[ECOS] possible bug in if_fcc ethernet driver?

[ECOS] possible bug in if_fcc ethernet driver?

[agyhoo]

Hi,

In the if_fcc ethernet driver, the fcc_eth_init routine
returns "false" if there is no LINK detected (ie, network
cable is unplugged). However, there is a compile switch
around the return false statement (CYGPKG_REDBOOT).
Therefore, when I run my eCos application, it hangs when
the network cable is unplugged. I think this is because the
network alarm thread and the background thread think the
device status is available and try to access it.

Am I missing something? Can someone tell me why the compile
switch is present?

Thanks.

- Anunoy


	
		
__________________________________
Do you Yahoo!?
Friends.  Fun.  Try the all-new Yahoo! Messenger.
http://messenger.yahoo.com/ 

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] possible bug in if_fcc ethernet driver?

[Gary Thomas]

On Mon, 2004-06-14 at 09:15, agyhoo wrote:
> Hi,
> 
> In the if_fcc ethernet driver, the fcc_eth_init routine
> returns "false" if there is no LINK detected (ie, network
> cable is unplugged). However, there is a compile switch
> around the return false statement (CYGPKG_REDBOOT).
> Therefore, when I run my eCos application, it hangs when
> the network cable is unplugged. I think this is because the
> network alarm thread and the background thread think the
> device status is available and try to access it.
> 
> Am I missing something? Can someone tell me why the compile
> switch is present?

We explicitly want RedBoot to not try and use the network if there
is no cable in place (it just wastes time).  In the case of eCos,
I doubt that your program "hangs", but rather it is waiting to
get an ethernet IP address via DHCP (the default behaviour).  If
the network is unplugged, this process will take quite a long time
(more than 30 seconds), waiting for the network to come available.
Eventually, the DHCP will timeout and your code will continue,
but if it needs the network, it will probably have other problems.

-- 
Gary Thomas <gary@mlbassoc.com>
MLB Associates


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] possible bug in if_fcc ethernet driver?

[agyhoo]

In my eCos configuration, I've chosen to initialize
the eth0 interface manually. My initialization routine
(called from within my application routine), tries to
get params via bootp first. If that fails, it then
sets up a bootp record (using data from flash) and
then initializes the network calling the init_net fn.

However, I dont even get to my application's main routine.
The code seems to "hang" in the cyg_net_init function where
the alarm thread and the background network threads are
created.

- Anunoy

--- Gary Thomas <gary@mlbassoc.com> wrote:
> On Mon, 2004-06-14 at 09:15, agyhoo wrote:
> > Hi,
> > 
> > In the if_fcc ethernet driver, the fcc_eth_init routine
> > returns "false" if there is no LINK detected (ie,
> network
> > cable is unplugged). However, there is a compile switch
> > around the return false statement (CYGPKG_REDBOOT).
> > Therefore, when I run my eCos application, it hangs
> when
> > the network cable is unplugged. I think this is because
> the
> > network alarm thread and the background thread think
> the
> > device status is available and try to access it.
> > 
> > Am I missing something? Can someone tell me why the
> compile
> > switch is present?
> 
> We explicitly want RedBoot to not try and use the network
> if there
> is no cable in place (it just wastes time).  In the case
> of eCos,
> I doubt that your program "hangs", but rather it is
> waiting to
> get an ethernet IP address via DHCP (the default
> behaviour).  If
> the network is unplugged, this process will take quite a
> long time
> (more than 30 seconds), waiting for the network to come
> available.
> Eventually, the DHCP will timeout and your code will
> continue,
> but if it needs the network, it will probably have other
> problems.
> 
> -- 
> Gary Thomas <gary@mlbassoc.com>
> MLB Associates
> 
> 



	
		
__________________________________
Do you Yahoo!?
Friends.  Fun.  Try the all-new Yahoo! Messenger.
http://messenger.yahoo.com/ 

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] possible bug in if_fcc ethernet driver?

