Re: [Libguestfs] [libnbd PATCH v2 4/5] states: Allow in-flight read while writing next command

As already noted in our state machine, a client that batches up a large read followed by large writes, coupled with a server that only processes commands in order, can result in deadlock (the server won't read more until we unblock its ability to write out its reply to our first command; but we aren't willing to read until we are done writing out our second command). Break the deadlock by teaching the generator that while we are in the middle of writing a command, we must remain responsive to read_notify events; if the server has data for us to read, we should consume that before jumping back into the middle of our command issue (and consuming a reply can invalidate sbuf, so we have to drop an assertion in PREPARE_WRITE_PAYLOAD).

--- generator/generator | 26 ++++++++++++++++++-------- generator/states-issue-command.c | 25 ++++++++++++++++++++++++- lib/internal.h | 1 + 3 files changed, 43 insertions(+), 9 deletions(-)

@@ -43,12 +52,18 @@ } return 0; + ISSUE_COMMAND.PAUSE_SEND_REQUEST: + assert (conn->wlen); + assert (conn->cmds_to_issue != NULL); + conn->in_write_payload = false; + SET_NEXT_STATE (%^REPLY.START); + return 0;

+ ISSUE_COMMAND.PREPARE_WRITE_PAYLOAD: struct command_in_flight *cmd; assert (conn->cmds_to_issue != NULL); cmd = conn->cmds_to_issue; - assert (cmd->handle == be64toh (conn->sbuf.request.handle));