lu)", ctx->next_leader_slot, ctx->reset_slot ));
1651 0 : }
1652 :
1653 0 : if( FD_UNLIKELY( is_leader && ctx->slot>ctx->next_leader_slot ) ) {
1654 : /* We ticked while leader and are no longer leader... transition
1655 : the state machine. */
1656 0 : FD_TEST( !max_remaining_microblocks );
1657 0 : publish_plugin_slot_end( ctx, ctx->next_leader_slot, ctx->cus_used );
1658 0 : FD_LOG_INFO(( "fd_poh_ticked_outof_leader(slot=%lu)", ctx->next_leader_slot ));
1659 :
1660 0 : no_longer_leader( ctx );
1661 0 : ctx->expect_sequential_leader_slot = ctx->slot;
1662 :
1663 0 : double tick_per_ns = fd_tempo_tick_per_ns( NULL );
1664 0 : fd_histf_sample( ctx->slot_done_delay, (ulong)((double)(fd_log_wallclock()-ctx->reset_slot_start_ns)/tick_per_ns) );
1665 0 : ctx->next_leader_slot = next_leader_slot( ctx );
1666 :
1667 0 : if( FD_UNLIKELY( ctx->slot>=ctx->next_leader_slot ) ) {
1668 : /* We finished a leader slot, and are immediately leader for the
1669 : following slot... transition. */
1670 0 : publish_plugin_slot_start( ctx, ctx->next_leader_slot, ctx->next_leader_slot-1UL );
1671 0 : FD_LOG_INFO(( "fd_poh_ticked_into_leader(slot=%lu, reset_slot=%lu)", ctx->next_leader_slot, ctx->next_leader_slot-1UL ));
1672 0 : }
1673 0 : }
1674 0 : }
1675 :
1676 : static inline void
1677 0 : during_housekeeping( fd_poh_ctx_t * ctx ) {
1678 0 : if( FD_UNLIKELY( maybe_change_identity( ctx, 0 ) ) ) {
1679 0 : ctx->next_leader_slot = next_leader_slot( ctx );
1680 0 : FD_LOG_INFO(( "fd_poh_identity_changed(next_leader_slot=%lu)", ctx->next_leader_slot ));
1681 :
1682 : /* Signal replay to check if we are leader again, in-case it's stuck
1683 : because everything already replayed. */
1684 0 : FD_COMPILER_MFENCE();
1685 0 : fd_ext_poh_signal_leader_change( ctx->signal_leader_change );
1686 0 : }
1687 0 : }
1688 :
1689 : static inline void
1690 0 : metrics_write( fd_poh_ctx_t * ctx ) {
1691 0 : FD_MHIST_COPY( POH, BEGIN_LEADER_DELAY_SECONDS, ctx->begin_leader_delay );
1692 0 : FD_MHIST_COPY( POH, FIRST_MICROBLOCK_DELAY_SECONDS, ctx->first_microblock_delay );
1693 0 : FD_MHIST_COPY( POH, SLOT_DONE_DELAY_SECONDS, ctx->slot_done_delay );
1694 0 : FD_MHIST_COPY( POH, BUNDLE_INITIALIZE_DELAY_SECONDS, ctx->bundle_init_delay );
1695 0 : }
1696 :
1697 : static int
1698 : before_frag( fd_poh_ctx_t * ctx,
1699 : ulong in_idx,
1700 : ulong seq,
1701 0 : ulong sig ) {
1702 0 : (void)seq;
1703 :
1704 0 : if( FD_LIKELY( ctx->in_kind[ in_idx ]==IN_KIND_BANK ) ) {
1705 0 : ulong microblock_idx = fd_disco_bank_sig_microblock_idx( sig );
1706 0 : FD_TEST( microblock_idx>=ctx->expect_microblock_idx );
1707 :
1708 : /* Return the fragment to stem so we can process it later, if it's
1709 : not next in the sequence. */
1710 0 : if( FD_UNLIKELY( microblock_idx>ctx->expect_microblock_idx ) ) return -1;
1711 :
1712 0 : ctx->expect_microblock_idx++;
1713 0 : }
1714 :
1715 0 : return 0;
1716 0 : }
1717 :
1718 : static inline void
1719 : during_frag( fd_poh_ctx_t * ctx,
1720 : ulong in_idx,
1721 : ulong seq FD_PARAM_UNUSED,
1722 : ulong sig,
1723 : ulong chunk,
1724 : ulong sz,
1725 0 : ulong ctl FD_PARAM_UNUSED ) {
1726 :
1727 0 : ctx->skip_frag = 0;
1728 :
1729 0 : if( FD_UNLIKELY( ctx->in_kind[ in_idx ]==IN_KIND_STAKE ) ) {
