Possible bug with CYGSEM_REDBOOT_NETWORK_INIT_ONE_DEVICE

Possible bug with CYGSEM_REDBOOT_NETWORK_INIT_ONE_DEVICE

[David Fernandez]

Hi there,

I'was about to use the feature CYGSEM_REDBOOT_NETWORK_INIT_ONE_DEVICE,
and I've noticed that may be it doesn't work as per the description in
the cdl option:

It says that it should stop initializing network devices as soon as one
is initialized successfully. But it seems that the initialization logic
for a default interface has someway confused the things.

It tries to initialize the default interface, but if it fails, then it
seems to initialize all of them... I think there should be a break
statement within the "if (t->init())" block (net_io.c: net_init()) if
this options is active and enabled.

May be the we can get rid of CYGSEM_REDBOOT_FLASH_CONFIG condition in
CYGDAT_REDBOOT_DEFAULT_NETWORK_DEVICE (doing the appropriate changes in
net_io.c: net_init()), so we could provide a default interface if there
is no flash configured.

Anybody thinks a patch to change all this is interesting?

Re: Possible bug with CYGSEM_REDBOOT_NETWORK_INIT_ONE_DEVICE

[Gary Thomas]

David Fernandez wrote:
> Hi there,
> 
> I'was about to use the feature CYGSEM_REDBOOT_NETWORK_INIT_ONE_DEVICE,
> and I've noticed that may be it doesn't work as per the description in
> the cdl option:
> 
> It says that it should stop initializing network devices as soon as one
> is initialized successfully. But it seems that the initialization logic
> for a default interface has someway confused the things.
> 
> It tries to initialize the default interface, but if it fails, then it
> seems to initialize all of them... I think there should be a break
> statement within the "if (t->init())" block (net_io.c: net_init()) if
> this options is active and enabled.
> 
> May be the we can get rid of CYGSEM_REDBOOT_FLASH_CONFIG condition in
> CYGDAT_REDBOOT_DEFAULT_NETWORK_DEVICE (doing the appropriate changes in
> net_io.c: net_init()), so we could provide a default interface if there
> is no flash configured.
> 
> Anybody thinks a patch to change all this is interesting?

Propose one and explain why it's useful.  Then we can discuss it.

Patches (and the discussion surrounding them) should be sent to:
   ecos-patches@ecos.sourceware.org

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------