Archive for July 14, 2009

Is the “Wrong AP” Issue I Described Kernel-Related?

Posted in acer aspire one, linux on July 14, 2009 by lucky

It didn’t take too long to compile and install kernel 2.6.30-1 on the Aspire One. I didn’t apply any patches, just vanilla sources. After I installed and rebooted, I looked at a few things. More on that in a second because after I looked at those things I looked at the kuki Linux site and saw they had 2.6.30-1 with a few patches including one I want: acerhfd to better control the AA1 fan. (Edited: My vanilla kernel has LED.)

After I installed it, I looked again at those things I mentioned before. Here’s the one thing that I found interesting.

Under my kernel, I saw that the first AP I associated with was my own. Under the kuki kernel, it went to the unencrypted SSID I’ve mentioned before. And under kuki it did a bit more. Under the generic Ubuntu kernel in crunchbang, there were only a couple attempts to associate with that AP. This kernel is like the fucking Energizer bunny: it just keeps going.

wlan0: direct probe to AP xx:xx:xx:xx:xx:xx try 1
wlan0: direct probe to AP xx:xx:xx:xx:xx:xx try 2
wlan0 direct probe responded
wlan0: authenticate with AP xx:xx:xx:xx:xx:xx
wlan0: authenticated
wlan0: associate with AP xx:xx:xx:xx:xx:xx
wlan0: RX AssocResp from xx:xx:xx:xx:xx:xx (capab=0x401 status=12 aid=0)
wlan0: AP denied association (code=12)
wlan0: RX AssocResp from xx:xx:xx:xx:xx:xx (capab=0x401 status=12 aid=0)
wlan0: AP denied association (code=12)
wlan0: RX AssocResp from xx:xx:xx:xx:xx:xx (capab=0x401 status=12 aid=0)
wlan0: AP denied association (code=12)
wlan0: RX AssocResp from xx:xx:xx:xx:xx:xx (capab=0x401 status=12 aid=0)
wlan0: AP denied association (code=12)
wlan0: RX AssocResp from xx:xx:xx:xx:xx:xx (capab=0x401 status=12 aid=0)
wlan0: AP denied association (code=12)
wlan0: RX AssocResp from xx:xx:xx:xx:xx:xx (capab=0x401 status=12 aid=0)
wlan0: AP denied association (code=12)
wlan0: associate with AP xx:xx:xx:xx:xx:xx
wlan0: RX AssocResp from xx:xx:xx:xx:xx:xx (capab=0x401 status=12 aid=0)
wlan0: AP denied association (code=12)
wlan0: RX AssocResp from xx:xx:xx:xx:xx:xx (capab=0x401 status=12 aid=0)
wlan0: AP denied association (code=12)
wlan0: associate with AP xx:xx:xx:xx:xx:xx
wlan0: association with AP xx:xx:xx:xx:xx:xx timed out
wlan0: authenticate with AP yy:yy:yy:yy:yy:yy
wlan0: authenticated
wlan0: associate with AP yy:yy:yy:yy:yy:yy
wlan0: RX AssocResp from yy:yy:yy:yy:yy:yy (capab=0x431 status=0 aid=4)
wlan0: associated

x=neighbor’s
y=mine

I didn’t believe that the first time I saw it so I rebooted into each kernel. This is my  third time switching between them and it holds true so far. When my kernel booted, it didn’t try to associate with anything but my own router — which is how it should work. When the kuki kernel is running (as it is right now), it keeps trying to associate with that other one, even more times than under the stock Ubuntu/#! kernel.

I don’t know if the difference between the kuki kernel and mine is just coincidence or if there’s some issue between patched kernels and vanilla that might have something to do with what I’ve been complaining about how it’s been trying to associate with an AP other than what’s specified in the relevant wireless configuration (which right now is still NetworkManager — and which I thought might have more to do with this).

For what it’s worth, my fan is running non-stop still even though lsmod shows acerhfd is loaded (along with some crypto modules I’ll need to blacklist). That fan module is the primary reason I’m using the kuki kernel right now. If I continue to see the same AP issues described above, I’m switching back to the vanilla kernel until I can figure out wtf is happening — and why.

PREVIOUS REFERENCES TO ISSUE:

http://lucky13linux.wordpress.com/2009/07/12/oh-no-not-again/

http://lucky13linux.wordpress.com/2009/07/11/getting-ready-for-crunchbang-de-bloat-etc/

http://lucky13linux.wordpress.com/2009/07/10/crunchbang-more-jwm-screenshots/

http://lucky13linux.wordpress.com/2009/07/10/crunchbang-update-jwm-aterm-acpiwifi-issues-and-office-software/

UPDATE1: Maybe it’s not the kernel after all. I removed the kuki kernel and booted my own for about the fifth time. This time it showed a pattern more similar to the generic/default kernel:

[   28.717834] ADDRCONF(NETDEV_UP): wlan0: link is not ready
[   30.585365] wlan0: direct probe to AP xx:xx:xx:xx:xx:xx try 1
[   30.784127] wlan0: direct probe to AP xx:xx:xx:xx:xx:xx try 2
[   30.984132] wlan0: direct probe to AP xx:xx:xx:xx:xx:xx try 3
[   31.185114] wlan0: direct probe to AP xx:xx:xx:xx:xx:xx timed out
[  129.061657] wlan0: authenticate with AP yy:yy:yy:yy:yy:yy
[  129.063526] wlan0: authenticated
[  129.063540] wlan0: associate with AP yy:yy:yy:yy:yy:yy
[  129.065835] wlan0: RX AssocResp from yy:yy:yy:yy:yy:yy (capab=0x431 status=0 aid=4)
[  129.065852] wlan0: associated
[  129.067151] ADDRCONF(NETDEV_CHANGE): wlan0: link becomes ready

That’s from the last boot with the generic/default kernel from #!. There’s no difference between that and what my vanilla kernel just did except sequence numbers.