1730 0 : if( FD_UNLIKELY( chunk<ctx->in[ in_idx ].chunk0 || chunk>ctx->in[ in_idx ].wmark ) )
1731 0 : FD_LOG_ERR(( "chunk %lu %lu corrupt, not in range [%lu,%lu]", chunk, sz,
1732 0 : ctx->in[ in_idx ].chunk0, ctx->in[ in_idx ].wmark ));
1733 :
1734 0 : uchar const * dcache_entry = fd_chunk_to_laddr_const( ctx->in[ in_idx ].mem, chunk );
1735 0 : fd_stake_ci_stake_msg_init( ctx->stake_ci, dcache_entry );
1736 0 : return;
1737 0 : }
1738 :
1739 0 : ulong pkt_type;
1740 0 : ulong slot;
1741 0 : switch( ctx->in_kind[ in_idx ] ) {
1742 0 : case IN_KIND_BANK: {
1743 0 : pkt_type = POH_PKT_TYPE_MICROBLOCK;
1744 0 : slot = fd_disco_bank_sig_slot( sig );
1745 0 : break;
1746 0 : }
1747 0 : case IN_KIND_PACK: {
1748 0 : pkt_type = fd_disco_poh_sig_pkt_type( sig );
1749 0 : slot = fd_disco_poh_sig_slot( sig );
1750 0 : break;
1751 0 : }
1752 0 : default:
1753 0 : FD_LOG_ERR(( "unexpected in_kind %d", ctx->in_kind[ in_idx ] ));
1754 0 : }
1755 :
1756 0 : int is_frag_for_prior_leader_slot = 0;
1757 0 : if( FD_LIKELY( pkt_type==POH_PKT_TYPE_DONE_PACKING || pkt_type==POH_PKT_TYPE_MICROBLOCK ) ) {
1758 : /* The following sequence is possible...
1759 :
1760 : 1. We become leader in slot 10
1761 : 2. While leader, we switch to a fork that is on slot 8, where
1762 : we are leader
1763 : 3. We get the in-flight microblocks for slot 10
1764 :
1765 : These in-flight microblocks need to be dropped, so we check
1766 : against the high water mark (highwater_leader_slot) rather than
1767 : the current hashcnt here when determining what to drop.
1768 :
1769 : We know if the slot is lower than the high water mark it's from a stale
1770 : leader slot, because we will not become leader for the same slot twice
1771 : even if we are reset back in time (to prevent duplicate blocks). */
1772 0 : is_frag_for_prior_leader_slot = slot<ctx->highwater_leader_slot;
1773 0 : }
1774 :
1775 0 : if( FD_UNLIKELY( ctx->in_kind[ in_idx ]==IN_KIND_PACK ) ) {
1776 : /* We now know the real amount of microblocks published, so set an
1777 : exact bound for once we receive them. */
1778 0 : ctx->skip_frag = 1;
1779 0 : if( pkt_type==POH_PKT_TYPE_DONE_PACKING ) {
1780 0 : if( FD_UNLIKELY( is_frag_for_prior_leader_slot ) ) return;
1781 :
1782 0 : FD_TEST( ctx->microblocks_lower_bound<=ctx->max_microblocks_per_slot );
1783 0 : fd_done_packing_t const * done_packing = fd_chunk_to_laddr( ctx->in[ in_idx ].mem, chunk );
1784 0 : FD_LOG_INFO(( "done_packing(slot=%lu,seen_microblocks=%lu,microblocks_in_slot=%lu)",
1785 0 : ctx->slot,
1786 0 : ctx->microblocks_lower_bound,
1787 0 : done_packing->microblocks_in_slot ));
1788 0 : ctx->microblocks_lower_bound += ctx->max_microblocks_per_slot - done_packing->microblocks_in_slot;
1789 0 : }
1790 0 : return;
1791 0 : } else {
1792 0 : if( FD_UNLIKELY( chunk<ctx->in[ in_idx ].chunk0 || chunk>ctx->in[ in_idx ].wmark || sz>USHORT_MAX ) )
1793 0 : FD_LOG_ERR(( "chunk %lu %lu corrupt, not in range [%lu,%lu]", chunk, sz, ctx->in[ in_idx ].chunk0, ctx->in[ in_idx ].wmark ));
1794 :
1795 0 : uchar * src = (uchar *)fd_chunk_to_laddr( ctx->in[ in_idx ].mem, chunk );
1796 :
1797 0 : fd_memcpy( ctx->_txns, src, sz-sizeof(fd_microblock_trailer_t) );
1798 0 : fd_memcpy( ctx->_microblock_trailer, src+sz-sizeof(fd_microblock_trailer_t), sizeof(fd_microblock_trailer_t) );
