- Put garlicked DSM messages in the InNetMessagePool instead of handling directly,
so the IterativeSearchJob will see the response and declare success.
Only affected LS lookups as we do not request encrypted replies for RI lookups.
Similar to the fix for verifies 6dc5bed94321ae2b290cfe351511d18465e08f91
This bug was causing initial lookups to fail but subsequent ones to succeed.
Broken since 0.9.7 when encrypted replies was introduced.
Ticket #1125
- Fix several 'increment of volatile is not atomic' all over
Remaining: UDP PeerState.java, to be checked in separately
- Comment out all of unused MessageStateMonitor
- Reject type 3 (undefined) delivery instructions at both OBEP and IBEP,
would be treated as LOCAL
- Reject type 0 (LOCAL) delivery instructions at OBEP, no current usage,
would cause NPE if handled
- Increase threshold for idle timeout reduction from 33% to 60%.
This was reduced from 75% to 33% in -10 and it caused significant
tunnel build problems, for reasons that aren't entirely clear.
- Don't change the SSU idle timeout quicker when firewalled.
- log tweaks
- Build a new exploratory fallback tunnel in the BuildExecutor
loop if we run out.
- Don't use closest expl. tunnel as the paired tunnel for a build,
use a random one instead (partially back out change from -12)
- Log tweaks
- Don't discard an encrypted DSRM received
down a tunnel, just strip the hashes like we do for unencrypted
- Send a store of our own encrypted LS received down a tunnel to
the InNetMessagePool so the FloodfillVerifyStoreJob will see it.
* NetDB: Fix LS store verifies with encrypted replies
by storing the tagset with the correct SKM for the inbound tunnel used.
Broken since 0.9.7 when it was introduced.
- Don't use fallback expl. tunnels as the paired tunnel
for a client tunnel build.
This will cause more client tunnel breakage but it's
really bad to use a zero-hop for this.
- Fix selection of an expl. tunnel close to a hash
- Don't update build generation stat if the generation failed
- log tweaks
- Check blacklist/whitelist before connection limits, so
a blacklisted peer does not increment the counters
- Don't increment total throttle if peer is throttled
- Fix blacklist NPE after config change (ticket #1070)