Back to square one.

UPDATE 2: I booted into Windows (yea, it’s Patch Tuesday!) and then back into #! (which had its own Ubuntu patches today in addition to new kernels and my own upgrade of libmtp), and checked again. This is what I’d been seeing earlier — no attempt to authenticate with the other AP.

[   47.750281] wlan0: authenticate with AP yy:yy:yy:yy:yy:yy
[   47.751795] wlan0: authenticated
[   47.751805] wlan0: associate with AP yy:yy:yy:yy:yy:yy
[   47.754153] wlan0: RX AssocResp from yy:yy:yy:yy:yy:yy (capab=0x431 status=0 aid=4)
[   47.754168] wlan0: associated

Look, no xx:xx*. Just mine. Like it SHOULD do every fucking time.

Nothing like consistency, eh. Five boots with my own kernel and only one boot  did it try to associate with any other AP; four boots with kuki kernel and all four with extracurricular association attempts; every boot with the stock kernel associates with the other AP if it’s available.

I’m giving up for tonight.

continued MTP Issues – crunchbang and Ubuntu Jaunty Jacksh*t

Posted in MTP, Samsung S3, crunchbang, my stuff on July 14, 2009 by lucky

NOTE: Resolved, see update at end of article.

After doing a little hunting through Ubuntu-related forums, I double-checked the version of libmtp that came with crunchbang and saw it was version 0.3.0. I tried one of the older suggested “fixes” by setting a udev rule and adding myself to group audio. This gave me partial access to my Samsung S3.

What do I mean by partial? First, the device remained “unrecognized” using mtp-detect. That’s probably why it continued to not connect when running rhythmbox; I gave up on gpodder because it puts out sqlite-related errors every time I open it (saw that when tried launching from terminal). Second, I could send files to the device but they wouldn’t go to specified directories. The files went to the default files directory rather than to either the Music or Datacasts (podcasts) directory; I sent using the number code for the Datacasts directory and that didn’t work.

I decided to download and install the source tarball. This required installing gcc and related packages, as well as libusb-dev (compiling libmtp requires libusb headers). I used –prefix=/usr to build over the installed package and –program-prefix=mtp- as suggested in the documentation to prevent conflicts with other applications I may install later (e. g., sox). INSTALL also suggested –enable-hotplugging but this resulted in an error that it’s not a recognized option. Go figure.

After installing, I plugged in my S3 and fired up rhythmbox. I saw the S3 was available in the left panel and, once its files were read, I was able to browse it. While that’s a very good improvement, rhythmbox doesn’t show things in the S3’s file hierarchy but rather only as artist, album, etc.

rbox-S3-01

I just tried to send the same podcast to the Datacasts directory with mtp-sendfile and got a segfault. Great fuck, just what I need. I’m not asking for a miracle here. I just want to be able to at least manipulate things one way or another, and hopefully with enough control that I can put files where I want them and delete them when I want. I also looked to see if I can use rhythmbox to delete the podcast I copied last night (which went to the default files folder). Nope because it’s in the generic file folder rather than Music; I can still listen to it by navigating to the files folder. So now I have to see if I can figure out why the segfault, see if I can use mtp-tools to do it at all (did under Fedora), or see if there’s some way (via plugin?) to get rhythmbox to see the directory structure on the S3 rather than just what’s in Music.

Geez. Always fucking something, huh.

I’m not as pissed off at libmtp or Linux in general about this as I am Samsung for using MTP in the first place and peddling it as an ogg-friendly device. What a fucking nightmare this has been. First it didn’t really support ogg because of the firmware it came with and now it’s a pain in the ass getting it to work under non-Windows operating systems.

That’s it for today. Back to Windows to get some real work done. I can get more done without all these fucking headaches.

UPDATE 14 July 2009 @ 13:05 – Fixed by compiling libmtp 3.7 from Debian (INSTALL no longer mentions the hotplugging option for configure). No segfaults, deleted one old podcast and sent a test file to Datacasts without any errors:

% mtp-folders |grep Datacast                                                                                    
32773    Datacasts

% mtp-sendfile startx_old.ogg 32773                                                                             
libmtp version: 0.3.7

Sending startx_old.ogg to 32773
type:ogg,3
Sending file...
Progress: 14866 of 14866 (100%)
New file ID: -2147155907

I’m happier now. Wonder how long that’ll last if I decide to compile a custom kernel later while the All Star game is on.