1799 :
1800 0 : ctx->skip_frag = is_frag_for_prior_leader_slot;
1801 0 : }
1802 0 : }
1803 :
1804 : static void
1805 : publish_microblock( fd_poh_ctx_t * ctx,
1806 : fd_stem_context_t * stem,
1807 : ulong slot,
1808 : ulong hashcnt_delta,
1809 0 : ulong txn_cnt ) {
1810 0 : uchar * dst = (uchar *)fd_chunk_to_laddr( ctx->shred_out->mem, ctx->shred_out->chunk );
1811 0 : FD_TEST( slot>=ctx->reset_slot );
1812 0 : fd_entry_batch_meta_t * meta = (fd_entry_batch_meta_t *)dst;
1813 0 : meta->parent_offset = 1UL+slot-ctx->reset_slot;
1814 0 : meta->reference_tick = (ctx->hashcnt/ctx->hashcnt_per_tick) % ctx->ticks_per_slot;
1815 0 : meta->block_complete = !ctx->hashcnt;
1816 :
1817 0 : dst += sizeof(fd_entry_batch_meta_t);
1818 0 : fd_entry_batch_header_t * header = (fd_entry_batch_header_t *)dst;
1819 0 : header->hashcnt_delta = hashcnt_delta;
1820 0 : fd_memcpy( header->hash, ctx->hash, 32UL );
1821 :
1822 0 : dst += sizeof(fd_entry_batch_header_t);
1823 0 : ulong payload_sz = 0UL;
1824 0 : ulong included_txn_cnt = 0UL;
1825 0 : for( ulong i=0UL; i<txn_cnt; i++ ) {
1826 0 : fd_txn_p_t * txn = (fd_txn_p_t *)(ctx->_txns + i*sizeof(fd_txn_p_t));
1827 0 : if( FD_UNLIKELY( !(txn->flags & FD_TXN_P_FLAGS_EXECUTE_SUCCESS) ) ) continue;
1828 :
1829 0 : fd_memcpy( dst, txn->payload, txn->payload_sz );
1830 0 : payload_sz += txn->payload_sz;
1831 0 : dst += txn->payload_sz;
1832 0 : included_txn_cnt++;
1833 0 : }
1834 0 : header->txn_cnt = included_txn_cnt;
1835 :
1836 : /* We always have credits to publish here, because we have a burst
1837 : value of 3 credits, and at most we will publish_tick() once and
1838 : then publish_became_leader() once, leaving one credit here to
1839 : publish the microblock. */
1840 0 : ulong tspub = (ulong)fd_frag_meta_ts_comp( fd_tickcount() );
1841 0 : ulong sz = sizeof(fd_entry_batch_meta_t)+sizeof(fd_entry_batch_header_t)+payload_sz;
1842 0 : ulong new_sig = fd_disco_poh_sig( slot, POH_PKT_TYPE_MICROBLOCK, 0UL );
1843 0 : fd_stem_publish( stem, ctx->shred_out->idx, new_sig, ctx->shred_out->chunk, sz, 0UL, 0UL, tspub );
1844 0 : ctx->shred_seq = stem->seqs[ ctx->shred_out->idx ];
1845 0 : ctx->shred_out->chunk = fd_dcache_compact_next( ctx->shred_out->chunk, sz, ctx->shred_out->chunk0, ctx->shred_out->wmark );
1846 0 : }
1847 :
1848 : static inline void
1849 : after_frag( fd_poh_ctx_t * ctx,
1850 : ulong in_idx,
1851 : ulong seq,
1852 : ulong sig,
1853 : ulong sz,
1854 : ulong tsorig,
1855 0 : fd_stem_context_t * stem ) {
1856 0 : (void)in_idx;
1857 0 : (void)seq;
1858 0 : (void)tsorig;
1859 :
1860 0 : if( FD_UNLIKELY( ctx->skip_frag ) ) return;
1861 :
1862 0 : if( FD_UNLIKELY( ctx->in_kind[ in_idx ]==IN_KIND_STAKE ) ) {
1863 0 : fd_stake_ci_stake_msg_fini( ctx->stake_ci );
1864 : /* It might seem like we do not need to do state transitions in and
1865 : out of being the leader here, since leader schedule updates are
1866 : always one epoch in advance (whether we are leader or not would
1867 : never change for the currently executing slot) but this is not
1868 : true for new ledgers when the validator first boots. We will
1869 : likely be the leader in slot 1, and get notified of the leader
1870 : schedule for that slot while we are still in it.
