[PATCH v1] ESIA : Dummy eisa_driver_register should return error code

[PATCH v1] ESIA : Dummy eisa_driver_register should return error code

[Arvind Yadav]

The inline eisa_driver_register stub simply allows compilation on
systems with CONFIG_EISA disabled. the dummy eisa_driver_register
does not register an *_eisa_driver at all. The inline
eisa_driver_register should return to indicate lack of support
when attempting to register an *_eisa_driver on such a system with
CONFIG_EISA disabled.

Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
---
 include/linux/eisa.h | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/include/linux/eisa.h b/include/linux/eisa.h
index 6925249..cbfc684 100644
--- a/include/linux/eisa.h
+++ b/include/linux/eisa.h
@@ -70,7 +70,10 @@ void eisa_driver_unregister (struct eisa_driver *edrv);
 
 #else /* !CONFIG_EISA */
 
-static inline int eisa_driver_register (struct eisa_driver *edrv) { return 0; }
+static inline int eisa_driver_register(struct eisa_driver *edrv)
+{
+	return -ENODEV;
+}
 static inline void eisa_driver_unregister (struct eisa_driver *edrv) { }
 
 #endif /* !CONFIG_EISA */
-- 
1.9.1

Re: [PATCH v1] ESIA : Dummy eisa_driver_register should return error code

[Christoph Hellwig]

On Tue, Jul 19, 2016 at 12:15:01AM +0530, Arvind Yadav wrote:
> The inline eisa_driver_register stub simply allows compilation on
> systems with CONFIG_EISA disabled. the dummy eisa_driver_register
> does not register an *_eisa_driver at all. The inline
> eisa_driver_register should return to indicate lack of support
> when attempting to register an *_eisa_driver on such a system with
> CONFIG_EISA disabled.

Why?  The idea is that you can simply leave the stub in if registering
say PCI / EISA.  With your change such a driver will fail to load even
if it could serve PCI device in a !CONFIG_EISA config, which is the
behavior we want.

Re: [PATCH v1] ESIA : Dummy eisa_driver_register should return error code

[arvind Yadav]

SCSI_AHA1740 : This is support for a SCSI host adapter.
If CONFIG_EISA is disable then we should not register eisa driver.
static __init int aha1740_init (void)
{
     return eisa_driver_register (&aha1740_driver);

}

static __exit void aha1740_exit (void)
{
         eisa_driver_unregister (&aha1740_driver);
}

module_init (aha1740_init);
module_exit (aha1740_exit);
----------------------------------------
-Shall we add check by using CONFIG_EISA. We can avoid to register a dummy
driver.

static __init int aha1740_init (void)
{
         int err = 0;
#ifdef CONFIG_EISA
         err = eisa_driver_register (&aha1740_driver);
#endif
         return err;
}

static __exit void aha1740_exit (void)
{
#ifdef CONFIG_EISA
         eisa_driver_unregister (&aha1740_driver);
#endif
}

On Tuesday 19 July 2016 10:01 AM, Christoph Hellwig wrote:
> On Tue, Jul 19, 2016 at 12:15:01AM +0530, Arvind Yadav wrote:
>> The inline eisa_driver_register stub simply allows compilation on
>> systems with CONFIG_EISA disabled. the dummy eisa_driver_register
>> does not register an *_eisa_driver at all. The inline
>> eisa_driver_register should return to indicate lack of support
>> when attempting to register an *_eisa_driver on such a system with
>> CONFIG_EISA disabled.
> Why?  The idea is that you can simply leave the stub in if registering
> say PCI / EISA.  With your change such a driver will fail to load even
> if it could serve PCI device in a !CONFIG_EISA config, which is the
> behavior we want.