[PATCH linux 15/15] drivers/fsi: Initialize CFAMs for read/write access

[Christopher Bostic]

On Thu, Oct 6, 2016 at 8:21 PM, Jeremy Kerr <jk at ozlabs.org> wrote:
> Hi Chris,
>
>> A break can be issued on any available link.
>
> OK.
>
>> Due to hardware bugs that vary depending on the combination of
>> master type and cfam type the break command must be issued to
>> different ID's and addresses accordingly.  This was the reason for
>> adding the cfam_id, break_id, etc.. to the fsi_master struct.
>
> But (according to the docs I have) a break is just a logical 1 for 256
> clocks - it contains no addressing ID / information. What do the
> cfam_ids and break_ids represent "on the wire"?

Right, at the protocol level a break is simply 256 1's.   One of the
cascaded masters on P8,P9 though has quirks where you must
adjust the target address and ID when writing to the actual hardware.
I don't know the hardware specifics as to what the bug is but
the logic responsible for the translation of a firmware initiated MMIO
onto a link that results in the 256 1's has a bug that corrupts that
message in some way.  Note that this is even an issue for the gpio
master since the bug I just described is in the P8,P9 cascaded master
hardware.

Sounds like we need  to schedule a meeting to probably
go over all the hardware quirks that need to be dealt with.  Exchanging
emails probably isn't the quickest way.

Long story short we'll still need a master->break( ) operation defined
for each master type.

-Chris

>
> Cheers,
>
>
> Jeremy