It's me again... In I2c.h I found: /* Mainboard arch_initcall() code should register all its I2C devices. * This is done at arch_initcall time, before declaring any i2c adapters. * Modules for add-on boards must use other calls. */ extern int i2c_register_board_info(int busnum, struct i2c_board_info const *info, unsigned n); I guess I have to call this function near the arch_initcall(). I greped my folder for arch_initcall() but couldn't figure out witch file to use... Any idea ? Another thing, what would be the equivalent of MODULE_DEVICE_TABLE(i2c, example_idtable); As you said, the i2c_device_id structure doesn't exist in this kernel... -----Original Message----- From: linux-i2c-owner@xxxxxxxxxxxxxxx [mailto:linux-i2c-owner@xxxxxxxxxxxxxxx] On Behalf Of Jean Delvare Sent: Wednesday, March 11, 2009 12:23 PM To: Ayman KHAMOUMA Cc: linux-i2c@xxxxxxxxxxxxxxx Subject: Re: Need help on selecting one (and only one) i2c bus Hi Ayman, On Wed, 11 Mar 2009 11:55:22 +0100, Ayman KHAMOUMA wrote: > After some researches I guess I have to use the normal_i2c and ignore > arrays: > > static unsigned short normal_i2c[] = { 0x71, I2C_CLIENT_END}; static > unsigned short ignore[] = { 0x01, 0x71, 0x02, 0x71, I2C_CLIENT_END}; > > static struct i2c_client_address_data addr_data = { > .normal_i2c = normal_i2c, > .ignore = ignore, > }; > > Int the init function I have to allocate a major to the driver, and > than add the I2c driver: > i2c_add_driver(&i2c_my_drv); > > After doing that, i2c_my_drv.attach_adapter is called. > > Which will call (among others) my probe function, in which I call > i2c_attach_client. > > > Am I in the right way ? No. You're using the old, deprecated way to instantiate devices, and you are abusing it (the ignore array wasn't meant to be used the way you do.) You said you're using kernel 2.6.23.17, it's not brand new, but it's not too old either. At any rate, it already includes the new-style device binding. So, instead of using the old binding model and then fight to get your driver to _not_ attach to improper devices/addresses, you should use the new model and explicitly instantiate your device, so that your driver won't even try to bind to the wrong devices. I have posted an explanation about how to do this recently: http://marc.info/?l=linux-i2c&m=123659553713018&w=2 You should convert your driver to a new-style one and use either method 1 or method 2, whichever is easier for you. See the following document for a guide on how to update your code: http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob_plai n;f=Documentation/i2c/upgrading-clients;hb=HEAD It's for kernel 2.6.26 and later but it should mostly apply to 2.6.23 too. The only big difference I can think of is that 2.6.23 doesn't have struct i2c_device_id, instead the device driver binding is based on the i2c driver name. I can help you with this if you can't figure it out. As a side note, the fact that you needed to add buses 1 and 2 to the ignore list suggests that something is wrong with your I2C bus driver. It is rather unlikely that all 3 I2C buses on your system have a device at address 0x71, so your old-style driver shouldn't have instantiated 3 devices. If it did, it strongly suggests that the quick write i2c-core is using to check for the presence of a device at a given address improperly reported success in all cases. You really should check your I2C bus driver and in particular the implementation of 0-byte messages before going further with the device driver. -- Jean Delvare http://khali.linux-fr.org/wishlist.html -- To unsubscribe from this list: send the line "unsubscribe linux-i2c" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-i2c" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
