IBD slowed down by failing to deal with a bad checksum block message #18755

issue rebroad opened this issue on April 24, 2020

rebroad commented at 10:59 AM on April 24, 2020: contributor

2020-04-24 10:05:11.299 Added outbound connection 47.184.48.254:8333 peer=58
2020-04-24 10:05:11.300 send version 70015, blocks=438212, relay=0, services=0x9, us=[::]:0, them=47.184.48.254:8333, peer=58
2020-04-24 10:05:11.515 recv version: /Satoshi:0.18.1/: version 0, blocks=-1, relay=0, services=0x40d us=203.126.139.7:30374, them=47.184.48.254:8333, peer=58
2020-04-24 10:05:11.752 send getdata block(1/3,0/0) 00000000000000000370b870cb38da4165f5f3c9214e3dfd3745a5b49a7c19a2 (438215) age=1260.9d behind=189203 peer=58
2020-04-24 10:05:11.752 send getdata block(2/4,0/0) 000000000000000002a8fda1da1355095093882bd434281f5a866590684512e2 (438218) age=1260.9d behind=189200 peer=58
2020-04-24 10:05:39.848 ProcessMessages(block, 998209 bytes): CHECKSUM ERROR expected=97d355bb is=97d355bb peer=58
2020-04-24 10:06:20.123 recv block(2/10,0/0) 000000000000000002a8fda1da1355095093882bd434281f5a866590684512e2 (438218) age=1260.9d behind=189200 size=855918 peer=58
2020-04-24 10:11:07.799 Stall started peer=58
2020-04-24 10:11:11.557 block download stalled. LastRecv=0s LastRecvBlk=0s LastSend=17s disconnect peer=58

It takes 6 minutes to detect a stall and request the block from another peer (and continues to download subsequent blocks from this peer). This is on v0.14.2 so may be fixed in later versions. (I didn't see any issues raised regarding this though).

rebroad added the label Bug on Apr 24, 2020
fanquake commented at 11:02 AM on April 24, 2020: member

Please don’t open issues for versions that are well past there EOL. Test with a newer release.
maflcko removed the label Bug on Apr 24, 2020
maflcko added the label Brainstorming on Apr 24, 2020
maflcko added the label P2P on Apr 24, 2020
maflcko commented at 11:14 AM on April 24, 2020: member

CHECKSUM ERROR expected=97d355bb is=97d355bb

The checksum matches. I guess you are running a modified version of Bitcoin Core?
rebroad closed this on Apr 24, 2020
rebroad commented at 11:43 AM on April 24, 2020: contributor

@MarcoFalke yes, and thank you, I'd not spotted that they matched. Closing as clearly I broke it (again).
rebroad commented at 11:48 AM on April 24, 2020: contributor

the only thing I broke is the reporting of the actual checksum received, so the checksum did differ, and it did slow down the IBD. Re-opening as the fix (IMHO) is to disconnect the node after the failed checksum (during IBD at least), which isn't being done as of the latest master.
rebroad reopened this on Apr 24, 2020
rebroad referenced this in commit 2e7add838d on Apr 25, 2020
rebroad referenced this in commit 119bc784ea on Apr 25, 2020
rebroad referenced this in commit d2c9d85ba4 on Apr 25, 2020
rebroad referenced this in commit e2d37e55dd on Apr 25, 2020
rebroad referenced this in commit 17a1ff0462 on Apr 25, 2020
rebroad commented at 6:33 AM on April 25, 2020: contributor

#15206 would also fix this
pinheadmz commented at 4:14 PM on March 7, 2023: member

I think this was closed by #19107. @rebroad are you able to still reproduce this issue with the latest release of Bitcoin Core?
maflcko commented at 4:45 PM on March 7, 2023: member

Can be re-opened if this is confirmed to be happening on a recent vanilla release of Bitcoin Core. Bonus points will be assigned for steps to reproduce.
maflcko closed this on Mar 7, 2023
bitcoin locked this on Mar 6, 2024

Contributors

Labels

Linked (view graph)

#18760 Disconnect nodes with bad messages during IBD