Bug 1062 - rx_packets increasing without traffic
: rx_packets increasing without traffic
Status: VERIFIED FIXED
: IPW3945
Data Transfer
: 1.0.5
: All All
: P2 normal
Assigned To:
:
:
:
:
:
  Show dependency treegraph
 
Reported: 2006-06-07 11:37 by
Modified: 2006-08-10 22:50 (History)


Attachments
Fix for the increasing rx_packets counter (549 bytes, patch)
2006-06-11 12:43, Markus Reitt
Details | Diff
patch for rx_packets (617 bytes, patch)
2006-06-12 09:32, Markus Reitt
Details | Diff
Move rx_packets increment into data frame handler. (701 bytes, patch)
2006-06-15 08:14, James Ketrenos
Details | Diff


Note

You need to log in before you can comment on or make changes to this bug.


Description From 2006-06-07 11:37:59
Although there is no traffic on the link (ethereal),
/sys/class/net/eth1/statistics/rx_packets increases around 8 packets every
second. This screws up programs showing/analysing network-traffic, sine they
indicate incoming traffic when there is none.
------- Comment #1 From 2006-06-08 08:01:41 -------
the packets are beacons, can you set the mode to monitor to check the packets 
by ethereal?
------- Comment #2 From 2006-06-08 09:44:33 -------
Neither tcpdump nor ethereal show any packets in monitor mode. But when
associated, the counter increases
------- Comment #3 From 2006-06-10 22:32:41 -------
They're indeed beacons (forgot to compile RADIOTAP in). Nonetheless, this bug is
valid IMO since not one single driver I used so far counts beacons (ipw2200,
rt2500, madwifi).
Is see no reason why rx_packets counter should increase for packets that are not
directly meant for the host.
------- Comment #4 From 2006-06-11 12:43:27 -------
Created an attachment (id=823) [details]
Fix for the increasing rx_packets counter

The attached patch fixes the problem for me, found in the ML. Now the counter
works as expected. Any reason for not including this?
------- Comment #5 From 2006-06-12 09:32:15 -------
Created an attachment (id=825) [details]
patch for rx_packets

Sorry, the other patch was b0rked...
------- Comment #6 From 2006-06-15 08:14:02 -------
Created an attachment (id=828) [details]
Move rx_packets increment into data frame handler.

Makes sense to only count rx_packets for data frames since we only count
rx_bytes for data payload.  

Attached is the patch I'll merge in for 1.0.10 (I haven't pushed the other
snapshots to SF but you can pull them from GIT at
http://bughost.org/repos/ipw3945.git/ if you want)  

Let me know if it does what you expect.

Thanks,
James
------- Comment #7 From 2006-06-17 03:38:41 -------
Can't check it currently since the driver fails to build with
sim_rx.c:63:36: Error: net/ieee80211_radiotap.h: No such file or directory
Maybe it will work if I patch the kernel with ieee80211, but I'm not to eager to
do that (it's annoying enought to have two ieee80211 stacks, on in the kernel
and one slightly different standalone one).
------- Comment #8 From 2006-06-17 05:16:10 -------
OK, put some ieee80211-stuff into /usr/include/net, but the driver (1.0.10kdmr)
from today git) still counts incoming packets without any actual traffic. My
patch worked better :)
------- Comment #9 From 2006-06-18 03:13:45 -------
OK, the reason the problem didn't disappear was that the patch is not in git.
Oops, didn't notice that at first. OK, the patch works fine here.
Thanks for the fix.
------- Comment #10 From 2006-06-22 07:53:38 -------
Wait this patch integrate in the 1.0.11.
------- Comment #11 From 2006-06-23 15:44:55 -------
Merged for 1.0.11.  Setting to fixed.
------- Comment #12 From 2006-07-18 22:23:26 -------
Markus, can you try ipw3945-1.0.12 or ipw3945-1.1.0-pre2 and verify this bug 
if it has been fixed?
------- Comment #13 From 2006-07-18 22:46:18 -------
Yes, it works just fine now. Thanks.