1871 :
1872 : For safety we just handle both transitions, in and out, although
1873 : the only one possible should be into leader. */
1874 0 : ulong next_leader_slot_after_frag = next_leader_slot( ctx );
1875 :
1876 0 : int currently_leader = ctx->slot>=ctx->next_leader_slot;
1877 0 : int leader_after_frag = ctx->slot>=next_leader_slot_after_frag;
1878 :
1879 0 : FD_LOG_INFO(( "stake_update(before_leader=%lu,after_leader=%lu)",
1880 0 : ctx->next_leader_slot,
1881 0 : next_leader_slot_after_frag ));
1882 :
1883 0 : ctx->next_leader_slot = next_leader_slot_after_frag;
1884 0 : if( FD_UNLIKELY( currently_leader && !leader_after_frag ) ) {
1885 : /* Shouldn't ever happen, otherwise we need to do a state
1886 : transition out of being leader. */
1887 0 : FD_LOG_ERR(( "stake update caused us to no longer be leader in an active slot" ));
1888 0 : }
1889 :
1890 : /* Nothing to do if we transition into being leader, since it
1891 : will just get picked up by the regular tick loop. */
1892 0 : if( FD_UNLIKELY( !currently_leader && leader_after_frag ) ) {
1893 0 : publish_plugin_slot_start( ctx, next_leader_slot_after_frag, ctx->reset_slot );
1894 0 : }
1895 :
1896 0 : return;
1897 0 : }
1898 :
1899 0 : if( FD_UNLIKELY( !ctx->microblocks_lower_bound ) ) {
1900 0 : double tick_per_ns = fd_tempo_tick_per_ns( NULL );
1901 0 : fd_histf_sample( ctx->first_microblock_delay, (ulong)((double)(fd_log_wallclock()-ctx->reset_slot_start_ns)/tick_per_ns) );
1902 0 : }
1903 :
1904 0 : ulong target_slot = fd_disco_bank_sig_slot( sig );
1905 :
1906 0 : if( FD_UNLIKELY( target_slot!=ctx->next_leader_slot || target_slot!=ctx->slot ) ) {
1907 0 : FD_LOG_ERR(( "packed too early or late target_slot=%lu, current_slot=%lu. highwater_leader_slot=%lu",
1908 0 : target_slot, ctx->slot, ctx->highwater_leader_slot ));
1909 0 : }
1910 :
1911 0 : FD_TEST( ctx->current_leader_bank );
1912 0 : FD_TEST( ctx->microblocks_lower_bound<ctx->max_microblocks_per_slot );
1913 0 : ctx->microblocks_lower_bound += 1UL;
1914 :
1915 0 : ulong txn_cnt = (sz-sizeof(fd_microblock_trailer_t))/sizeof(fd_txn_p_t);
1916 0 : fd_txn_p_t * txns = (fd_txn_p_t *)(ctx->_txns);
1917 0 : ulong executed_txn_cnt = 0UL;
1918 0 : ulong cus_used = 0UL;
1919 0 : for( ulong i=0UL; i<txn_cnt; i++ ) {
1920 0 : if( FD_LIKELY( txns[ i ].flags & FD_TXN_P_FLAGS_EXECUTE_SUCCESS ) ) {
1921 0 : executed_txn_cnt++;
1922 0 : cus_used += txns[ i ].bank_cu.actual_consumed_cus;
1923 0 : }
1924 0 : }
1925 :
1926 : /* We don't publish transactions that fail to execute. If all the
1927 : transactions failed to execute, the microblock would be empty,
1928 : causing agave to think it's a tick and complain. Instead, we just
1929 : skip the microblock and don't hash or update the hashcnt. */
1930 0 : if( FD_UNLIKELY( !executed_txn_cnt ) ) return;
1931 :
1932 0 : uchar data[ 64 ];
1933 0 : fd_memcpy( data, ctx->hash, 32UL );
1934 0 : fd_memcpy( data+32UL, ctx->_microblock_trailer->hash, 32UL );
1935 0 : fd_sha256_hash( data, 64UL, ctx->hash );
1936 :
1937 0 : ctx->hashcnt++;
1938 0 : FD_TEST( ctx->hashcnt>ctx->last_hashcnt );
1939 0 : ulong hashcnt_delta = ctx->hashcnt - ctx->last_hashcnt;
1940 :
1941 : /* The hashing loop above will never leave us exactly one away from
1942 : crossing a tick boundary, so this increment will never cause the
1943 : current tick (or the slot) to change, except in low power mode
1944 : for development, in which case we do need to register the tick
1945 : with the leader bank. We don't need to publish the tick since
1946 : sending the microblock below is the publishing action. */
