Re: [Patch 2/3] driver data: pass zfcp data through blktrace

[Martin Peschke <mpeschke@linux.vnet.ibm.com>]

On Mon, 2008-10-06 at 12:59 +0200, Heiko Carstens wrote:
> > +	trace.flags = 0;
> > +	trace.magic = cpu_to_be32(ZFCP_BLK_DRV_DATA_MAGIC);
> > +	if (fsf_req->adapter->adapter_features & FSF_FEATURE_MEASUREMENT_DATA) {
> > +		trace.flags |= ZFCP_BLK_LAT_VALID;
> > +		lat_inf = &fsf_req->qtcb->prefix.prot_status_qual.latency_info;
> > +		trace.channel_lat = cpu_to_be64(lat_inf->channel_lat * ticks);
> > +		trace.fabric_lat = cpu_to_be64(lat_inf->fabric_lat * ticks);
> > +	}
> > +	if (fsf_req->status & ZFCP_STATUS_FSFREQ_ERROR)
> > +		trace.flags |= ZFCP_BLK_REQ_ERROR;
> > +	trace.inb_usage = cpu_to_be16(fsf_req->qdio_inb_usage);
> > +	trace.outb_usage = cpu_to_be16(fsf_req->qdio_outb_usage);
> 
> Could we please not add any cpu_to_be* calls to s390 only code? s390 is a big
> endian architecture and therefore these calls are confusing and pointless.
> 
> > +#else
> > +#define zfcp_fsf_trace_latency(fsf_req)	do { } while (0)
> > +#endif
> 
> Please convert that to a static inline function for type checking.


Correct. Fixed patch will follow..

Thanks,
Martin