On Wed, 2010-05-19 at 11:16 -0700, Joe Eykholt wrote:
> >> */
> >> -static void fcoe_ctlr_age_fcfs(struct fcoe_ctlr *fip)
> >> +static unsigned long fcoe_ctlr_age_fcfs(struct fcoe_ctlr *fip)
> >> {
> >> struct fcoe_fcf *fcf;
> >> struct fcoe_fcf *next;
> >> + unsigned long next_timer = jiffies +
> msecs_to_jiffies(FIP_DEF_FKA);
>
> Also, this should be further out than 8 seconds, in case there's
> nothing sooner needed. That's why I used the 90 second time before.
> The rest of the code in age_fcfs or the timeout should come up with
> a closer value unless the 90 second keep-alive is the next thing
> needed
> or there's nothing at all to do.
>
This default value mostly gets written when fd_flag is zero due to
sooner E node KA with typical 8 sec KA period and large value was always
effective in case fd_flag set that was an issue since in that case
periodic ka timeouts didn't kick after 90 second of fcf selection while
fd_flag set.
I can roll back this change but then have to set KA period timeout for
this case when fcf is selected in fcoe_ctlr_timeout, so it would add few
more lines there after fcoe_ctlr_select calling as:-
if (sel && sel->fd_flags &&
time_after(next_timer, sel->time + sel->fka_period))
next_timer = sel->time + fcf->fka_period;
I'm fine with either and let me know if you want me to change this.
Thanks
Vasu
_______________________________________________
devel mailing list
[email protected]
http://www.open-fcoe.org/mailman/listinfo/devel