1947 0 : if( FD_UNLIKELY( !(ctx->hashcnt%ctx->hashcnt_per_slot ) ) ) {
1948 0 : ctx->slot++;
1949 0 : ctx->hashcnt = 0UL;
1950 0 : }
1951 :
1952 0 : ctx->last_slot = ctx->slot;
1953 0 : ctx->last_hashcnt = ctx->hashcnt;
1954 :
1955 0 : ctx->cus_used += cus_used;
1956 :
1957 0 : if( FD_UNLIKELY( !(ctx->hashcnt%ctx->hashcnt_per_tick ) ) ) {
1958 0 : fd_ext_poh_register_tick( ctx->current_leader_bank, ctx->hash );
1959 0 : if( FD_UNLIKELY( ctx->slot>ctx->next_leader_slot ) ) {
1960 : /* We ticked while leader and are no longer leader... transition
1961 : the state machine. */
1962 0 : publish_plugin_slot_end( ctx, ctx->next_leader_slot, ctx->cus_used );
1963 :
1964 0 : no_longer_leader( ctx );
1965 :
1966 0 : if( FD_UNLIKELY( ctx->slot>=ctx->next_leader_slot ) ) {
1967 : /* We finished a leader slot, and are immediately leader for the
1968 : following slot... transition. */
1969 0 : publish_plugin_slot_start( ctx, ctx->next_leader_slot, ctx->next_leader_slot-1UL );
1970 0 : }
1971 0 : }
1972 0 : }
1973 :
1974 0 : publish_microblock( ctx, stem, target_slot, hashcnt_delta, txn_cnt );
1975 0 : }
1976 :
1977 : static void
1978 : privileged_init( fd_topo_t * topo,
1979 0 : fd_topo_tile_t * tile ) {
1980 0 : void * scratch = fd_topo_obj_laddr( topo, tile->tile_obj_id );
1981 :
1982 0 : FD_SCRATCH_ALLOC_INIT( l, scratch );
1983 0 : fd_poh_ctx_t * ctx = FD_SCRATCH_ALLOC_APPEND( l, alignof( fd_poh_ctx_t ), sizeof( fd_poh_ctx_t ) );
1984 :
1985 0 : if( FD_UNLIKELY( !strcmp( tile->poh.identity_key_path, "" ) ) )
1986 0 : FD_LOG_ERR(( "identity_key_path not set" ));
1987 :
1988 0 : const uchar * identity_key = fd_keyload_load( tile->poh.identity_key_path, /* pubkey only: */ 1 );
1989 0 : fd_memcpy( ctx->identity_key.uc, identity_key, 32UL );
1990 :
1991 0 : if( FD_UNLIKELY( tile->poh.bundle.enabled ) ) {
1992 0 : if( FD_UNLIKELY( !fd_base58_decode_32( tile->poh.bundle.vote_account_path, ctx->bundle.vote_account.uc ) ) ) {
1993 0 : const uchar * vote_key = fd_keyload_load( tile->poh.bundle.vote_account_path, /* pubkey only: */ 1 );
1994 0 : fd_memcpy( ctx->bundle.vote_account.uc, vote_key, 32UL );
1995 0 : }
1996 0 : }
1997 0 : }
1998 :
1999 : /* The Agave client needs to communicate to the shred tile what
2000 : the shred version is on boot, but shred tile does not live in the
2001 : same address space, so have the PoH tile pass the value through
2002 : via. a shared memory ulong. */
2003 :
2004 : static volatile ulong * fd_shred_version;
2005 :
2006 : void
2007 0 : fd_ext_shred_set_shred_version( ulong shred_version ) {
2008 0 : while( FD_UNLIKELY( !fd_shred_version ) ) FD_SPIN_PAUSE();
2009 0 : *fd_shred_version = shred_version;
2010 0 : }
2011 :
2012 : void
2013 : fd_ext_poh_publish_gossip_vote( uchar * data,
2014 0 : ulong data_len ) {
2015 0 : poh_link_publish( &gossip_dedup, 1UL, data, data_len );
2016 0 : }
2017 :
2018 : void
2019 : fd_ext_poh_publish_leader_schedule( uchar * data,
2020 0 : ulong data_len ) {
2021 0 : poh_link_publish( &stake_out, 2UL, data, data_len );
2022 0 : }
2023 :
2024 : void
2025 : fd_ext_poh_publish_cluster_info( uchar * data,
2026 0 : ulong data_len ) {
2027 0 : poh_link_publish( &crds_shred, 2UL, data, data_len );
2028 0 : }
2029 :
2030 : void
2031 : fd_ext_plugin_publish_replay_stage( ulong sig,
2032 : uchar * data,
2033 0 : ulong data_len ) {
2034 0 : poh_link_publish( &replay_plugin, sig, data, data_len );
2035 0 : }
2036 :
2037 : void
2038 : fd_ext_plugin_publish_genesis_hash( ulong sig,
2039 : uchar * data,
2040 0 : ulong data_len ) {
2041 0 : poh_link_publish( &replay_plugin, sig, data, data_len );
