[Skiboot] [PATCH 1/2] include: Add PRD message definition

[Benjamin Herrenschmidt]

On Mon, 2015-02-16 at 15:42 +1100, Stewart Smith wrote:
> Neelesh Gupta <neelegup at linux.vnet.ibm.com> writes:
> > From: Jeremy Kerr <jk at ozlabs.org>
> >
> > Signed-off-by: Jeremy Kerr <jk at ozlabs.org>
> > Signed-off-by: Neelesh Gupta <neelegup at linux.vnet.ibm.com>
> > ---
> >  include/opal.h  |   41 ++++++++++++++++++++++++++++++++++++++++-
> >  include/xscom.h |    8 ++++++++
> >  2 files changed, 48 insertions(+), 1 deletion(-)
> 
> Add something to doc/opal-api/opal-messages.txt ?
> 
> > +enum opal_prd_msg_type {
> > +	OPAL_PRD_MSG_TYPE_INIT = 0,	/* RT --> FW */
> 
> RT?
> 
> > +struct opal_prd_msg {
> > +	uint8_t		type;
> > +	uint8_t		pad[3];
> > +	uint32_t	token;
> 
> Do we want a version in here?
> 
> I'm thinking of future when we have either:
> a) more types
> b) some types have extended information.
> 
> Please clearly document how OS should behave in the event of unknown
> types or newer version.

We just make a new message in that case, though we could use pad as
flags ...

> > +	union {
> > +		struct {
> > +			uint32_t	version;
> > +			uint64_t	ipoll;
> > +		} init;
> > +		struct {
> > +			uint64_t	proc;
> > +			uint64_t	ipoll_status;
> > +			uint64_t	ipoll_mask;
> > +		} attn;
> > +		struct {
> > +			uint64_t	proc;
> > +			uint64_t	ipoll_ack;
> > +		} attn_ack;
> > +		struct {
> > +			uint64_t	chip;
> > +		} occ_error;
> > +		struct {
> > +			uint64_t	chip;
> > +		} occ_reset;
> > +	};
> > +};
> 
> seeing as this is an opal_msg, can we also have a BUILD_ASSERT() of it
> being < sizeof(struct opal_msg) ?

Ack.

> Actually, recent discoveries suggest we should actually BUILD_ASSERT()
> on it being <= 72 as well.

Ben.