[Gary Thomas]

On Mon, 2004-06-14 at 09:46, agyhoo wrote:
> In my eCos configuration, I've chosen to initialize
> the eth0 interface manually. My initialization routine
> (called from within my application routine), tries to
> get params via bootp first. If that fails, it then
> sets up a bootp record (using data from flash) and
> then initializes the network calling the init_net fn.
> 
> However, I dont even get to my application's main routine.
> The code seems to "hang" in the cyg_net_init function where
> the alarm thread and the background network threads are
> created.

How do you know that it doesn't get to your code?
Does it work properly if the network is plugged in?

The network threads that you mention don't do any network I/O
without an application first doing something, plus they should
never "hang".  I think there must be some other problem here.

> 
> - Anunoy
> 
> --- Gary Thomas <gary@mlbassoc.com> wrote:
> > On Mon, 2004-06-14 at 09:15, agyhoo wrote:
> > > Hi,
> > > 
> > > In the if_fcc ethernet driver, the fcc_eth_init routine
> > > returns "false" if there is no LINK detected (ie,
> > network
> > > cable is unplugged). However, there is a compile switch
> > > around the return false statement (CYGPKG_REDBOOT).
> > > Therefore, when I run my eCos application, it hangs
> > when
> > > the network cable is unplugged. I think this is because
> > the
> > > network alarm thread and the background thread think
> > the
> > > device status is available and try to access it.
> > > 
> > > Am I missing something? Can someone tell me why the
> > compile
> > > switch is present?
> > 
> > We explicitly want RedBoot to not try and use the network
> > if there
> > is no cable in place (it just wastes time).  In the case
> > of eCos,
> > I doubt that your program "hangs", but rather it is
> > waiting to
> > get an ethernet IP address via DHCP (the default
> > behaviour).  If
> > the network is unplugged, this process will take quite a
> > long time
> > (more than 30 seconds), waiting for the network to come
> > available.
> > Eventually, the DHCP will timeout and your code will
> > continue,
> > but if it needs the network, it will probably have other
> > problems.
> > 
> > -- 
> > Gary Thomas <gary@mlbassoc.com>
> > MLB Associates
> > 
> > 
> 
> 
> 
> 	
> 		
> __________________________________
> Do you Yahoo!?
> Friends.  Fun.  Try the all-new Yahoo! Messenger.
> http://messenger.yahoo.com/
-- 
Gary Thomas <gary@mlbassoc.com>
MLB Associates


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] possible bug in if_fcc ethernet driver?

[agyhoo]

Gary,

See my comments below:

--- Gary Thomas <gary@mlbassoc.com> wrote:
> On Mon, 2004-06-14 at 09:46, agyhoo wrote:
> > In my eCos configuration, I've chosen to initialize
> > the eth0 interface manually. My initialization routine
> > (called from within my application routine), tries to
> > get params via bootp first. If that fails, it then
> > sets up a bootp record (using data from flash) and
> > then initializes the network calling the init_net fn.
> > 
> > However, I dont even get to my application's main
> routine.
> > The code seems to "hang" in the cyg_net_init function
> where
> > the alarm thread and the background network threads are
> > created.
> 
> How do you know that it doesn't get to your code?
[AG]: I set a break point at my application main.
I also have other routines in my main function that 
I put breakpoints on but I never get there. I am using
a POSIX style main function btw.

> Does it work properly if the network is plugged in?
[AG]: Yes. I tried this with both my networking code
(ie., bootp and static ip initialization fallback) compiled
in and compiled out. When the network cable is plugged
in, it works fine.

I also tried making the fcc_eth_init return false. When
I do this, the application executes correctly. Hence,
my suspicions....