2042 0 : }
2043 :
2044 : void
2045 : fd_ext_plugin_publish_start_progress( ulong sig,
2046 : uchar * data,
2047 0 : ulong data_len ) {
2048 0 : poh_link_publish( &start_progress_plugin, sig, data, data_len );
2049 0 : }
2050 :
2051 : void
2052 : fd_ext_plugin_publish_vote_listener( ulong sig,
2053 : uchar * data,
2054 0 : ulong data_len ) {
2055 0 : poh_link_publish( &vote_listener_plugin, sig, data, data_len );
2056 0 : }
2057 :
2058 : void
2059 : fd_ext_plugin_publish_validator_info( ulong sig,
2060 : uchar * data,
2061 0 : ulong data_len ) {
2062 0 : poh_link_publish( &validator_info_plugin, sig, data, data_len );
2063 0 : }
2064 :
2065 : void
2066 : fd_ext_plugin_publish_periodic( ulong sig,
2067 : uchar * data,
2068 0 : ulong data_len ) {
2069 0 : poh_link_publish( &gossip_plugin, sig, data, data_len );
2070 0 : }
2071 :
2072 : void
2073 : fd_ext_resolv_publish_root_bank( uchar * data,
2074 0 : ulong data_len ) {
2075 0 : poh_link_publish( &replay_resolv, 0UL, data, data_len );
2076 0 : }
2077 :
2078 : void
2079 : fd_ext_resolv_publish_completed_blockhash( uchar * data,
2080 0 : ulong data_len ) {
2081 0 : poh_link_publish( &replay_resolv, 1UL, data, data_len );
2082 0 : }
2083 :
2084 : static inline fd_poh_out_ctx_t
2085 : out1( fd_topo_t const * topo,
2086 : fd_topo_tile_t const * tile,
2087 0 : char const * name ) {
2088 0 : ulong idx = ULONG_MAX;
2089 :
2090 0 : for( ulong i=0UL; i<tile->out_cnt; i++ ) {
2091 0 : fd_topo_link_t const * link = &topo->links[ tile->out_link_id[ i ] ];
2092 0 : if( !strcmp( link->name, name ) ) {
2093 0 : if( FD_UNLIKELY( idx!=ULONG_MAX ) ) FD_LOG_ERR(( "tile %s:%lu had multiple output links named %s but expected one", tile->name, tile->kind_id, name ));
2094 0 : idx = i;
2095 0 : }
2096 0 : }
2097 :
2098 0 : if( FD_UNLIKELY( idx==ULONG_MAX ) ) FD_LOG_ERR(( "tile %s:%lu had no output link named %s", tile->name, tile->kind_id, name ));
2099 :
2100 0 : void * mem = topo->workspaces[ topo->objs[ topo->links[ tile->out_link_id[ idx ] ].dcache_obj_id ].wksp_id ].wksp;
2101 0 : ulong chunk0 = fd_dcache_compact_chunk0( mem, topo->links[ tile->out_link_id[ idx ] ].dcache );
2102 0 : ulong wmark = fd_dcache_compact_wmark ( mem, topo->links[ tile->out_link_id[ idx ] ].dcache, topo->links[ tile->out_link_id[ idx ] ].mtu );
2103 :
2104 0 : return (fd_poh_out_ctx_t){ .idx = idx, .mem = mem, .chunk0 = chunk0, .wmark = wmark, .chunk = chunk0 };
2105 0 : }
2106 :
2107 : static void
2108 : unprivileged_init( fd_topo_t * topo,
2109 0 : fd_topo_tile_t * tile ) {
2110 0 : void * scratch = fd_topo_obj_laddr( topo, tile->tile_obj_id );
2111 :
2112 0 : FD_SCRATCH_ALLOC_INIT( l, scratch );
2113 0 : fd_poh_ctx_t * ctx = FD_SCRATCH_ALLOC_APPEND( l, alignof( fd_poh_ctx_t ), sizeof( fd_poh_ctx_t ) );
2114 0 : void * stake_ci = FD_SCRATCH_ALLOC_APPEND( l, fd_stake_ci_align(), fd_stake_ci_footprint() );
2115 0 : void * sha256 = FD_SCRATCH_ALLOC_APPEND( l, FD_SHA256_ALIGN, FD_SHA256_FOOTPRINT );
2116 :
2117 0 : #define NONNULL( x ) (__extension__({ \
2118 0 : __typeof__((x)) __x = (x); \
2119 0 : if( FD_UNLIKELY( !__x ) ) FD_LOG_ERR(( #x " was unexpectedly NULL" )); \
2120 0 : __x; }))
2121 :
2122 0 : ctx->stake_ci = NONNULL( fd_stake_ci_join( fd_stake_ci_new( stake_ci, &ctx->identity_key ) ) );
2123 0 : ctx->sha256 = NONNULL( fd_sha256_join( fd_sha256_new( sha256 ) ) );
2124 0 : ctx->current_leader_bank = NULL;
2125 0 : ctx->signal_leader_change = NULL;
2126 :
2127 0 : ctx->shred_seq = ULONG_MAX;
2128 0 : ctx->halted_switching_key = 0;
