Re: [Madwifi-devel] HAL_MIB_STATS implemented?

[Sam L. <sa...@er...>]

On Monday 12 April 2004 10:21 pm, Tsung Chwen Woo wrote:
> Hi Sam,
>
> I'm using atheros 5212 chipset (Netgear WG511T) with ath_hal: 0.9.8.6,
> wlan: 0.7.3.1 BETA, ath_pci: 0.8.5.4 BETA
>
> Here's the code I've used:
>
> In if_athvar.h, i've included the following:
>
> #define     ath_hal_updatemibcounters(_ah, _mibstats) \
>             ((*(_ah)->ah_updateMibCounters)((_ah), (_mibstats)))
>
>
>
> In ath_ioctl() function (about line 3253) in if_ath.c, i've added an extra
> case to retrieve the mib counters (in pretty much the same manner as
> athstats):
>
>    case SIOCGMIBCOUNTERS: {
>      struct ath_hal *ah = sc->sc_ah;
>      struct ieee80211com *ic = &sc->sc_ic;
> !    HAL_MIB_STATS hmibstats;
> !    ath_hal_updatemibcounters(ah, &hmibstats);
> !    ic->mib_counters.mib_rx_rts_successes = hmibstats.rts_good;
> !    ic->mib_counters.mib_rx_rts_failures = hmibstats.rts_bad;
> !    ic->mib_counters.mib_rx_ack_failures = hmibstats.ackrcv_bad;
> !    ic->mib_counters.mib_rx_fcs_errors = hmibstats.fcs_bad;
>      ic->mib_counters.mib_tx_frames = ic->mib_counters.mib_tx_fragments;
>      if (copy_to_user(ifr->ifr_data, &ic->mib_counters,
>          sizeof (ic->mib_counters)))
>       return -EFAULT;
>      return 0;
>    }

updatemibcounters does a += for each statistic so you're incrementing 
uninitialized variables.  It appears you have a mib_counters variable which 
you should just supply directly instead.  (BTW, this belongs in the driver 
softc structure and not the 802.11 layer as it's Atheros-specific.)  Beware 
also that each call to updatemibcounters will also reset the h/w counters so 
you need to insure two operations don't happen concurrently.  The best way 
(probably) to do this is to poll periodically in a single-threaded context 
(e.g. off a timer) and then return the current values through the ioctl.

>
>
> The lines with a "!" is how I tried to extract the HAL_MIB_STATS struct.
> The ic->mib_counters struct contains just a list of MIB counters members.
> Outputting the stats will involve an application similar in workings to
> athstats :-) . Am I on the right track in extracting these stats?
>
> Will I be correct to assume that sc_stats.ast_rx_crcerr is in fact
> equivalent to hmibstats.fcs_bad?

fcs_bad is probably more accurate as ast_rx_crcerr only counts frames that 
actually get sent up to the host.

>
> Also that sc_stats.ast_bmiss is equivalent to hmibstats.beacons? Or would
> hmibstats.beacons be count for all received beacon mgt frames?

beacons is the count of beacon frames received. ast_bmiss is the count of 
beacon miss interrupts.

	Sam