Ticket #7950 (new defect)

Opened 18 months ago

Last modified 17 months ago

Sudden lost connectivity

Reported by: joe Owned by: mbletsas
Priority: blocker Milestone: 8.2.0 (was Update.2)
Component: wireless Version: not specified
Keywords: blocks-:8.2.0 Cc: charlie, mstone, kimquirk, wad, ashish
Action Needed: never set Verified: no
Deployments affected: Blocked By:
Blocking:

Description

Two of the C2 laptops (CSN74801BDF and CSN74702114) running joyride-2263 suddenly lost connectivity while being connected to Internet running Firefox.

The log files are attached.

Attachments

logs.CSN74801BDF.2008-08-13.21-00-13.tar.bz2 (60.7 kB) - added by joe 18 months ago.
logs.CSN74702114.2008-08-13.20-57-48.tar.bz2 (44.3 kB) - added by joe 18 months ago.
net_hang_logs_pgf.tgz (79.0 kB) - added by pgf 18 months ago.

Change History

Changed 18 months ago by joe

Changed 18 months ago by joe

Changed 18 months ago by joe

Was able to wake them up by touching (multiple times) the power buttons. Looks like the problem is related to the "suspend" state.

Eventually was able to reconnect machines to the Internet.

Changed 18 months ago by joe

Eventually those machines (plus one more) just died... and required a hard reboot (power cycle) to work again.

What's suspected (after a discussion w/Kim, Michael and Chris): the combination of an 1cc-specific wireless environment and a newly introduced "suspect" feature might be a culprit.

Changed 18 months ago by joe

Sorry, "suspend", not "suspect" ;-)

Changed 18 months ago by ashish

  • cc ashish added

Changed 18 months ago by cjb

  • keywords blocks?:8.2 added

Someone needs to diagnose this based on the logs.

Changed 18 months ago by pgf

i have another, possibly-related network lockup.

joyride-2294, with inhibit-idle-suspend in place. not using NM (using my own network stumbler for setup -- it browses with iwlist, and configures with iwconfig. no daemon, and no iwpriv involved.)

was connected this morning to 1cc wifi. joined network at 11:42 am. at 12:32, closed lid, went to lunch. 13:04, opened laptop. did testing of touchpad. sometime before 13:25 noticed that browser couldn't make connections. neither could ssh. ifconfig and iwconfig outputs appeared normal (up and running, and associated, respectively), and dhclient was still running. suspended and resumed with power button. didn't help. ran my network stumbler, requested reassociation, problem gone.

will attach log.

Changed 18 months ago by pgf

Changed 18 months ago by pgf

i think i understand my issue, though it may or may not be "expected". if the machine goes out of range of an AP, at least while suspended, it will not reassociate when it comes back in range of that AP. (with NM not running.) with NM running, it seems to reconnect, though it takes a surprising amount of time (minutes?).

we should probably make sure "AP goes out of range" is in our test plan. on a former project i was involved with, reassociation was sometimes problematic in that case. (and it was with marvell h/w.)

Changed 18 months ago by kimquirk

This might be fixed with latest wireless firmware. #7973

Changed 17 months ago by cjb

  • keywords blocks-:8.2.0 added; blocks?:8.2 removed

Downgrading until we have this reproduced some more times, since it might be fixed by firmware.

Note: See TracTickets for help on using tickets.