Hi,
I'm using NetwokManager-0.7(using dhcpcd) with wifi, and see a lot of
messages like "Received DHCP event from unexpected PID xxxx (expected
xxxx)" in the syslog.
I also see more than one dhcpcd processes running. Sometimes, only one
dhcpcd process lefts at last, but sometimes not. (searching google, find the
same problem someone has reported:
https://bugs.launchpad.net/ubuntu/+source/network-manager/+bug/255829)
Following the source code(NM-0.7), in the function
<nm_dhcp_client_start> (nm-dhcp-dhcpcd.c), it kills the previous dhcpcd
process and saves the new one's pid in device->pid. But the function
<dhcp_watch_cb>, knowing the termination of the previous dhcpd process, sets
device->pid to zero. So the current dhcp process's pid is lost, and
<nm_dhcp_manager_handle_event> will print log like "Received DHCP event from
unexpected PID 8812 (expected 0)".
Also the previous dhcpcd process's time_out handler may expire, and
affects the current running dhcpcd process?
It seems the function <supplicant_interface_acquire> (nm-device-wifi.c)
registers a function <supplicant_iface_connection_state_cb> as the handler
of the signal "connection-state" .
And:
<supplicant_iface_connection_state_cb> shcedules
<supplicant_iface_connection_state_cb_handler>, which calls
<nm_device_activate_schedule_stage3_ip_config_start>, which shcedules
<nm_device_activate_stage3_ip_config_start> , which calls
<real_act_stage3_ip_config_start>, which calls
<nm_dhcp_manager_begin_transaction>.
So it is possiable that more than one dhcpcd processes running meanwhile.
To fix this, shoud I do:
1) Protect the corresponding data struct
or
2) Add a clean step when a second dhcpcd process to be started
Any ideas? Thanks.
_______________________________________________
NetworkManager-list mailing list
[email protected]
http://mail.gnome.org/mailman/listinfo/networkmanager-list