2129 0 : ctx->keyswitch = fd_keyswitch_join( fd_topo_obj_laddr( topo, tile->keyswitch_obj_id ) );
2130 0 : FD_TEST( ctx->keyswitch );
2131 :
2132 0 : ctx->slot = 0UL;
2133 0 : ctx->hashcnt = 0UL;
2134 0 : ctx->last_hashcnt = 0UL;
2135 0 : ctx->highwater_leader_slot = ULONG_MAX;
2136 0 : ctx->next_leader_slot = ULONG_MAX;
2137 0 : ctx->reset_slot = ULONG_MAX;
2138 :
2139 0 : ctx->lagged_consecutive_leader_start = tile->poh.lagged_consecutive_leader_start;
2140 0 : ctx->expect_sequential_leader_slot = ULONG_MAX;
2141 :
2142 0 : ctx->microblocks_lower_bound = 0UL;
2143 :
2144 0 : ctx->max_active_descendant = 0UL;
2145 :
2146 0 : if( FD_UNLIKELY( tile->poh.bundle.enabled ) ) {
2147 0 : ctx->bundle.enabled = 1;
2148 0 : NONNULL( fd_bundle_crank_gen_init( ctx->bundle.gen, (fd_acct_addr_t const *)tile->poh.bundle.tip_distribution_program_addr,
2149 0 : (fd_acct_addr_t const *)tile->poh.bundle.tip_payment_program_addr,
2150 0 : (fd_acct_addr_t const *)ctx->bundle.vote_account.uc,
2151 0 : (fd_acct_addr_t const *)ctx->bundle.vote_account.uc, 0UL ) ); /* last two arguments are properly bogus */
2152 0 : } else {
2153 0 : ctx->bundle.enabled = 0;
2154 0 : }
2155 :
2156 0 : ulong poh_shred_obj_id = fd_pod_query_ulong( topo->props, "poh_shred", ULONG_MAX );
2157 0 : FD_TEST( poh_shred_obj_id!=ULONG_MAX );
2158 :
2159 0 : fd_shred_version = fd_fseq_join( fd_topo_obj_laddr( topo, poh_shred_obj_id ) );
2160 0 : FD_TEST( fd_shred_version );
2161 :
2162 0 : poh_link_init( &gossip_dedup, topo, tile, out1( topo, tile, "gossip_dedup" ).idx );
2163 0 : poh_link_init( &stake_out, topo, tile, out1( topo, tile, "stake_out" ).idx );
2164 0 : poh_link_init( &crds_shred, topo, tile, out1( topo, tile, "crds_shred" ).idx );
2165 0 : poh_link_init( &replay_resolv, topo, tile, out1( topo, tile, "replay_resol" ).idx );
2166 :
2167 0 : if( FD_LIKELY( tile->poh.plugins_enabled ) ) {
2168 0 : poh_link_init( &replay_plugin, topo, tile, out1( topo, tile, "replay_plugi" ).idx );
2169 0 : poh_link_init( &gossip_plugin, topo, tile, out1( topo, tile, "gossip_plugi" ).idx );
2170 0 : poh_link_init( &start_progress_plugin, topo, tile, out1( topo, tile, "startp_plugi" ).idx );
2171 0 : poh_link_init( &vote_listener_plugin, topo, tile, out1( topo, tile, "votel_plugin" ).idx );
2172 0 : poh_link_init( &validator_info_plugin, topo, tile, out1( topo, tile, "valcfg_plugi" ).idx );
2173 0 : } else {
2174 : /* Mark these mcaches as "available", so the system boots, but the
2175 : memory is not set so nothing will actually get published via.
2176 : the links. */
2177 0 : FD_COMPILER_MFENCE();
2178 0 : replay_plugin.mcache = (fd_frag_meta_t*)1;
2179 0 : gossip_plugin.mcache = (fd_frag_meta_t*)1;
2180 0 : start_progress_plugin.mcache = (fd_frag_meta_t*)1;
2181 0 : vote_listener_plugin.mcache = (fd_frag_meta_t*)1;
2182 0 : validator_info_plugin.mcache = (fd_frag_meta_t*)1;
2183 0 : FD_COMPILER_MFENCE();
2184 0 : }
2185 :
2186 0 : FD_LOG_INFO(( "PoH waiting to be initialized by Agave client... %lu %lu", fd_poh_waiting_lock, fd_poh_returned_lock ));
2187 0 : FD_VOLATILE( fd_poh_global_ctx ) = ctx;
2188 0 : FD_COMPILER_MFENCE();
2189 0 : for(;;) {
2190 0 : if( FD_LIKELY( FD_VOLATILE_CONST( fd_poh_waiting_lock ) ) ) break;
2191 0 : FD_SPIN_PAUSE();
2192 0 : }
2193 0 : FD_VOLATILE( fd_poh_waiting_lock ) = 0UL;
2194 0 : FD_VOLATILE( fd_poh_returned_lock ) = 1UL;
2195 0 : FD_COMPILER_MFENCE();
2196 0 : for(;;) {
2197 0 : if( FD_UNLIKELY( !FD_VOLATILE_CONST( fd_poh_returned_lock ) ) ) break;
2198 0 : FD_SPIN_PAUSE();
2199 0 : }
2200 0 : FD_COMPILER_MFENCE();
2201 :
