Re: [patch 2.6.28-rc5] i2c: remove i2c_adapter.clist_lock

[Jean Delvare <khali-PUYAD+kWke1g9hUCZPvPmw@public.gmane.org>]

On Tue, 25 Nov 2008 03:27:08 -0800, David Brownell wrote:
> On Tuesday 25 November 2008, Jean Delvare wrote:
> > 
> > > I did test this with the legacy "eeprom" and "i2c-stub" as you had
> > > suggested.  Worked fine ... this is different from previous passes,
> > > as it removes the lock instead of the list, and doesn't attempt to
> > > change how the list is (mis/ab)used.
> > 
> > Which kernel version? Since 2.6.27, the eeprom driver is no longer a
> > legacy driver. Instead it's a new-style driver with the
> > optional .detect() callback.
> 
> 2.6.28-rc6 ... hmm, seeing the top of that driver with
> 
>   /* Addresses to scan */
>   static const unsigned short normal_i2c[] = { 0x50, 0x51, 0x52, 0x53, 0x54,
>                                         0x55, 0x56, 0x57, I2C_CLIENT_END };
> 
>   /* Insmod parameters */
>   I2C_CLIENT_INSMOD_1(eeprom);
> 
> strongly implied that it's still a legacy driver.  But
> at the bottom, I see otherwise.

I voluntarily made the .detect() callback reuse the same data
structures, to make the transition easier. I still hope to be able to
kill the I2C_CLIENT_INSMOD*() macros someday though.

> (...) I suppose your "how to
> test such stuff" message predated 2.6.27 ... :)

Definitely.

-- 
Jean Delvare