Re: [PATCH RFC v3 1/2] tpm: Keep CLKRUN enabled throughout the duration of transmit_cmd()

[Jason Gunthorpe <jgg@ziepe.ca>]

On Mon, Nov 20, 2017 at 07:34:07PM +0000, Shaikh, Azhar wrote:

> >However, why not have check_locality, release_locality, wait_startup use
> >clk_toggle instead? That seems better to me??
> >
> 
> I think, better to have the clk disable/enable at one place instead
> of adding it all locations where TPM access is done. The clk_toggle
> is kind of a special case where, for the complete duration of the
> TPM command processing, the CLKRUN protocol was supposed to be
> disabled. For rest of the places we can disable and re-enable the
> clkrun as soon as possible.

> Also since it is part of the read/write APIs we won't miss any other
> location in the code where TPM access is done.

But I wonder if you will hit the same bug that motivated this change
in the other places?

It seems clearer to me to have a strong guard of the clock across the
entire high level action, since this hardware is so broken.

You could add a debug to ensure that read/write is never without
clk_enable being held.

Jason