4d397f5ebf
fix minor robustness issue in DHT bootstrap logic
2017-03-21 12:29:08 -04:00
c00a25a645
back-port aldenml's fix to not save replacement nodes twice to 1.1 ( #1661 )
...
back-port aldenml's fix to not save replacement nodes twice to 1.1
2017-02-04 19:58:32 -05:00
4df1ecbe25
re-enable some warnings and fix code
2016-11-20 01:05:16 -05:00
25195a7260
back port DHT fixes from master ( #1209 )
...
back port DHT fixes from master
2016-10-14 23:47:59 -04:00
6bb2d7c017
Sanitize 1.1 ( #1197 )
...
enable address sanitizer
2016-10-09 17:00:20 -04:00
2e0140b6f6
fix peer picking algorithm ( #1141 )
...
fix peer picking algorithm. The old code was always picking the first to_pick peers from the set.
2016-09-24 06:08:52 -07:00
b554909486
hint DHT bootstrap nodes of actual bootstrap request
2016-08-30 21:15:35 -04:00
3a665a1040
fix inconsistant prefix calulcation ( #1030 )
...
The special case for calculating the prefix in the last bucket was being
applied to the condidate node but not the existing nodes in the bucket.
2016-08-25 19:26:48 -04:00
3321ca10f1
update our id in the rpc manager when it changes ( #942 )
...
update our id in the rpc manager when it changes
2016-07-24 21:30:36 -04:00
8f3bd1052e
fix mutable dht_get_item failure when salt is non-empty ( #909 )
2016-07-18 07:12:10 -07:00
5ecd00c5c5
fix osx build ( #804 )
...
fix osx build with asserts enabled
2016-06-11 00:48:13 -04:00
98eabc63e9
using _WIN32 appears safer than WIN32
2016-05-23 23:02:52 -04:00
53bd03558c
Dht stats 1.1 ( #720 )
...
fix missing dht stats counter updates
2016-05-12 21:27:11 -04:00
4998bfedc6
Documentation typos
2016-03-04 10:43:25 -05:00
1f9f588e75
merge copyright year update and changelog from RC_1_0
2016-01-17 18:57:46 -05:00
f4d643bd2d
instead of restarting the whole DHT when changing external IP, just rebuild the routing table and change the node IDs
2016-01-11 21:17:25 -05:00
0f442f59f3
Avoid pass IPv6 address to current implementation of DHT while adding a node
2016-01-09 18:46:19 -05:00
e06daa68a4
overhaul the dht simulator (setup_dht) to be more flexible and cheaper to run thousands of nodes
2016-01-09 01:08:52 -05:00
d954ae1b18
simplify dht routing table add
2016-01-02 20:03:18 -05:00
07ddb010c5
split buckets when exceeding the next bucket's size, to make sure we split before risking discarding nodes because the next bucket is smaller
2016-01-02 00:45:44 -05:00
e3ba811ae4
improve name of m_timer in dht_tracker
2016-01-01 16:42:44 -05:00
01e6b93854
simplify and improve unit test for distance_exp. make some immutable variables const in the DHT implementation. instead of waking up periodically just to check if it's time to refresh the DHT secret key, set the timer to only wake up to refresh the key. If we don't have a DHT observer (to ask for our external IP) or if we don't know our external IP, don't generate a node ID based on 0.0.0.0, just generate a random ID instead. Simplified and improved node replacement logic in the routing table a little bit
2016-01-01 16:42:37 -05:00
ffb78f0378
Merge pull request #312 from ssiloti/export-verify-message
...
export verify_message and incoming_error
2015-12-30 22:51:38 -05:00
ac01ccc43a
add templated wrapper around verify_message
...
It was too easy to forget to update the ret and size parameters
when the size of desec changed. With this change we now automatically
get the size of desc and enforce that ret has the same size.
It would be nice if we could use bdecode_node ret[static Size] from C99
but that is not widely supported.
