#441: duplicate node list entries in ad-hoc mode
----------------------------------------+-----------------------------------
Reporter: [EMAIL PROTECTED] | Owner:
Type: defect | Status: new
Priority: major | Milestone: version 0.9.0 - move to
new codebase
Component: madwifi: 802.11 stack | Version: trunk
Resolution: | Keywords: adhoc duplicate node
Patch_attached: 0 |
----------------------------------------+-----------------------------------
Changes (by Vikram):
* patch_attached: 1 => 0
Comment:
Impact of this on the [http://madwifi.org/ticket/334] ticket
Good News ! Maybe !
My setup had three ad-hoc nodes (s1, s2 and r). 'r' is the receiver
running two netserver intances. 's1' and 's2' ran netperf (UDP_STREAM) to
the netservers on 'r'.
If 'r' somehow has duplicate entries for 's1', then the traffic coming
into 'r' from 's1' causes heavy memory leaks. NOTE that if in this
situation 's2' sends traffic to 'r', that does NOT cause any problems.
Sometimes the duplicate entry does not happen, I am not sure why. But when
they do, it is virtually assured that the memory increase will cause a
panic in the kernel.
I downloaded the r1457+above patch and did the above experiment again. The
duplicate entries did not occur, and hence the massive leaks leading to
panic did not happen. There was still some relatively minor leak, I am not
sure what the source of that is. But that one existed in r1454 as well.
Probably nothing to do with madwifi, but I am not sure.
I think someone else should test the r1457-along-with-the-patch using
iperf or netperf and see if [http://madwifi.org/ticket/334] ticket can be
deemed partially closed !
--
Ticket URL: <http://madwifi.org/ticket/441>
MadWifi <http://madwifi.org/>
Multiband Atheros Driver for Wireless Fidelity