2202 0 : if( FD_UNLIKELY( ctx->reset_slot==ULONG_MAX ) ) FD_LOG_ERR(( "PoH was not initialized by Agave client" ));
2203 :
2204 0 : fd_histf_join( fd_histf_new( ctx->begin_leader_delay, FD_MHIST_SECONDS_MIN( POH, BEGIN_LEADER_DELAY_SECONDS ),
2205 0 : FD_MHIST_SECONDS_MAX( POH, BEGIN_LEADER_DELAY_SECONDS ) ) );
2206 0 : fd_histf_join( fd_histf_new( ctx->first_microblock_delay, FD_MHIST_SECONDS_MIN( POH, FIRST_MICROBLOCK_DELAY_SECONDS ),
2207 0 : FD_MHIST_SECONDS_MAX( POH, FIRST_MICROBLOCK_DELAY_SECONDS ) ) );
2208 0 : fd_histf_join( fd_histf_new( ctx->slot_done_delay, FD_MHIST_SECONDS_MIN( POH, SLOT_DONE_DELAY_SECONDS ),
2209 0 : FD_MHIST_SECONDS_MAX( POH, SLOT_DONE_DELAY_SECONDS ) ) );
2210 :
2211 0 : fd_histf_join( fd_histf_new( ctx->bundle_init_delay, FD_MHIST_SECONDS_MIN( POH, BUNDLE_INITIALIZE_DELAY_SECONDS ),
2212 0 : FD_MHIST_SECONDS_MAX( POH, BUNDLE_INITIALIZE_DELAY_SECONDS ) ) );
2213 :
2214 0 : for( ulong i=0UL; i<tile->in_cnt; i++ ) {
2215 0 : fd_topo_link_t * link = &topo->links[ tile->in_link_id[ i ] ];
2216 0 : fd_topo_wksp_t * link_wksp = &topo->workspaces[ topo->objs[ link->dcache_obj_id ].wksp_id ];
2217 :
2218 0 : ctx->in[ i ].mem = link_wksp->wksp;
2219 0 : ctx->in[ i ].chunk0 = fd_dcache_compact_chunk0( ctx->in[ i ].mem, link->dcache );
2220 0 : ctx->in[ i ].wmark = fd_dcache_compact_wmark ( ctx->in[ i ].mem, link->dcache, link->mtu );
2221 :
2222 0 : if( FD_UNLIKELY( !strcmp( link->name, "stake_out" ) ) ) {
2223 0 : ctx->in_kind[ i ] = IN_KIND_STAKE;
2224 0 : } else if( FD_UNLIKELY( !strcmp( link->name, "pack_bank" ) ) ) {
2225 0 : ctx->in_kind[ i ] = IN_KIND_PACK;
2226 0 : } else if( FD_LIKELY( !strcmp( link->name, "bank_poh" ) ) ) {
2227 0 : ctx->in_kind[ i ] = IN_KIND_BANK;
2228 0 : } else {
2229 0 : FD_LOG_ERR(( "unexpected input link name %s", link->name ));
2230 0 : }
2231 0 : }
2232 :
2233 0 : *ctx->shred_out = out1( topo, tile, "poh_shred" );
2234 0 : *ctx->pack_out = out1( topo, tile, "poh_pack" );
2235 0 : ctx->plugin_out->mem = NULL;
2236 0 : if( FD_LIKELY( tile->poh.plugins_enabled ) ) {
2237 0 : *ctx->plugin_out = out1( topo, tile, "poh_plugin" );
2238 0 : }
2239 :
2240 0 : ulong scratch_top = FD_SCRATCH_ALLOC_FINI( l, 1UL );
2241 0 : if( FD_UNLIKELY( scratch_top > (ulong)scratch + scratch_footprint( tile ) ) )
2242 0 : FD_LOG_ERR(( "scratch overflow %lu %lu %lu", scratch_top - (ulong)scratch - scratch_footprint( tile ), scratch_top, (ulong)scratch + scratch_footprint( tile ) ));
2243 0 : }
2244 :
2245 : /* One tick, one microblock, one plugin slot end, one plugin slot start,
2246 : and one leader update. */
2247 0 : #define STEM_BURST (5UL)
2248 :
2249 : /* See explanation in fd_pack */
2250 0 : #define STEM_LAZY (128L*3000L)
2251 :
2252 0 : #define STEM_CALLBACK_CONTEXT_TYPE fd_poh_ctx_t
2253 0 : #define STEM_CALLBACK_CONTEXT_ALIGN alignof(fd_poh_ctx_t)
2254 :
2255 0 : #define STEM_CALLBACK_DURING_HOUSEKEEPING during_housekeeping
2256 0 : #define STEM_CALLBACK_METRICS_WRITE metrics_write
2257 0 : #define STEM_CALLBACK_AFTER_CREDIT after_credit
2258 0 : #define STEM_CALLBACK_BEFORE_FRAG before_frag
2259 0 : #define STEM_CALLBACK_DURING_FRAG during_frag
2260 0 : #define STEM_CALLBACK_AFTER_FRAG after_frag
2261 :
2262 : #include "../../../../disco/stem/fd_stem.c"
2263 :
2264 : fd_topo_run_tile_t fd_tile_poh = {
2265 : .name = "poh",
2266 : .populate_allowed_seccomp = NULL,
2267 : .populate_allowed_fds = NULL,
2268 : .scratch_align = scratch_align,
2269 : .scratch_footprint = scratch_footprint,
2270 : .privileged_init = privileged_init,
2271 : .unprivileged_init = unprivileged_init,
2272 : .run = stem_run,
2273 : };
|