On Mon, Nov 2, 2009 at 9:00 PM, Sasha Khapyorsky <[email protected]> wrote:
> On 06:50 Mon 02 Nov     , Hal Rosenstock wrote:
>>
>> Move some code out into separate handle_babbling_port routine
>> for readability/maintainability
>>
>> Signed-off-by: Hal Rosenstock <[email protected]>
>> ---
>> diff --git a/opensm/opensm/osm_trap_rcv.c b/opensm/opensm/osm_trap_rcv.c
>> index 5790461..1621fbc 100644
>> --- a/opensm/opensm/osm_trap_rcv.c
>> +++ b/opensm/opensm/osm_trap_rcv.c
>> @@ -312,6 +312,61 @@ static void log_trap_info(osm_log_t *p_log, 
>> ib_mad_notice_attr_t *p_ntci,
>>                       cl_ntoh16(source_lid), cl_ntoh64(trans_id));
>>  }
>>
>> +static int handle_babbling_port(osm_sm_t *sm, ib_mad_notice_attr_t *p_ntci,
>> +                             uint32_t num_received,
>> +                             boolean_t physp_change_trap,
>> +                             osm_physp_t **pp_physp,
>> +                             boolean_t *run_heavy_sweep,
>> +                             uint64_t *event_wheel_timeout)
>> +{
>> +     if (print_num_received(num_received))
>> +             OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3804: "
>> +                     "Received trap %u times consecutively\n",
>> +                     num_received);
>> +     /* If the trap provides info about a bad port, mark it as unhealthy. */
>> +     if (physp_change_trap == TRUE) {
>> +             /* get the port */
>> +             *pp_physp = get_physp_by_lid_and_num(sm,
>> +                                                  
>> cl_ntoh16(p_ntci->data_details.ntc_129_131.lid),
>> +                                                  
>> p_ntci->data_details.ntc_129_131.port_num);
>> +
>> +             if (!*pp_physp)
>> +                     OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3805: "
>> +                             "Failed to find physical port by lid:%u 
>> num:%u\n",
>> +                             
>> cl_ntoh16(p_ntci->data_details.ntc_129_131.lid),
>> +                             p_ntci->data_details.ntc_129_131.port_num);
>> +             else {
>> +                     /* When babbling port policy option is enabled and
>> +                        Threshold for disabling a "babbling" port is 
>> exceeded */
>> +                     if (sm->p_subn->opt.babbling_port_policy &&
>> +                         num_received >= 250 &&
>> +                         disable_port(sm, *pp_physp) == 0)
>> +                             return 1;
>> +
>> +                     OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
>> +                             "Marking unhealthy physical port by lid:%u 
>> num:%u\n",
>> +                             
>> cl_ntoh16(p_ntci->data_details.ntc_129_131.lid),
>> +                             p_ntci->data_details.ntc_129_131.port_num);
>> +                     /* check if the current state of the p_physp is 
>> healthy. If
>> +                        it is - then this is a first change of state. Run a 
>> heavy sweep.
>> +                        if it is not - no need to mark it again - just 
>> restart the timer. */
>> +                     if (osm_physp_is_healthy(*pp_physp)) {
>> +                             osm_physp_set_health(*pp_physp, FALSE);
>> +                             /* Make sure we sweep again - force a heavy 
>> sweep. */
>> +                             /* The sweep should be done only after the 
>> re-registration, or
>> +                                else we'll be losing track of the timer. */
>> +                             *run_heavy_sweep = TRUE;
>> +                     }
>> +                     /* If we are marking the port as unhealthy - we want to
>> +                        keep this for a longer period of time than the
>> +                        OSM_DEFAULT_TRAP_SUPRESSION_TIMEOUT. Use the
>> +                        OSM_DEFAULT_UNHEALTHY_TIMEOUT */
>> +                     *event_wheel_timeout = OSM_DEFAULT_UNHEALTHY_TIMEOUT;
>> +     }
>> +     return 0;
>> +}
>> +
>
> It looks that you are moving the code as it is to a separate function
> where many flow switchers (run_heavy_sweep, event_wheel_timeout,
> pp_physp) are magically modified.
>
> What about something like below instead?

