Re: [PATCH v2 3/3] tpm_tis: Clean up the force=1 module parameter

[Jason Gunthorpe <jgunthorpe@obsidianresearch.com>]

On Tue, Dec 01, 2015 at 08:33:58PM +0100, Uwe Kleine-König wrote:
> Hello,
> 
> On Tue, Dec 01, 2015 at 11:58:29AM -0700, Jason Gunthorpe wrote:
> > The TPM core has long assumed that every device has a driver attached,
> > however commit b8b2c7d845d5 ("base/platform: assert that dev_pm_domain
> > callbacks are called unconditionally") breaks that assumption.
> 
> you asked for an alternative wording here. What about:
> 
> 	The TPM core has long assumed that every device has a driver
> 	attached, which is not valid.

But it is valid, it is an invariant of the tpm core that a driver be
attached, and prior to 'b8b that has been satisfied.

>       This was noticed with commit
> 	b8b2c7d845d5 ("base/platform: assert that dev_pm_domain
> 	callbacks are called unconditionally") which made probing of the
> 	tpm_tis device fail by mistake and resulted in an oops later on.

The probe didn't fail, the 'b8b causes a NULL probe function to result
in no driver being attached.

How about:

 The TPM has for a long time required that every device it uses has an
 attached driver. In the force case the tpm_tis driver met this via
 platform_register_simple and a NULL probe function for the driver.
 However, commit b8b2c7d845d5 ("base/platform: assert that dev_pm_domain
 callbacks are called unconditionally") causes NULL probe functions
 to no longer bind a driver.

Did we ever reach a conclusion if Martin's patch should go ahead?

Jason