- Anunoy
> 
> The network threads that you mention don't do any network
> I/O
> without an application first doing something, plus they
> should
> never "hang".  I think there must be some other problem
> here.
> 
> > 
> > - Anunoy
> > 
> > --- Gary Thomas <gary@mlbassoc.com> wrote:
> > > On Mon, 2004-06-14 at 09:15, agyhoo wrote:
> > > > Hi,
> > > > 
> > > > In the if_fcc ethernet driver, the fcc_eth_init
> routine
> > > > returns "false" if there is no LINK detected (ie,
> > > network
> > > > cable is unplugged). However, there is a compile
> switch
> > > > around the return false statement (CYGPKG_REDBOOT).
> > > > Therefore, when I run my eCos application, it hangs
> > > when
> > > > the network cable is unplugged. I think this is
> because
> > > the
> > > > network alarm thread and the background thread
> think
> > > the
> > > > device status is available and try to access it.
> > > > 
> > > > Am I missing something? Can someone tell me why the
> > > compile
> > > > switch is present?
> > > 
> > > We explicitly want RedBoot to not try and use the
> network
> > > if there
> > > is no cable in place (it just wastes time).  In the
> case
> > > of eCos,
> > > I doubt that your program "hangs", but rather it is
> > > waiting to
> > > get an ethernet IP address via DHCP (the default
> > > behaviour).  If
> > > the network is unplugged, this process will take
> quite a
> > > long time
> > > (more than 30 seconds), waiting for the network to
> come
> > > available.
> > > Eventually, the DHCP will timeout and your code will
> > > continue,
> > > but if it needs the network, it will probably have
> other
> > > problems.
> > > 
> > > -- 
> > > Gary Thomas <gary@mlbassoc.com>
> > > MLB Associates
> > > 
> > > 
> > 
> > 
> > 
> > 	
> > 		
> > __________________________________
> > Do you Yahoo!?
> > Friends.  Fun.  Try the all-new Yahoo! Messenger.
> > http://messenger.yahoo.com/
> -- 
> Gary Thomas <gary@mlbassoc.com>
> MLB Associates
> 
> 



	
		
__________________________________
Do you Yahoo!?
Friends.  Fun.  Try the all-new Yahoo! Messenger.
http://messenger.yahoo.com/ 

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] possible bug in if_fcc ethernet driver?

[Peter Graf]

Gary Thomas wrote:

>On Mon, 2004-06-14 at 09:15, agyhoo wrote:
>  
>
>>Hi,
>>
>>In the if_fcc ethernet driver, the fcc_eth_init routine
>>returns "false" if there is no LINK detected (ie, network
>>cable is unplugged). However, there is a compile switch
>>around the return false statement (CYGPKG_REDBOOT).
>>Therefore, when I run my eCos application, it hangs when
>>the network cable is unplugged. I think this is because the
>>network alarm thread and the background thread think the
>>device status is available and try to access it.
>>
>>Am I missing something? Can someone tell me why the compile
>>switch is present?
>>    
>>
>
>We explicitly want RedBoot to not try and use the network if there
>is no cable in place (it just wastes time).  In the case of eCos,
>I doubt that your program "hangs", but rather it is waiting to
>get an ethernet IP address via DHCP (the default behaviour).  If
>the network is unplugged, this process will take quite a long time
>(more than 30 seconds), waiting for the network to come available.
>Eventually, the DHCP will timeout and your code will continue,
>but if it needs the network, it will probably have other problems.
>
>  
>
It might be a matter of timeout(s) length(s) around phy inialization. I 
vaguely remember I had to shorten them, so it recognizes the missing 
cable sooner.

All the best
Peter


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

[ECOS] Enabling Second Eth interface causes system stop

[Rajeev Gaba]

Hi,

We are using a Intel IXP 425 based board. In this if we remove the PHY
device for second ethernet port it stops executing and says No Network
Interfaces found.
I'm very new to this and would require some useful tips for this problem.

Regards,
Rajeev Gaba


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] possible bug in if_fcc ethernet driver?

[agyhoo]

Thanks for your response Peter.

I dont think that the PHY timeout is an issue in my case.
The phy initialization and indeed the MAC initialization
completes successfully. The hardware independent ethernet
driver sets the state as ACTIVE so the FreeBSD stack
thinks the device is ready to send/receive.

I still haven't figured out where the problem is. All I 
know is that something goes wrong in the initialization
code before my main routine kicks in. I suspect the
alarm timeout network thread because it calls the
ethdrv "tickle" API call that seems to do a send operation.

- Anunoy


		
__________________________________
Do you Yahoo!?
Yahoo! Mail Address AutoComplete - You start. We finish.
http://promotions.yahoo.com/new_mail 

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss