"Max retry while checking receiveddata..." in RMCP mode with Vadatech MCH

[j-marjanovic]

Hi,

when I try to use the mmcterm with Vadatech MCH UTC056 I get the following error:

$ mmcterm 10.2.21.173 0x7e -c 0
Traceback (most recent call last):
  File "/home/jan-marjanovic/miniconda3/envs/mmc-env/bin/mmcterm", line 8, in <module>
    sys.exit(main())
             ^^^^^^
  File "/home/jan-marjanovic/Atom/Software/mmcterm/mmcterm/mmcterm.py", line 536, in main
    if not conn.session_ctrl(args.channel, True, args.max_pkt_size):
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jan-marjanovic/Atom/Software/mmcterm/mmcterm/mmcterm.py", line 142, in session_ctrl
    status, _ = self.raw_cmd(
                ^^^^^^^^^^^^^
  File "/home/jan-marjanovic/Atom/Software/mmcterm/mmcterm/mmcterm.py", line 112, in raw_cmd
    raw_reply = self.conn.raw_command(0, 0x30, data)
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jan-marjanovic/miniconda3/envs/mmc-env/lib/python3.12/site-packages/pyipmi/__init__.py", line 225, in raw_command
    return self.interface.send_and_receive_raw(self.target, lun, netfn,
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jan-marjanovic/miniconda3/envs/mmc-env/lib/python3.12/site-packages/pyipmi/interfaces/rmcp.py", line 656, in send_and_receive_raw
    return self._send_and_receive(target=target,
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jan-marjanovic/miniconda3/envs/mmc-env/lib/python3.12/site-packages/pyipmi/interfaces/rmcp.py", line 632, in _send_and_receive
    raise RetryError("Max retry while checking received"
pyipmi.errors.RetryError: Max retry while checking receiveddata against request header for rmcphost 10.2.21.173
Exception ignored in: <function IpmiConn.__del__ at 0x7f5ceeb46d40>
Traceback (most recent call last):
  File "/home/jan-marjanovic/Atom/Software/mmcterm/mmcterm/mmcterm.py", line 54, in __del__
    self.conn.session.close()
  File "/home/jan-marjanovic/miniconda3/envs/mmc-env/lib/python3.12/site-packages/pyipmi/session.py", line 122, in close
    self.interface.close_session()
  File "/home/jan-marjanovic/miniconda3/envs/mmc-env/lib/python3.12/site-packages/pyipmi/interfaces/rmcp.py", line 563, in close_session
    rsp = self.send_and_receive(req)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jan-marjanovic/miniconda3/envs/mmc-env/lib/python3.12/site-packages/pyipmi/interfaces/rmcp.py", line 670, in send_and_receive
    rx_data = self._send_and_receive(target=req.target,
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jan-marjanovic/miniconda3/envs/mmc-env/lib/python3.12/site-packages/pyipmi/interfaces/rmcp.py", line 632, in _send_and_receive
    raise RetryError("Max retry while checking received"
pyipmi.errors.RetryError: Max retry while checking receiveddata against request header for rmcphost 10.2.21.173

it works fine (although extremely slow) if I use ipmitool option with the fix from #5 . Not surprisingly, IPMI tool also works with no issue:

$ ipmitool -H 10.2.21.173 -P "" -B 0 -b 7 -T 0x82 -t 0x7e raw 0x30 0xf0 0
 4d 4d 43 20 43 6f 6e 73 6f 6c 65
$ ipmitool -H 10.2.21.173 -P "" -B 0 -b 7 -T 0x82 -t 0x7e raw 0x30 0xf0 1
 52 46 53 6f 43

[patrislav1]

This looks a lot like either a Vadatech IPMI quirk or a pyipmi bug. I would really like to ensure mmcterm compatibility with Vadatech MCHs, but since I don't have any, my hands are tied here. 🤷

Your best bet, IMHO, is to sniff the RMCP packets of both the failing mmcterm and the successful ipmitool invocation (they should both do the same, in theory?) and open an issue in python-ipmi.

[jan-atom]

Don't (yet) have a solution, but have discovered some new details.

The difference seems to be that the NAT sends one response per a request:

while Vadatech sends two responses per a requests, with the second one actually containing the data:

However, by some "magic" (need to investigate this better) the ipmitool is able to somehow figure out that the first packet does not contain the right data and parse the data from the second packet, while the python-ipmi library does not do that.

Ad just to confirm, in the Vadatech case, on the I2C on the backplane everything looks normal, there is a request and ~900 us a response is transmitted.