Re: [2.6 patch] SCSI_AIC94XX must depend on SCSI

[Adrian Bunk <bunk@kernel.org>]

On Thu, Mar 06, 2008 at 02:31:07PM -0600, James Bottomley wrote:
> On Thu, 2008-03-06 at 22:22 +0200, Adrian Bunk wrote:
> > This patch fixes the following build error reported by Toralf Förster:
> > 
> > <--  snip  -->
> > 
> > ...
> >   MODPOST 243 modules
> > ERROR: "sas_bios_param" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_change_queue_type" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_change_queue_depth" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_target_destroy" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_target_alloc" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_slave_destroy" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_slave_configure" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_eh_bus_reset_handler" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_eh_device_reset_handler" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_queuecommand" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_ioctl" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_register_ha" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_domain_attach_transport" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_ssp_task_response" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_slave_alloc" [drivers/scsi/mvsas.ko] undefined!
> > ERROR: "sas_unregister_ha" [drivers/scsi/mvsas.ko] undefined!
> > make[1]: *** [__modpost] Error 1
> > 
> > <--  snip  -->
> > 
> > Signed-off-by: Adrian Bunk <bunk@kernel.org>
> > 
> > ---
> > --- a/drivers/scsi/aic94xx/Kconfig
> > +++ b/drivers/scsi/aic94xx/Kconfig
> > @@ -26,7 +26,7 @@
> >  
> >  config SCSI_AIC94XX
> >  	tristate "Adaptec AIC94xx SAS/SATA support"
> > -	depends on PCI
> > +	depends on SCSI && PCI
> >  	select SCSI_SAS_LIBSAS
> >  	select FW_LOADER
> 
> This doesn't look to be a fix for your error: the symbols you've listed
> as missing are all libsas ones, not core SCSI ones.  Also they're all
> claiming to be missing from mvsas, not aic94xx.  There's something else
> about the configuration that produced this error.

This is the right fix for this bug.

We have:
- SCSI=m
- SCSI_AIC94XX=y
- SCSI_MVSAS=m
- SCSI_SAS_LIBSAS=y

The problem is:
- SCSI_AIC94XX selects (and requires) SCSI_SAS_LIBSAS
- SCSI_MVSAS selects (and requires) SCSI_SAS_LIBSAS
- with SCSI=m a SCSI_SAS_LIBSAS=y is not linked into the kernel

SCSI_AIC94XX=y also results in the AIC94xx driver not being linked into 
the kernel here (and that's also fixed by this patch), but this alone 
wouldn't result in a build bug...

> James

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed