Ring Buffer Overflow needs Manual Flushing

[HaydosR]

I have an OSDP network of approximately 80 devices deployed, without logging from anything but the CP. At present I believe one of my devices has gone rogue and is talking atop other devices, contributing to overflowed receive buffers (ie, its sending random junk data atop other valid devices). Obviously this in itself is an issue, but my concern is that if it was disabled, I dont seem to see the Rx buffer properly recover itself.

E (1900603) LIBOSDP: [0] osdp_phy.c:73 - RX ring buffer overflow!                          
E (1900603) LIBOSDP: [3] osdp_phy.c:549 - Packet sequence mismatch (expected: 0, got: 1)                                   
W (1900603) LIBOSDP: [4] osdp_cp.c:752 - NAK(SEQ_NUM); restarting communication                                                 
E (1902823) LIBOSDP: [0] osdp_phy.c:73 - RX ring buffer overflow!                          
E (1902823) LIBOSDP: [3] osdp_phy.c:549 - Packet sequence mismatch (expected: 1, got: 0)                                   
W (1902823) LIBOSDP: [4] osdp_cp.c:752 - NAK(SEQ_NUM); restarting communication                                                 
E (1907053) LIBOSDP: [0] osdp_phy.c:73 - RX ring buffer overflow!                          
E (1907053) LIBOSDP: [3] osdp_phy.c:549 - Packet sequence mismatch (expected: 0, got: 1)                                   
W (1907053) LIBOSDP: [4] osdp_cp.c:752 - NAK(SEQ_NUM); restarting communication                                                 
E (1909093) LIBOSDP: [0] osdp_phy.c:73 - RX ring buffer overflow!                                                 
E (1911313) LIBOSDP: [0] osdp_phy.c:73 - RX ring buffer overflow!                                               
E (1911313) LIBOSDP: [3] osdp_phy.c:475 - Invalid crc 0x81f3/0x0045                                                                                                
I (1911483) CMDS: Board info: Vin=19.945V, Batt=23.97V, Charge State=Disconnected,

I suspect the fix could be as simple as clearing the ring buffer in the error path of osdp_phy_state_reset()

  if (is_error) {
      pd->phy_retry_count = 0;
      phy_reset_seq_number(pd);
      if (pd->channel.flush) {
          pd->channel.flush(pd->channel.data);  // Flushes HARDWARE buffer
      }
      pd->rx_rb.head = pd->rx_rb.tail = 0;      // Reset Ring Buffer
  }

Can anyone see any potential issues with this modification? I cannot replicate reliably on our testbench, and am too cautious to deploy to a prod system.

[sidcha]

Your idea is correct, but the placement could be improved. I created #261 to track this issue. Thanks for reporting it.