2015-12-29 20:35:36 -08:00
626baade75
fix dht_storage bug
2015-12-19 22:06:25 -05:00
28749466a2
fix build
2015-12-19 02:40:10 -05:00
15ec839d33
extend the unit test of dht_storage
2015-12-19 02:09:06 -05:00
c6f1e0bd7b
export verify_message and incoming_error
...
Also move these functions to a new msg.cpp. They are useful for plugins
which implement DHT extensions.
2015-12-05 18:20:08 -08:00
8732863a9a
another fix to receiving malformed DHT error message
2015-12-03 01:08:27 -05:00
7540a6e5cc
correctly check incoming DHT error messages
2015-11-30 18:02:00 -05:00
332799bff2
make alert types declared final and disable the log alert declarations when logging is disabled
2015-11-29 10:16:02 -05:00
603ef84c65
Fix observer_size.
2015-11-23 11:06:20 -05:00
9c7edf803e
add override and a TODO
2015-11-22 18:58:32 -08:00
9f7aa7f3a6
fix segfault in put_data
...
If less than three nodes are found to put an item to then traversal_algorithm::start
will add router nodes. This leads to a crash in put_data::invoke when it tries to
read a token from uninitialized memory in a null_observer.
2015-11-22 16:05:24 -08:00
1f13343a70
const correctness
2015-11-22 10:02:26 -08:00
8a984effa7
allow done() with a non-zero invoke count
2015-11-22 10:02:26 -08:00
6c5bed63d3
whitespace cleanup
2015-11-22 10:02:26 -08:00
ab70ad3e05
fix comments and unit test
2015-11-20 00:06:08 -05:00
194ad410dc
Make dht_put_alert more accurate.
2015-11-20 00:06:08 -05:00
86116caf4c
fix assert after a DHT get/put request
...
If an observer is abandoned due to there being more than 100 results in a
traversal then the traversal may complete while the observer is still
outstanding. When the observer times out or is aborted it calls
traversal_algorithm::failed() which asserts because m_invoke_count has been
cleared to zero by traversal_algorithm::done().
To fix this add a check when abandoning observers to see if they have an
outstanding query. If they do then flag them as done to prevent them from
calling back into the traversal and decrement the invoke count.
Fixes #271
2015-11-16 21:05:32 -08:00
ba1ccdec66
fix assert in dht::observer
...
This is was introduced by c2277b3ea5
2015-11-14 18:30:42 -08:00
c2277b3ea5
fix traversal_algorithm::done() being invoked more than once
...
A traversal can be done while there are still outstanding queries (i.e.
m_invoke_count is non-zero) if K good responses have already been received and
none of the outstanding queries are closer than the nodes which have responded.
When this happens and the outstanding queries eventually complete or timeout
they call traversal_algorithm::failed() or traversal_algorithm::finished() as
usual which will cause traversal_algorithm::done() to be called yet again.
The fix is to always set the done flag on all queried observers in
traversal_algorithm::done() so that the observers will refrain from calling
back into the traversal.
This also makes traversal_algorithm::abort() redundant since this was the only
thing it did before it called into done().
2015-11-09 20:47:33 -08:00
3cc464805f
update libsimulator submodule
2015-10-25 10:55:20 -04:00
a75dfe2eb6
silence warnings in boost headers in dht_storage
2015-10-25 09:30:56 -04:00
4151c8e366
assert that the signature is valid when assigning a mutable value to an item
2015-10-21 20:29:29 -07:00
62b24d8b67
Add set_dht_storage to session API.
2015-10-02 09:00:07 -04:00
d7cdcf9369
Merge pull request #183 from thomas-yuan/ta_init
...
start 'find' from near nodes instead of all of nodes in routing table.
2015-09-30 00:43:14 -04:00
197a443465
Fix of dht_storage_counters::peers counter.
2015-09-24 10:50:04 -04:00
0d2644802d
start 'find' from near nodes instead of all of nodes in routing table.
2015-09-23 17:04:33 -04:00
f9fa8ffdec
Use ping instead of get_peers if current bucket is full.
2015-09-22 10:35:10 -04:00
arvidn