IBD: do not check PoW (Yespower) during Download headers

[decryp2kanon]

Background

#119

TL;DR

• CURRENT:
  • when downloading headers, check header (PoW: Yespower)
  • bottle neck at downloading headers
• PR:
  • skip check header (PoW: Yespower), only in downloading headers
  • bottle neck gone! 🎉

Safe to skip?

CheckProofOfWork() is already multiple checking in several place. i believe that even if our node is attacked, it will defense well.

---

RESULT

• ✔️ total IBD time ++23% faster
  

• ✔️ data traffic --60% reduced

  • before
    

  • after
    

• ✔️ less CPU usage around --5% when IBD

---

What we are skipping?

skipped check CheckProofOfWork() at pow.cpp when Downloading headers

bool CheckProofOfWork(uint256 hash, unsigned int nBits, const Consensus::Params& params)
{
    bool fNegative;
    bool fOverflow;
    arith_uint256 bnTarget;

    bnTarget.SetCompact(nBits, &fNegative, &fOverflow);

    // Check range
    if (fNegative || bnTarget == 0 || fOverflow || bnTarget > UintToArith256(params.powLimit))
        return false;

    // Check proof of work matches claimed amount
    if (UintToArith256(hash) > bnTarget)
        return false;

    return true;
}

How to skip and when?

do not PoW check when Downloading headers of IBD (include fImporting and fReindex). if not, we do not skip.

if ( IsInitialBlockDownload() == false )

Safe enough?

i believe so. its already multiple checked in CheckBlock() and this can be found in 4 places:

CLICK ME

• 1. AcceptBlock()

    if (!CheckBlock(block, state, chainparams.GetConsensus()) ||
        !ContextualCheckBlock(block, state, chainparams.GetConsensus(), pindex->pprev)) {
        if (state.IsInvalid() && !state.CorruptionPossible()) {
            pindex->nStatus |= BLOCK_FAILED_VALID;
            setDirtyBlockIndex.insert(pindex);
        }
        return error("%s: %s", __func__, FormatStateMessage(state));
    }

• 2. TestBlockValidity()

    // NOTE: CheckBlockHeader is called by CheckBlock
    if (!ContextualCheckBlockHeader(block, state, chainparams, pindexPrev, GetAdjustedTime()))
        return error("%s: Consensus::ContextualCheckBlockHeader: %s", __func__, FormatStateMessage(state));
    if (!CheckBlock(block, state, chainparams.GetConsensus(), fCheckPOW, fCheckMerkleRoot))
        return error("%s: Consensus::CheckBlock: %s", __func__, FormatStateMessage(state));

• 3. VerifyDB()

        // check level 1: verify block validity
        if (nCheckLevel >= 1 && !CheckBlock(block, state, chainparams.GetConsensus()))
            return error("%s: *** found bad block at %d, hash=%s (%s)\n", __func__,
                         pindex->nHeight, pindex->GetBlockHash().ToString(), FormatStateMessage(state));

• 4. ConnectBlock()

    // Check it again in case a previous version let a bad block in
    // NOTE: We don't currently (re-)invoke ContextualCheckBlock() or
    // ContextualCheckBlockHeader() here. This means that if we add a new
    // consensus rule that is enforced in one of those two functions, then we
    // may have let in a block that violates the rule prior to updating the
    // software, and we would NOT be enforcing the rule here. Fully solving
    // upgrade from one software version to the next after a consensus rule
    // change is potentially tricky and issue-specific (see RewindBlockIndex()
    // for one general approach that was used for BIP 141 deployment).
    // Also, currently the rule against blocks more than 2 hours in the future
    // is enforced in ContextualCheckBlockHeader(); we wouldn't want to
    // re-enforce that rule here (at least until we make it impossible for
    // GetAdjustedTime() to go backward).
    if (!CheckBlock(block, state, chainparams.GetConsensus(), !fJustCheck, !fJustCheck))
        return error("%s: Consensus::CheckBlock: %s", __func__, FormatStateMessage(state));

Screenshot during IBD

• 1. This means PoW check during IBD is not actually skipped, but still its checking.
     

[decryp2kanon]

when do PoW check?

1. IBD to first user ✔️
2. reindex-chainstate (fReindex) ✔️
3. reindex (fReindex) ✔️
4. fImporting ❓

[akshaynexus]

Looks like a good change to me,i had alot of delay and cpu usage when syncing

[decryp2kanon]

Tested ACK 032184c

[decryp2kanon]

IBD around 10~30% faster, however it does not skip checking PoW, but only remove duplicated at downloading stage

[decryp2kanon]

reference about

if (!x() && !y())

https://stackoverflow.com/questions/8300519/c-if-statement-with-which-function-will-execute-first

[volbil]

Looking good to me 👍