Reviewed-by: Hal Rosenstock <[email protected]>

>
> Sasha
>
>
> >From 5ed9114a32f9536123af4bd135c7cc603d340f92 Mon Sep 17 00:00:00 2001
> From: Sasha Khapyorsky <[email protected]>
> Date: Tue, 3 Nov 2009 03:46:54 +0200
> Subject: [PATCH] opensm/osm_trap_rcv.c: some consolidation
>
> Consolidate noisy port handling code in shutup_noisy_port() function.
> As before it will try to disable port or to mark it unhealthy.
>
> Signed-off-by: Sasha Khapyorsky <[email protected]>
> ---
>  opensm/opensm/osm_trap_rcv.c |  113 
> ++++++++++++++++++++----------------------
>  1 files changed, 53 insertions(+), 60 deletions(-)
>
> diff --git a/opensm/opensm/osm_trap_rcv.c b/opensm/opensm/osm_trap_rcv.c
> index 0ee4e77..dae892e 100644
> --- a/opensm/opensm/osm_trap_rcv.c
> +++ b/opensm/opensm/osm_trap_rcv.c
> @@ -226,7 +226,6 @@ static int disable_port(osm_sm_t *sm, osm_physp_t *p)
>        uint8_t payload[IB_SMP_DATA_SIZE];
>        osm_madw_context_t context;
>        ib_port_info_t *pi = (ib_port_info_t *)payload;
> -       int ret;
>
>        /* select the nearest port to master opensm */
>        if (p->p_remote_physp &&
> @@ -236,10 +235,6 @@ static int disable_port(osm_sm_t *sm, osm_physp_t *p)
>        /* If trap 131, might want to disable peer port if available */
>        /* but peer port has been observed not to respond to SM requests */
>
> -       OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3810: "
> -               "Disabling physical port 0x%016" PRIx64 " num:%u\n",
> -               cl_ntoh64(osm_physp_get_port_guid(p)), p->port_num);
> -
>        memcpy(payload, &p->port_info, sizeof(ib_port_info_t));
>
>        /* Set port to disabled/down */
> @@ -253,15 +248,10 @@ static int disable_port(osm_sm_t *sm, osm_physp_t *p)
>        context.pi_context.light_sweep = FALSE;
>        context.pi_context.active_transition = FALSE;
>
> -       ret = osm_req_set(sm, osm_physp_get_dr_path_ptr(p),
> -                         payload, sizeof(payload), IB_MAD_ATTR_PORT_INFO,
> -                         cl_hton32(osm_physp_get_port_num(p)),
> -                         CL_DISP_MSGID_NONE, &context);
> -       if (ret)
> -               OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3811: "
> -                       "Request to set PortInfo failed\n");
> -
> -       return ret;
> +       return osm_req_set(sm, osm_physp_get_dr_path_ptr(p),
> +                          payload, sizeof(payload), IB_MAD_ATTR_PORT_INFO,
> +                          cl_hton32(osm_physp_get_port_num(p)),
> +                          CL_DISP_MSGID_NONE, &context);
>  }
>
>  static void log_trap_info(osm_log_t *p_log, ib_mad_notice_attr_t *p_ntci,
> @@ -301,6 +291,42 @@ static void log_trap_info(osm_log_t *p_log, 
> ib_mad_notice_attr_t *p_ntci,
>                        cl_ntoh16(source_lid), cl_ntoh64(trans_id));
>  }
>
> +static int shutup_noisy_port(osm_sm_t *sm, uint16_t lid, uint8_t port,
> +                            unsigned num)
> +{
> +       osm_physp_t *p = get_physp_by_lid_and_num(sm, lid, port);
> +       if (!p) {
> +               OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3805: "
> +                       "Failed to find physical port by lid:%u num:%u\n",
> +                       lid, port);
> +               return -1;
> +       }
> +
> +       /* When babbling port policy option is enabled and
> +          Threshold for disabling a "babbling" port is exceeded */
> +       if (sm->p_subn->opt.babbling_port_policy && num >= 250) {
> +               OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
> +                       "Disabling noisy physical port 0x%016" PRIx64
> +                       ": lid %u, num %u\n",
> +                       cl_ntoh64(osm_physp_get_port_guid(p)), lid, port);
> +               if (disable_port(sm, p))
> +                       OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3811: "
> +                               "Failed to disable.\n");

Might want to include port information in this log message.

