Re: SAR panic

Hi, Igor.
I think, first of all, panic must be eliminated.
So reduce client sdu size to SAR_SEGM_SIZE*255 is Ok.
I'm not sure about changing SAR_SEGM_SIZE. If nokia gw uses value 576,
may be some nokia models doesn't work stable with large SAR_SEGM_SIZE?
Any way, current values are tested already, so new values must be also
tested first.

Igor Ivoilov wrote:

> Dear Slava,
> 
> While ESAR is not implemented I think a graceful way to hanlde it is
> to reduce client sdu size sent to wsp to SAR_SEGM_SIZE*255.
> Also I can change SAR_SEGM_SIZE to be let's say 1300.
> it will change the maximum file size to 331K from 146K.
> But it may cause a problem for the phones that are connected using CSD.

> Actually SAR_SEGM_SIZE has to be equal to MTU.
> Do you know a way to detect MTU to the phone that works fast enough?
> 
> Let me know what you think
> 
> Igor
> 
>  > -----Original Message-----
>  > From: Vjacheslav Chekushin [mailto:[email protected]]
>  > Sent: Thursday, January 23, 2003 9:31 AM
>  > To: devel
>  > Subject: SAR panic
>  >
>  >
>  > Hi, list.
>  > Seems that SAR incorrectly handles cases when segments count>255
>  >
>  > wapbox.log:
>  >
>  > 2003-01-22 19:14:01 [6] DEBUG: WTP: resp_machine 718, state
>  > RESULT_RESP_WAIT,
>  > event RcvAck.
>  > 2003-01-22 19:14:01 [6] DEBUG: WTP: continue_sar_result():
>  > lsegm=254, nsegm=440,
>  > csegm=251
>  > 2003-01-22 19:14:01 [6] DEBUG: WTP: dispath_to_wdp(): psn = 255
>  > 2003-01-22 19:14:01 [6] PANIC: gwlib/octstr.c:1677:
>  > octstr_set_bits: Assertion
>  > `value <= mask' failed.
>  >
>  >
>  > --
>  > Vjacheslav Chekushin                                mailto:[email protected]
>  > Latvian Mobile Phone Company                        http://www.lmt.lv
>  >
>  >
> 


-- 
Vjacheslav Chekushin                                mailto:[email protected]
Latvian Mobile Phone Company                        http://www.lmt.lv