[OOB 5.5.0] ublox_evk_odin_w2 never gets to main

[mray190]

ublox_evk_odin_w2 never gets to main. Compiles just fine, deploy to the device, but then hangs and never displays anything or does anything.

Compiler: IAR, GCC_ARM, and ARM
GCC_ARM version: 5.4.1 20160919 (release)
mbed CLI version: 1.1.1
mbed OS version: 92fbf2a9b3988d430482fc25a6077f2462e2a634

[0xc0170]

cc @andreaslarssonublox

[ashok-rao]

@mray19027 may not change anything major, but can you please try gcc v6 (mbed OS 5.5 officially supported)..

[andreaslarssonublox]

We'll have a look at it.
int ref TE_OM2-72

[tomasfrisbergublox]

@0xc0170

It seems to be two separate problems. One of them requires a new driver release from us. This will be done within the next couple of days.

Also, it seems to be a problem in how the interfaces are registered into the lwip stack.

In “lwip_stack.c”, the function “mbed_lwip_bringup_2” calls “mbed_lwip_emac_init(NULL)”, which calls the “netif_add” function in “lwip_netif.c”.

The NULL parameter will be a problem since it is forwarded as the state variable to the “netif_add” function.
/* remember netif specific state information data */
netif->state = state;

Further down, in “netif_add”, the installed “init” callback generates a call to “emac_lwip_if_init” in “emac_lwip.c”. Here the state variable is cast to the emac pointer which ends up in a null pointer crash.
err_t emac_lwip_if_init(struct netif *netif)
{
int err = ERR_OK;
emac_interface_t *mac = (emac_interface_t *)netif->state;

mac->ops.set_link_input_cb(mac, emac_lwip_input, netif);
mac->ops.set_link_state_cb(mac, emac_lwip_state_change, netif);

Hence, according to my understanding this cannot work for the DEVICE_EMAC solution.

The solution could be either of the two proposals

1. An “mbed_lwip_emac_initiated” variables that makes sure the code of the “mbed_lwip_emac_init” function is executed only once. Since we do a separate “mbed_lwip_emac_init” call when registering the interface this means it will not execute code from the bringup2 function.
2. Replace the “mbed_lwip_emac_init(NULL)” with “ret = NSAPI_ERROR_OK;” since it will not work for the DEVICE_EMAC case.
   #if !DEVICE_EMAC
   ret = mbed_lwip_emac_init(NULL);
   #else
   ret = NSAPI_ERROR_OK;
   #endif

[0xc0170]

@tomasfrisbergublox Thanks for the description. This seems is the same issue as this patch is fixing : ARMmbed/mbed-os#4444 ? Please review

In case it is not, this would be worth having it in mbed-os repository, can you please add it there?

[tomasfrisbergublox]

@0xc0170

The ARMmbed/mbed-os#4444 patch is working also for us. We will continue testing and generate a new binary hopefully tomorrow.

Thanks

[0xc0170]

The ARMmbed/mbed-os#4444 patch is working also for us. We will continue testing and generate a new binary hopefully tomorrow.

Please do, so it can be added to 5.5.1

[0xc0170]

The status update, this issue has some dependencies

• Fix STM32 crashes on boot due to unset VTOR mbed-os#4543
• resolve lwip init twice issue mbed-os#4444 (5.5.1)
• Updated u-blox ODIN-W2 driver binaries to 2.1 rc1 mbed-os#4567 (5.5.1/2)

Once the last one gets integrated, it should work . @tomasfrisbergublox please correct me if it is not the case

[MarceloSalazar]

As 5.5.1 is released, can @tomasfrisbergublox please confirm whether this is fixed and can be closed?

[bulislaw]

Closing. Please reopen if it's still a problem.