dependance loop on 2.6.14-rc1-mm1

dependance loop on 2.6.14-rc1-mm1

[trem]

Hi

I've tried to compile a 2.6.14-rc1-mm1 on my amd64. When I do the make 
modules_install,
I have this warning:

if [ -r System.map -a -x /sbin/depmod ]; then /sbin/depmod -ae -F 
System.map  2.6.14-rc1-mm1; fi
WARNING: Module 
/lib/modules/2.6.14-rc1-mm1/kernel/drivers/serial/serial_cs.ko ignored, 
due to loop
WARNING: Module 
/lib/modules/2.6.14-rc1-mm1/kernel/drivers/serial/serial_core.ko 
ignored, due to loop
WARNING: Module 
/lib/modules/2.6.14-rc1-mm1/kernel/drivers/serial/8250_pnp.ko ignored, 
due to loop
WARNING: Module 
/lib/modules/2.6.14-rc1-mm1/kernel/drivers/serial/8250_pci.ko ignored, 
due to loop
WARNING: Module 
/lib/modules/2.6.14-rc1-mm1/kernel/drivers/serial/8250_acpi.ko ignored, 
due to loop
WARNING: Loop detected: 
/lib/modules/2.6.14-rc1-mm1/kernel/drivers/serial/8250.ko needs 
serial_core.ko which needs 8250.ko again!
WARNING: Module 
/lib/modules/2.6.14-rc1-mm1/kernel/drivers/serial/8250.ko ignored, due 
to loop
WARNING: Module 
/lib/modules/2.6.14-rc1-mm1/kernel/drivers/parport/parport_serial.ko 
ignored, due to loop
WARNING: Module 
/lib/modules/2.6.14-rc1-mm1/kernel/drivers/char/mwave/mwave.ko ignored, 
due to loop



You can found the .config I've used here : 
http://www.zarb.org/~trem/config_loop.txt

It's a allmodconfig config with  all ISDN and "Digi International NEO 
PCI Support" set to OFF.
I've "removed" both option because they generate error when compiling.

I don't understand why I have this warning.

Thanks for any help,
trem

Re: dependance loop on 2.6.14-rc1-mm1

[Russell King]

On Sat, Sep 17, 2005 at 01:40:54PM +0200, trem wrote:
> I've tried to compile a 2.6.14-rc1-mm1 on my amd64. When I do the make 
> modules_install,
> I have this warning:
> 
> WARNING: Loop detected:
> /lib/modules/2.6.14-rc1-mm1/kernel/drivers/serial/8250.ko needs 
> serial_core.ko which needs 8250.ko again!

This looks suspicious.  8250 should need serial_core, but there's no
way in hell serial_core should require 8250. 

Seems to be caused by the kgdb patches, which add the following to
serial_core:

+#ifdef CONFIG_KGDB
+       {
+               extern int kgdb_irq;
+
+               if (port->irq == kgdb_irq)
+                       return;
+       }
+#endif
+

and kgdb_irq comes from the 8250 module.

Tom, can this dependency be solved before kgdb goes near mainline
please?

-- 
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 Serial core

Re: dependance loop on 2.6.14-rc1-mm1

[trem]

Hi

I've done a little test, I've remove this line from serial_core.c

+#ifdef CONFIG_KGDB
+       {
+               extern int kgdb_irq;
+
+               if (port->irq == kgdb_irq)
+                       return;
+       }
+#endif


and now the make modules_install works fine.

thanks for the help,
trem



Russell King a écrit :

>On Sat, Sep 17, 2005 at 01:40:54PM +0200, trem wrote:
>  
>
>>I've tried to compile a 2.6.14-rc1-mm1 on my amd64. When I do the make 
>>modules_install,
>>I have this warning:
>>
>>WARNING: Loop detected:
>>/lib/modules/2.6.14-rc1-mm1/kernel/drivers/serial/8250.ko needs 
>>serial_core.ko which needs 8250.ko again!
>>    
>>
>
>This looks suspicious.  8250 should need serial_core, but there's no
>way in hell serial_core should require 8250. 
>
>Seems to be caused by the kgdb patches, which add the following to
>serial_core:
>
>+#ifdef CONFIG_KGDB
>+       {
>+               extern int kgdb_irq;
>+
>+               if (port->irq == kgdb_irq)
>+                       return;
>+       }
>+#endif
>+
>
>and kgdb_irq comes from the 8250 module.
>
>Tom, can this dependency be solved before kgdb goes near mainline
>please?
>
>  
>

Re: dependance loop on 2.6.14-rc1-mm1

[Tom Rini]

On Sat, Sep 17, 2005 at 12:51:38PM +0100, Russell King wrote:
> On Sat, Sep 17, 2005 at 01:40:54PM +0200, trem wrote:
> > I've tried to compile a 2.6.14-rc1-mm1 on my amd64. When I do the make 
> > modules_install,
> > I have this warning:
> > 
> > WARNING: Loop detected:
> > /lib/modules/2.6.14-rc1-mm1/kernel/drivers/serial/8250.ko needs 
> > serial_core.ko which needs 8250.ko again!
> 
> This looks suspicious.  8250 should need serial_core, but there's no
> way in hell serial_core should require 8250. 
> 
> Seems to be caused by the kgdb patches, which add the following to
> serial_core:
> 
> +#ifdef CONFIG_KGDB
> +       {
> +               extern int kgdb_irq;
> +
> +               if (port->irq == kgdb_irq)
> +                       return;
> +       }
> +#endif
> +
> 
> and kgdb_irq comes from the 8250 module.
> 
> Tom, can this dependency be solved before kgdb goes near mainline
> please?

My KGDB isn't in -mm yet, that comes from George Anzginer's version.
This is a non-issue in mine, it really just adds yet another/different
release function.

-- 
Tom Rini
http://gate.crashing.org/~trini/