master_sync_reply is racy #266
Assignees
Labels
Comments
|
Yep, I hit the same bug stress testing. |
rustyrussell
added a commit
to rustyrussell/lightning
that referenced
this issue
Sep 8, 2017
We hit: assert(!peer->handle_master_reply); #4 0x000055bba3b030a0 in master_sync_reply (peer=0x55bba41c0030, msg=0x55bba41c6a80 "", replytype=WIRE_CHANNEL_GOT_COMMITSIG_REPLY, handle=0x55bba3b041cf <handle_reply_wake_peer>) at channeld/channel.c:518 #5 0x000055bba3b049bc in handle_peer_commit_sig (conn=0x55bba41c10d0, peer=0x55bba41c0030, msg=0x55bba41c6a80 "") at channeld/channel.c:959 #6 0x000055bba3b05c69 in peer_in (conn=0x55bba41c10d0, peer=0x55bba41c0030, msg=0x55bba41c67c0 "") at channeld/channel.c:1339 #7 0x000055bba3b123eb in peer_decrypt_body (conn=0x55bba41c10d0, pcs=0x55bba41c0030) at common/cryptomsg.c:155 #8 0x000055bba3b2c63b in next_plan (conn=0x55bba41c10d0, plan=0x55bba41c1100) at ccan/ccan/io/io.c:59 We got a commit_sig from the peer while waiting for the master to reply to acknowledge the commitsig we want to send (handle_sending_commitsig_reply). The fix is to go always talk to the master synchronous, and not try to process anything but messages from the master daemon. This avoids the whole class of problems. There's a fairly simple way to do this, as ccan/io lets you override its poll call: we process any outstanding master requests there, or add the master fd to the pollfds array. Fixes: ElementsProject#266 Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
master_sync_replyinchanneldassumes that there is only one such call pending at any point in time. This may not always be true, since we have 3 places in which this is called in response to timers and/or incoming messages from the peer.io_waitis used on the peer messages, so concurrent peer triggered calls should be safe, but if we have a commit timer running, followed by a call triggered by a peer message we'll run into the following error:This'll kill
channeld, requiring a reconnect.The text was updated successfully, but these errors were encountered: