[Click] packet pool assertion

[Eddie Kohler]

Ouch!  Thanks for this report, Roman.  The bug was in the user-level 
multithreading code, and was a bug in the assertion, not the code.  (The 
lost blocks were lost because the assertion aborted the code.)  I 
believe this is fixed.

Should v2.0.2 be released to address this error...?

Eddie


On 10/4/11 6:39 PM, rchertov wrote:
> The script below causes an assertion in the packet pool cleanup
> procedure.  Also, if
> this script is run in valgrind using options --leak-check=full
> --show-reachable=yes,
> then the tool detects a lot of lost but reachable blocks.  If the queue
> size is
> set to the default size, then no issues occur.
>
> q :: Queue(2000);
>
> src1p0  :: RatedSource(LENGTH 38, RATE 500000, LIMIT 3000, ACTIVE
> false)
>           ->  UDPIP6Encap(SRC 2000:10:1::01, SPORT 6667, DST
> 2000:20:1::01, DPORT 6667)
>           ->  SetIP6DSCP(0)
>           ->  EtherEncap(0x0800, 00:04:23:D0:93:63, 00:17:cb:0d:f8:01)
>           ->  uq1p0 :: Unqueue
>           ->  q;
>
> src2p0 :: RatedSource(LENGTH 38, RATE 500000, LIMIT 3000, ACTIVE false)
>           ->  UDPIP6Encap(SRC 2000:10:1::02, SPORT 6667, DST
> 2000:20:1::02, DPORT 6667)
>           ->  SetIP6DSCP(0)
>           ->  EtherEncap(0x0800, 00:04:23:D0:93:63, 00:17:cb:0d:f8:02)
>           ->  uq2p0 :: Unqueue
>           ->  q;
>
> q ->  Unqueue
>     ->  total_ctr :: AverageCounter
>     ->  cl :: Classifier(5/01, 5/02);
>
> cl[0] ->  ctr0 :: AverageCounter ->  Discard;
> cl[1] ->  ctr1 :: AverageCounter ->  Discard;
>
>
> Script(
> 	wait 2,
> 	write src1p0.active 1,
> 	write src2p0.active 1,
> 	wait 1,
> 	wait 1,
> 	wait 1,
> 	print "Flow 0",
> 	print ctr0.count,
> 	print ctr0.rate,
> 	print "Flow 1",
> 	print ctr1.count,
> 	print ctr1.rate,
> 	print "Total Counter",
> 	print total_ctr.count,
> 	print total_ctr.rate,
> );
>
>
> _______________________________________________
> click mailing list
> click at amsterdam.lcs.mit.edu
> https://amsterdam.lcs.mit.edu/mailman/listinfo/click