-- Hal

> +               else
> +                       return 1;
> +       }
> +
> +       /* check if the current state of the p_physp is healthy. If
> +          it is - then this is a first change of state. Run a heavy sweep. */
> +       if (osm_physp_is_healthy(p)) {
> +               OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
> +                       "Marking unhealthy physical port by lid:%u num:%u\n",
> +                       lid, port);
> +               osm_physp_set_health(p, FALSE);
> +               return 2;
> +       }
> +       return 0;
> +}
>  /**********************************************************************
>  **********************************************************************/
>  static void trap_rcv_process_request(IN osm_sm_t * sm,
> @@ -438,7 +464,7 @@ static void trap_rcv_process_request(IN osm_sm_t * sm,
>                /* Now we know how many times it provided this trap */
>                if (num_received > 10) {
>                        if (print_num_received(num_received))
> -                               OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3804: "
> +                               OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
>                                        "Received trap %u times 
> consecutively\n",
>                                        num_received);
>                        /*
> @@ -446,49 +472,17 @@ static void trap_rcv_process_request(IN osm_sm_t * sm,
>                         * we mark it as unhealthy.
>                         */
>                        if (physp_change_trap == TRUE) {
> -                               /* get the port */
> -                               p_physp = get_physp_by_lid_and_num(sm,
> -                                                                  cl_ntoh16
> -                                                                  
> (source_lid),
> -                                                                  port_num);
> -
> -                               if (!p_physp)
> -                                       OSM_LOG(sm->p_log, OSM_LOG_ERROR,
> -                                               "ERR 3805: "
> -                                               "Failed to find physical port 
> by lid:%u num:%u\n",
> -                                               cl_ntoh16(source_lid),
> -                                               port_num);
> -                               else {
> -                                       /* When babbling port policy option 
> is enabled and
> -                                          Threshold for disabling a 
> "babbling" port is exceeded */
> -                                       if (sm->p_subn->opt.
> -                                           babbling_port_policy
> -                                           && num_received >= 250
> -                                           && disable_port(sm, p_physp) == 0)
> -                                               goto Exit;
> -
> -                                       OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
> -                                               "Marking unhealthy physical 
> port by lid:%u num:%u\n",
> -                                               cl_ntoh16(source_lid),
> -                                               port_num);
> -                                       /* check if the current state of the 
> p_physp is healthy. If
> -                                          it is - then this is a first 
> change of state. Run a heavy sweep.
> -                                          if it is not - no need to mark it 
> again - just restart the timer. */
> -                                       if (osm_physp_is_healthy(p_physp)) {
> -                                               osm_physp_set_health(p_physp,
> -                                                                    FALSE);
> -                                               /* Make sure we sweep again - 
> force a heavy sweep. */
> -                                               /* The sweep should be done 
> only after the re-registration, or
> -                                                  else we'll be losing track 
> of the timer. */
> -                                               run_heavy_sweep = TRUE;
> -                                       }
> -                                       /* If we are marking the port as 
> unhealthy - we want to
> -                                          keep this for a longer period of 
> time than the
> -                                          
> OSM_DEFAULT_TRAP_SUPRESSION_TIMEOUT. Use the
> -                                          OSM_DEFAULT_UNHEALTHY_TIMEOUT */
> -                                       event_wheel_timeout =
> -                                           OSM_DEFAULT_UNHEALTHY_TIMEOUT;
> -                               }
> +                               int ret = shutup_noisy_port(sm,
> +                                                           
> cl_ntoh16(source_lid),
> +                                                           port_num,
> +                                                           num_received);
> +                               if (ret == 1) /* port disabled */
> +                                       goto Exit;
> +                               else if (ret == 2) /* unhealthy - run sweep */
> +                                       run_heavy_sweep = TRUE;
> +                               /* in any case increase timeout interval */
> +                               event_wheel_timeout =
> +                                   OSM_DEFAULT_UNHEALTHY_TIMEOUT;
>                        }
>                }
>
> @@ -508,8 +502,7 @@ static void trap_rcv_process_request(IN osm_sm_t * sm,
>                if (num_received > 10 && run_heavy_sweep == FALSE) {
>                        if (print_num_received(num_received))
>                                OSM_LOG(sm->p_log, OSM_LOG_VERBOSE,
> -                                       "Continuously received this trap %u 
> times. Ignoring\n",
> -                                       num_received);
> +                                       "Ignoring noisy traps.\n");
>                        goto Exit;
>                }
>        }
> --
> 1.6.5.1
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> the body of a message to [email protected]
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

Reply via email to