When running a discover with IPv6 resources I get an exception

[gbregman]

I created some resources, like a listener, using IPv6 addresses. Using the discovery.py service, when running "nvme discover -t tcp -a 192.168.13.3 -s 8009" I got:

ceph-nvmeof-nvmeof-1  | ERROR:control.server:GatewayServer exception occurred:
ceph-nvmeof-nvmeof-1  | Traceback (most recent call last):
ceph-nvmeof-nvmeof-1  |   File "/src/control/discovery.py", line 977, in nvmeof_tcp_connection
ceph-nvmeof-nvmeof-1  |     err = handle_opcode[opcode](conn, data, cmd_id)
ceph-nvmeof-nvmeof-1  |   File "/src/control/discovery.py", line 739, in reply_get_log_page
ceph-nvmeof-nvmeof-1  |     adrfam = ADRFAM_TYPES[allow_listeners[log_entry_counter]["adrfam"]]
ceph-nvmeof-nvmeof-1  |   File "/usr/lib64/python3.9/enum.py", line 432, in __getitem__
ceph-nvmeof-nvmeof-1  |     return cls._member_map_[name]
ceph-nvmeof-nvmeof-1  | KeyError: 'IPV6'
ceph-nvmeof-nvmeof-1  |
ceph-nvmeof-nvmeof-1  | The above exception was the direct cause of the following exception:
ceph-nvmeof-nvmeof-1  |
ceph-nvmeof-nvmeof-1  | Traceback (most recent call last):
ceph-nvmeof-nvmeof-1  |   File "/src/control/__main__.py", line 36, in <module>
ceph-nvmeof-nvmeof-1  |     gateway.serve()
ceph-nvmeof-nvmeof-1  |   File "/src/control/server.py", line 113, in serve
ceph-nvmeof-nvmeof-1  |     self._start_discovery_service()
ceph-nvmeof-nvmeof-1  |   File "/src/control/server.py", line 151, in _start_discovery_service
ceph-nvmeof-nvmeof-1  |     DiscoveryService(self.config).start_service()
ceph-nvmeof-nvmeof-1  |   File "/src/control/discovery.py", line 1030, in start_service
ceph-nvmeof-nvmeof-1  |     callback(key.fileobj, mask)
ceph-nvmeof-nvmeof-1  |   File "/src/control/discovery.py", line 979, in nvmeof_tcp_connection
ceph-nvmeof-nvmeof-1  |     raise UnknownNVMEOpcode(opcode) from e
ceph-nvmeof-nvmeof-1  | control.discovery.DiscoveryService.nvmeof_tcp_connection.<locals>.UnknownNVMEOpcode: unsupported opcode: 2
ceph-nvmeof-nvmeof-1  | INFO:control.server:Aborting SPDK(7205e8967caf) pid 22...
ceph-nvmeof-nvmeof-1  | INFO:control.server:Exiting the gateway process.
ceph-nvmeof-nvmeof-1  | Traceback (most recent call last):
ceph-nvmeof-nvmeof-1  |   File "/src/control/discovery.py", line 977, in nvmeof_tcp_connection
ceph-nvmeof-nvmeof-1  |     err = handle_opcode[opcode](conn, data, cmd_id)
ceph-nvmeof-nvmeof-1  |   File "/src/control/discovery.py", line 739, in reply_get_log_page
ceph-nvmeof-nvmeof-1  |     adrfam = ADRFAM_TYPES[allow_listeners[log_entry_counter]["adrfam"]]
ceph-nvmeof-nvmeof-1  |   File "/usr/lib64/python3.9/enum.py", line 432, in __getitem__
ceph-nvmeof-nvmeof-1  |     return cls._member_map_[name]
ceph-nvmeof-nvmeof-1  | KeyError: 'IPV6'
ceph-nvmeof-nvmeof-1  |
ceph-nvmeof-nvmeof-1  | The above exception was the direct cause of the following exception:
ceph-nvmeof-nvmeof-1  |
ceph-nvmeof-nvmeof-1  | Traceback (most recent call last):
ceph-nvmeof-nvmeof-1  |   File "/usr/lib64/python3.9/runpy.py", line 197, in _run_module_as_main
ceph-nvmeof-nvmeof-1  |     return _run_code(code, main_globals, None,
ceph-nvmeof-nvmeof-1  |   File "/usr/lib64/python3.9/runpy.py", line 87, in _run_code
ceph-nvmeof-nvmeof-1  |     exec(code, run_globals)
ceph-nvmeof-nvmeof-1  |   File "/src/control/__main__.py", line 36, in <module>
ceph-nvmeof-nvmeof-1  |     gateway.serve()
ceph-nvmeof-nvmeof-1  |   File "/src/control/server.py", line 113, in serve
ceph-nvmeof-nvmeof-1  |     self._start_discovery_service()
ceph-nvmeof-nvmeof-1  |   File "/src/control/server.py", line 151, in _start_discovery_service
ceph-nvmeof-nvmeof-1  |     DiscoveryService(self.config).start_service()
ceph-nvmeof-nvmeof-1  |   File "/src/control/discovery.py", line 1030, in start_service
ceph-nvmeof-nvmeof-1  |     callback(key.fileobj, mask)
ceph-nvmeof-nvmeof-1  |   File "/src/control/discovery.py", line 979, in nvmeof_tcp_connection
ceph-nvmeof-nvmeof-1  |     raise UnknownNVMEOpcode(opcode) from e
ceph-nvmeof-nvmeof-1  | control.discovery.DiscoveryService.nvmeof_tcp_connection.<locals>.UnknownNVMEOpcode: unsupported opcode: 2

[gbregman]

It seems the issue is in:

class ADRFAM_TYPES(enum.IntFlag):
    ipv4 = 0x1
    ipv6 = 0x2
    ib = 0x3
    fc = 0x4
    intra_host = 0xfe

we use lower case, but the value we get from OMAP is upper case, "IPV4" or "IPV6". when we try to use it we get an exception.

Notice that the SPDK code ignores the case when parsing this value:

spdk_nvme_transport_id_parse_adrfam(enum spdk_nvmf_adrfam *adrfam, const char *str)
{
        if (adrfam == NULL || str == NULL) {
                return -EINVAL;
        }

        if (strcasecmp(str, "IPv4") == 0) {
                *adrfam = SPDK_NVMF_ADRFAM_IPV4;
        } else if (strcasecmp(str, "IPv6") == 0) {
                *adrfam = SPDK_NVMF_ADRFAM_IPV6;
        } else if (strcasecmp(str, "IB") == 0) {
                *adrfam = SPDK_NVMF_ADRFAM_IB;
        } else if (strcasecmp(str, "FC") == 0) {
                *adrfam = SPDK_NVMF_ADRFAM_FC;
        } else {
                return -ENOENT;
        }
        return 0;
}

[gbregman]

So, running a create_listener command with a parameter of "--adrfam IPV4" or "--adrfam IPV6" would cause the crash when running "nvme discover".