On 8/16/19, 12:56 AM, "Burakov, Anatoly" <anatoly.bura...@intel.com> wrote:
<snip> > @@ -89,6 +96,65 @@ set_tsc_freq(void) > eal_tsc_resolution_hz = freq; > } > > +static void > +set_tsc_freq_secondary(void) > +{ > + struct rte_mp_msg mp_req; > + struct rte_mp_reply mp_reply; > + struct timer_mp_param *r; > + struct timespec ts = {.tv_sec = 1, .tv_nsec = 0}; > + > + memset(&mp_req, 0, sizeof(mp_req)); > + strcpy(mp_req.name, EAL_TIMER_MP); > + if (rte_mp_request_sync(&mp_req, &mp_reply, &ts) || mp_reply.nb_received != 1) { If rte_mp_request_sync returns 0 but mp_reply.nb_receieved isn't set to 1, you'll be leaking mp_reply.msgs. Ha - of course you're right. I didn't notice this when using the VFIO code as a reference for how to code this. I'll respin this patch and then push patches to fix the VFIO code separately. Thanks, -Jim