Caught internal_error: HandshakeManager::receive_succeeded(...) Unread data won't fit PCB's read buffer.

[slingamn]

Congratulations on the new release!

I tried applying the patchset from #134 to master (at the time of writing, eddcb1f). (Without the patchset, my rtorrent instance is basically unresponsive due to the high number of synchronous DNS lookups). rtorrent then crashes immediately with:

Caught internal_error: HandshakeManager::receive_succeeded(...) Unread data won't fit PCB's read buffer.
/home/shivaram/workspace/sandbox/lib/libtorrent.so.22(_ZN7torrent14internal_error10initializeERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE+0x74) [0x74b6be624484]
/home/shivaram/workspace/sandbox/bin/rtorrent(_ZN7torrent14internal_errorC1EPKc+0x88) [0x593e64ff8af8]
/home/shivaram/workspace/sandbox/lib/libtorrent.so.22(+0x460b4) [0x74b6be5eb0b4]
/home/shivaram/workspace/sandbox/lib/libtorrent.so.22(+0x45988) [0x74b6be5ea988]
/home/shivaram/workspace/sandbox/lib/libtorrent.so.22(_ZN7torrent9PollEPoll7performEv+0xe9) [0x74b6be62c5f9]
/home/shivaram/workspace/sandbox/lib/libtorrent.so.22(_ZN7torrent11thread_base10event_loopEPS0_+0x12a) [0x74b6be62014a]
/home/shivaram/workspace/sandbox/bin/rtorrent(+0x47c60) [0x593e64ff0c60]
/lib/x86_64-linux-gnu/libc.so.6(+0x2a1ca) [0x74b6bde2a1ca]
/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0x8b) [0x74b6bde2a28b]
/home/shivaram/workspace/sandbox/bin/rtorrent(+0x485d5) [0x593e64ff15d5]

I bisected this to 11c38cc / #254. So the release tag v0.14.0 and the subsequent commit a83dcb9 don't have this problem. But it's not clear to me what in #254 would cause this.

Thanks for your time.

[rakshasa]

There's logging instructions in #254, use those and send me the results.

Note that you need to also change DISABLED__USE_EXTRA_DEBUG to USE_EXTRA_DEBUG in the above commit.

[slingamn]

I was able to reproduce the issue quickly against bbeeab8 and b30a169 without my patchset. So it seems like this is an unresolved issue in master.

I'm trying to get logs for it with ./configure --with-extra-debug but I'm having trouble actually finding the logs. Are they supposed to display in the log window displayed when the l key is pressed? If so I think rtorrent may be crashing before the logs can actually be displayed.

[rakshasa]

In src/protocol/handshake.cc change #if DISABLED__USE_EXTRA_DEBUG to #if USE_EXTRA_DEBUG.

log.open_file = "connection", /run/self/logs/connection.log
log.add_output = "connection", "connection"

[slingamn]

0000000000 handshake_manager->11.22.33.44: created outgoing connection: fd:83 encryption:31 message:2
0000000000 handshake->11.22.33.44:65432: read_info
0000000000 handshake->11.22.33.44:65432: fill_read_buffer : size:68 remaining:0 reserved_left:1254
0000000000 handshake->11.22.33.44:65432: read_peer
0000000000 handshake->11.22.33.44:65432: fill_read_buffer : size:20 remaining:20 reserved_left:1186
0000000000 handshake->11.22.33.44:65432: event_read : READ_MESSAGE
0000000000 handshake->11.22.33.44:65432: fill_read_buffer : size:5 remaining:0 reserved_left:1186
0000000000 handshake->11.22.33.44:65432: read_port
0000000000 handshake->11.22.33.44:65432: fill_read_buffer : size:7 remaining:5 reserved_left:1181
0000000000 handshake->11.22.33.44:65432: event_read : READ_MESSAGE
0000000000 handshake->11.22.33.44:65432: fill_read_buffer : size:5 remaining:65534 reserved_left:1179
0000000000 handshake_manager->11.22.33.44: handshake success: type:leech id:-[REDACTED]

followed by

Caught internal_error: HandshakeManager::receive_succeeded(...) Unread data won't fit PCB's read buffer.

[slingamn]

I think I diagnosed this (I'm still testing): #264

[jmakovicka]

I think I diagnosed this (I'm still testing): #264

I had the same issue, fixed it exactly the same way.