Commit graph

655 commits

Author SHA1 Message Date
Roman Khimov
773ccc2b92 network: allow discoverer to reuse addresses
...and don't try to connect to the nodes we're already connected to.

Before this change we had a problem of discoverer throwing away good valid
addresses just because they are already known which lead to pool draining over
time (as address reuse was basically forbidden and getaddr may not get enough
new nodes).
2019-09-16 16:32:04 +03:00
Roman Khimov
46dc141c6c network: enlarge address request queue
Queuing one message is not reliable enough, the peer that gets it can fail to
actually make a request, so make this queue a bit deeper to have a higher
chance of success.
2019-09-16 16:32:04 +03:00
Roman Khimov
d3bb8ddf8f network: handle errors and connection close more correctly
This makes writer side handle errors properly and fixes communication between
reader and writer goroutine to always correctly unregister the peer. This is
especially important for the case where error occurs before handshake
completes as in this case we don't even have goroutine in startProtocol()
running.
2019-09-16 16:32:04 +03:00
Roman Khimov
76c7cff67f network: make node strictly follow handshake procedure
Don't accept other messages before handshake is completed, check handshake
message sequence.
2019-09-16 16:32:04 +03:00
Roman Khimov
c6487423ae network: close connection on disconnect
If it's already closed, this won't hurt, but in the case of logical error it
saves us from leaking this connection (and potentially, peer).
2019-09-16 16:26:30 +03:00
Roman Khimov
90e13debe4 network: implement getaddr sends and addr receival
Request new addresses if we're short on known addresses pool size.
2019-09-16 16:26:30 +03:00
Roman Khimov
2a49e68d77 network: start worker goroutine for every connection attempts
Prevents blocking on write to workCh which can be dangerous for the server.
2019-09-16 16:26:30 +03:00
Roman Khimov
b4e284f301 discovery: make pool management more reliable
Just drop excessive addresses, otherwise we can block for no good reason.
2019-09-16 16:26:30 +03:00
Roman Khimov
85f19936dd network: implement connection retries
It's worth to try a bit more than once.
2019-09-16 16:26:30 +03:00
Roman Khimov
be6c905e5d network: use and improve discovery mechanism for reconnections
This makes our node reconnect to other nodes if connection drops for some
reason. Fixes #390.
2019-09-16 16:26:30 +03:00
Roman Khimov
b263c97bbb core: add some mechanism to drop bad blocks from cache
If we can't persist anything from our cache there is something wrong with
it. Try to remediate some possible problems.
2019-09-16 16:26:30 +03:00
Evgenii Stratonikov
d2ccc3b017 vm: make REVERSE consume an item from stack 2019-09-16 15:50:51 +03:00
Evgenii Stratonikov
fd264cb06f vm: make REVERSE work with structs
Also add more tests.
Extend `Array` method of `Element` to work with `StructItem`.
2019-09-16 15:50:51 +03:00
Evgenii Stratonikov
f0426ac2d5 vm: do nothing if SHL/SHR by 0 2019-09-16 15:50:51 +03:00
Evgenii Stratonikov
9eca5ee317 vm: convert SIZE argument to bytearray 2019-09-16 15:50:51 +03:00
Evgenii Stratonikov
9780889239 vm: make NEWARRAY/NEWSTRUCT accept bytearray 2019-09-16 15:50:51 +03:00
Evgenii Stratonikov
09e197eaf3 vm: make PUSH0 create empty bytearray
Also make tests expect []byte{} instead of 0.
2019-09-16 15:50:51 +03:00
Evgenii Stratonikov
66501f9ef9 vm: make REMOVE consume array from stack 2019-09-16 15:50:14 +03:00
Evgenii Stratonikov
71cfd14b92 vm: create an array of false items in NEWARRAY/NEWSTRUCT 2019-09-16 15:24:58 +03:00
Evgenii Stratonikov
bafdb916a0 vm: make PICKITEM accept bytearrays 2019-09-16 15:24:58 +03:00
Evgenii Stratonikov
1881adabb9 vm: fail if NIP has not enough arguments 2019-09-16 15:24:58 +03:00
Evgenii Stratonikov
6e4014547d vm: fail if DROP has no argument 2019-09-16 15:24:58 +03:00
Evgenii Stratonikov
729b7a0b24 vm: fail if ROT has not enough arguments 2019-09-16 15:24:58 +03:00
Evgenii Stratonikov
cf7fbb80ba vm: fail if EQUAL has not enough arguments 2019-09-16 15:24:58 +03:00
Evgenii Stratonikov
d257044298 vm: fail if XTUCK argument is 0 2019-09-16 15:24:58 +03:00
Roman Khimov
b21a220712
Merge pull request #381 from nspcc-dev/add_boltd
storage: add boltdb support, closes #335.
2019-09-14 13:00:52 +03:00
Vsevolod Brekelov
2cabd5b274 storage: add boltDB unit tests 2019-09-14 10:29:08 +03:00
Vsevolod Brekelov
bfe5fd538e storage: add boltDB support
closes #335
2019-09-14 10:28:43 +03:00
Roman Khimov
cd3d22d150
Merge pull request #395 from nspcc-dev/fix/append-codegen
VM: fix codegen for APPEND

