[illumos-Developer] Code Review: WRITE_SAME support for COMSTAR's stmf_sbd

[Richard Elling]

On May 18, 2011, at 10:44 AM, Dan McDonald wrote:

> On Wed, May 18, 2011 at 10:22:03AM -0700, Richard Elling wrote:
>> I'd like to see a dtrace probe around the actual I/O, sbd_scsi.c line 2307
>> so we can measure how long the actual write same I/O takes.
> 
> Hmmm... you could measure this with judicious FBT, I would imagine:

[pedantic answer :-)]
Yes, of course, with fbt we can do many things. But it is much easier and less intrusive
for performance monitoring and troubleshooting to uplevel the probes to logical levels.
 -- richard


> 
> ===================== (Cut up to and including here.) =====================
> 
> sbd_write_same_data:entry
> {
>        self->trace = 1;
>        self->time = 0;
> }
> 
> sbd_data_write:entry
> /self->trace == 1/
> {
>        self->curtime = timestamp;
> }
> 
> sbd_data_write:return
> /self->trace == 1/
> {
>        self->time += (timestamp - self->curtime);
> }
> 
> sbd_write_same_data:return
> {
>        self->trace = 0;
>        printf("I/O time in one call is %d", self->time);
> }
> 
> ===================== (Cut up to and including here.) =====================
> 
> But if you've a more precise probe semantic in mind, I'm game.  Please expand
> a little so I know what you're shooting for.
> 
> Thanks,
> Dan
> 
> _______________________________________________
> Developer mailing list
> Developer at lists.illumos.org
> http://lists.illumos.org/m/listinfo/developer