wpa_supplicant ior NM crashes wireless router (dd-wrt)

[Fjodor]

Binary package hint: wpasupplicant

Upon upgrade to Natty, wifi-connection to my gf's router (Buffalo WZR-HP-G300NH with dd-wrt) drops with no possibility to reconnect until after a router power cycle. Since rsyslogd in Natty is non-functional atm, I can't provide logs, and dd-wrt seems not to keep logs over power cycles.

I upgraded to Natty over this connection without loss of network connectivity but upon booting Natty, I lose connectivity so quickly that fetching the latest kernel from kernel.org took 4 (!) router power cycles to complete.

Wireless NIC in my laptop is Intel 5100.

Please advise as to how to debug.

Note: I leave for home tomorrow, May 5th, but may return next week. Thus, expedient replies today, May 4th, would be preferable.

[Fjodor (sune-molgaard)]

Anyone?

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Miguelángel León (migueleonm)]

Same problen in oneiric with intel 5100 and 6200 wireless card. Both are dual band n and router crashes in both bands.

[Fjodor (sune-molgaard)]

commit 2249b011432ca3dcce112f0f71e0f531b4bb9347 in the 3.1 series fixes it for me. See https://lkml.org/lkml/2011/9/28/666 for discussion.

Thus I suggest pulling said commit.

What is interesting for me, and the reason why I reported it as a wpa_supplicant/NM bug is, that the original commit that this patch reverts was actually present in the kernel that I ran (2.6.39 - unsupported, I know) under Maverick, but the untoward behaviour only manifested itself on upgrading to Natty, which has a kernel (2.6.38) with this change, but that was still outdated relative to 2.6.39 that I was running, so I booted *with that exact same kernel* that I had used to download all the packages to upgrade to Natty under Maverick with no problems, but upon booting Natty (with that 2.6.39 kernel), the problem manifested itself immediately.

Thus, to work with upstream to find out why this seemingly benign change *that didn't cause trouble under Maverick* actually *is* a problem, looking into differences between wpa_supplicant/NM from Maverick to Natty could be beneficial.

[Fjodor (sune-molgaard)]

Hmmm, status is "Confirmed" but it is still unassigned. I have presented a full and working workaround, namely pulling a commit from 3.1 that undoes a change from 2.6.38 but no one seems interested?

What's up with that?