After changes in #391 some tests started to fail. This PR fixes codegen for APPEND.
2019-09-13 18:32:55 +03:00
Evgenii Stratonikov
56ec097d76 vm: fix codegen for APPEND 2019-09-13 11:24:21 +03:00
Roman Khimov
521a707e33 core: don't print contracts in persistBlock()
Yeah, this really needs more love, but for now just don't spit 'em out on the
screen, please.
2019-09-12 16:12:12 +03:00
Roman Khimov
d0ffd165fd
Merge pull request #393 from nspcc-dev/assorted-db-handling-fixes-and-optimizations
Assorted db handling fixes and optimizations based on the testnet
connection experience.
2019-09-11 22:31:19 +03:00
Roman Khimov
4395cea344 storage: fix reading overlapping hash blocks from the DB
In the unlikely event of overlapping hash block written to the DB we might end
up with wrong hash list. That happened to me for some reason when synching
with the testnet leading to the following keys with respective values:
  150000 -> 2000 hashes
  152000 -> 2000 hashes
  153999 -> 2000 hashes

Reading it hashes number 153999 and 154000 got the same values and the chain
couldn't sync correctly.
2019-09-11 20:33:41 +03:00
Roman Khimov
4f60fd3e8e core: optimize persist() for the case when no next block found
If there is no next block in the cache, looping through the whole cache won't
help.
2019-09-11 20:32:30 +03:00
Roman Khimov
7812fb28ea core: optimize blockchain init with existing DB
There is no need to generate genesis block (it's already in the DB) and we can
add all hashes at once without iterating over them.
2019-09-11 20:28:49 +03:00
Roman Khimov
6855e4e07a
Merge pull request #392 from nspcc-dev/fix/bug-append
VM: APPEND should push no value on stack. Fixes #391.
2019-09-11 19:02:05 +03:00
Evgenii Stratonikov
68c6c93980 vm: do not allow APPEND to operate on bytearray 2019-09-11 17:17:37 +03:00
Evgenii Stratonikov
723dcc6e25 vm: make APPEND push no value on stack 2019-09-11 17:17:34 +03:00
Roman Khimov
6f7e565b45
Merge pull request #389 from nspcc-dev/update-config-seeds
config: update mainnet and testnet seed lists
2019-09-11 15:15:06 +03:00
Roman Khimov
9281348846 config: update mainnet and testnet seed lists
Sync with neo-cli (master-2.x) branch.
2019-09-11 15:10:02 +03:00
Roman Khimov
794d375bfd
Merge pull request #388 from nspcc-dev/substr-left-indexes
VM: harden substr/left/right against bad indexes and improve
PUSHDATA along the way. Fixes #387.
2019-09-11 14:52:48 +03:00
Roman Khimov
8311bda355 vm: harden LEFT and RIGHT implementations against negative indexes 2019-09-11 14:51:04 +03:00
Roman Khimov
17f3a21e27 vm: harden SUBSTR implementation against bad index/offset values
The upper index bound for slices is capacity, not length. Check for negative
values also. Fixes #387.
2019-09-11 14:37:41 +03:00
Roman Khimov
fc1075bf75 vm: protect PUSHDATA from short reads
Same thing done in a2a8981979 for PUSHBYTES,
failing to read the amount of bytes specified should lead to FAULT. Also
makes readUint16() and readUint32() panic as this is the behavior we want in
these cases. Add some tests along the way.
2019-09-11 12:35:37 +03:00
Roman Khimov
14c02cb3b1
Merge pull request #385 from nspcc-dev/vm-contract-invoke-parameters-332-part1
VM CLI improvements, part 1 of #332 work.
2019-09-10 23:55:23 +03:00
Roman Khimov
9193ec45de
Merge pull request #383 from nspcc-dev/fix_blockchain_test
unitTests: fix for failing test
2019-09-10 23:53:50 +03:00
Roman Khimov
392a4c4f39 vm/cli: add some convenience heuristic to ease 'run' use 2019-09-10 23:31:06 +03:00
Roman Khimov
c5911c2f10 vm/cli: add support for bool type as 'run' parameter 2019-09-10 23:31:06 +03:00
Roman Khimov
982bdcd704 vm/cli: make operation mandatory if anything added to run
Matches neo-cli behavior.
2019-09-10 23:31:06 +03:00
Roman Khimov
5bd666b786 vm: return NOP as current instruction for just loaded VM
Before:
NEO-GO-VM > loadgo h.go
READY: loaded 16 instructions
NEO-GO-VM > ip
instruction pointer at -1 (PUSH0)

After:
NEO-GO-VM > loadgo h.go
READY: loaded 16 instructions
NEO-GO-VM > ip
instruction pointer at -1 (NOP)

I think NOP is a little less scary.
2019-09-10 23:31:06 +03:00