[Testing] some errors on iwconfig

[Ronak Chokshi]

Dan,
We investigated this issue.

The link quality, as reported by iwconfig, is computed by the libertas
driver based on RSSI, NF, and tx retries.
After a close inspection we found that minimum value of link quality is
always set to 100, that does not sound correct.

Please see the below code (see the text "read this" below).

../net/wireless/libertas/wext.c, function "wlan_get_wireless_stats"
{
...
priv->wstats.discard.misc = adapter->logmsg.ackfailure;
 
 /* Calculate quality */
 priv->wstats.qual.qual = max(quality, (u32)100);    <--- read this.
 priv->wstats.qual.updated = IW_QUAL_ALL_UPDATED | IW_QUAL_DBM;
...
}
 
I guess the above, instead, should be 'priv->wstats.qual.qual =
min(quality, (u32)100);'
 
Also, we checked the SNR (qual.level), NF (qual.noise) as reported by
the firmware and, found that firmware reports these values properly.

Thanks
Ronak

-----Original Message-----
From: Dan Williams [mailto:dcbw at redhat.com] 
Sent: Monday, July 02, 2007 4:55 AM
To: Ronak Chokshi
Cc: Michail Bletsas; supat at supat.eu.org; testing at laptop.org; Kim Quirk;
testing-bounces at lists.laptop.org
Subject: RE: [Testing] some errors on iwconfig

On Sun, 2007-07-01 at 10:11 -0700, Ronak Chokshi wrote:
> We have talked about populating link quality field with different
> sorts of values but we haven't gotten a consensus on that yet. I will
> dig up the latest email threads on that topic but suggestions are
> definitely welcome. Link quality in a mesh network usually doesn't
> hold too much of sense because there are several links to neighbor
> mesh nodes to start with. So, there is not one RF link that can be
> justified as a candidate for calculating link quality.

With WEXT, link "quality" (ie, qual.qual) is a linear scale field from 0
- 100 that may, at the option of the driver, take things like transmit
errors, decryption failures, etc, into account.

The "level" field (ie, qual.level) is where either dBm or raw RSSI
should go for the signal, and the "noise" field (ie, qual.noise) should
match the level semantics, but of course with noise and not signal.

So as long as these restrictions are met, whatever gets used in
qual.qual is probably fine by me.

Dan

>  
> 
> Thanks
> 
> Ronak
> 
> 
>                                    
> ______________________________________________________________________
> From:testing-bounces at lists.laptop.org
> [mailto:testing-bounces at lists.laptop.org] On Behalf Of Michail Bletsas
> Sent: Sunday, July 01, 2007 8:36 AM
> To: supat at supat.eu.org
> Cc: Dan Williams; testing at laptop.org; Kim Quirk;
> testing-bounces at lists.laptop.org
> Subject: Re: [Testing] some errors on iwconfig
> 
> 
>  
> 
> 
> I assume that you are using an XO. 
> What OS build does it run? 
> 
> I am not sure how link quality is calculated and I never pay too much
> attention to it. 
> Points 2,3 and 4 are normal 
> 
> M. 
> 
> 
> 
> 
> 
> supat at supat.eu.org 
> 
> 07/01/2007 10:46 AM 
> 
> 
>                To
> 
> 
> Michail Bletsas
> <mbletsas at laptop.org> 
> 
> 
>                cc
> 
> 
> Dan Williams
> <dcbw at redhat.com>, testing at laptop.org, Kim Quirk
<kim.quirk at gmail.com>, "OLPC Developer's List" <devel at laptop.org>,
testing-bounces at lists.laptop.org 
> 
> 
>           Subject
> 
> 
> some errors on
> iwconfig
> 
> 
> 
>  
> 
>  
> 
> 
>  
> 
> 
> 
> 
> 
> 
> 
> 
> always got below errors:
> 
> eth0      MRVL-USB8388  ESSID:"zxc555"  Nickname:"zxc55555"
>           Mode:Managed  Frequency:2.412 GHz  Access Point: 
> 00:15:E9:F9:29:CA
>           Bit Rate:11 Mb/s   Tx-Power=15 dBm
>           Retry limit:8   RTS thr=2347 B   Fragment thr=2346 B
>           Encryption key:off
>           Power Management:off
>           Link Quality=100/100  Signal level=-83 dBm  Noise level=-98
> dBm
>           Rx invalid nwid:0  Rx invalid crypt:0  Rx invalid frag:22023
>           Tx excessive retries:293  Invalid misc:648   Missed beacon:0
> 
> 
> 1. link quality alway 100% which is impossible
> 2. Rx invalid frag: always happen
> 3.  Tx excessive retries: always high
> 4. Invalid misc: alway increasing in numbers.
> 
> 
> 
>