We are running a BCM43362 in conjunction with an STM32F415. We are seeing an issue with roaming. Every 2 to 8 roams causes the system to no longer be able to connect to the AP. This is experienced with "wiced_wifi_scan" failing to return with WICED_SUCCESS. Sometimes the system, fails to jump to another ap and we can no longer send with no deauth being reported by the link_handler. Other times the roam succeeds, but soon after the system fails to send with a deauth also never being reported. When this occurs, "host_buffer_get" seems to always be failing on "pbuf_alloc".
The memory used for wiced and lwip buffers has a large number of fragments but still has a lot of space. Once we get into this state, the only recovery is a system restart.
Has anyone also experienced this?
Our process for roaming is fairly simple. We have the link_handler subscribed to the roamed event, which alerts us that a roam occured. When this happens, we renew the DHCP lease but don't wait for it to complete. Other than that, we don't stop any sockets or prevent them from send/receiving.
EDIT: we are also not receiving the link events for WLC_E_ROAM_START. This along with WLC_E_ROAM_PREP would be useful to manage our low power scheme, but even though they are added to the link_events list, they never seem to be reported by the platform. WLC_E_ROAM does get posted after a roam completes.
Message was edited by: Felix Turgeon
