acx-mac80211: 2.6.32, testable version avail

2010-01-03
2013-02-17
<< < 1 2 3 > >> (Page 2 of 3)
  • Joachim Frieben

    Joachim Frieben - 2010-04-11

    I have had no problems using 1.2.1.34 or 2.3.1.31. There is no point in using an outdated version like 0.4.11.9.

     
  • Hubbitus

    Hubbitus - 2010-04-11

    I can't find old firmware. With current it still got error and fully freeze system:
    Apr 11 14:03:37 x-www kernel: irq event 16: bogus return value f62e0b48
    Apr 11 14:03:37 x-www kernel: Pid: 0, comm: swapper Tainted: P           2.6.32.10-44.fc11.i586 #1
    Apr 11 14:03:37 x-www kernel: Call Trace:
    Apr 11 14:03:37 x-www kernel:  __report_bad_irq+0x33/0x74
    Apr 11 14:03:37 x-www kernel:  note_interrupt+0x5c/0x153
    Apr 11 14:03:37 x-www kernel:  handle_fasteoi_irq+0x83/0xa2
    Apr 11 14:03:37 x-www kernel:  handle_irq+0x40/0x4d
    Apr 11 14:03:37 x-www kernel:  do_IRQ+0x46/0x9a
    Apr 11 14:03:37 x-www kernel:  common_interrupt+0x30/0x38
    Apr 11 14:03:37 x-www kernel:  ? acpi_idle_enter_simple+0x112/0x142
    Apr 11 14:03:37 x-www kernel:  cpuidle_idle_call+0x72/0xc3
    Apr 11 14:03:37 x-www kernel:  cpu_idle+0x96/0xb0
    Apr 11 14:03:37 x-www kernel:  rest_init+0x58/0x5a
    Apr 11 14:03:37 x-www kernel:  start_kernel+0x330/0x335
    Apr 11 14:03:37 x-www kernel:  i386_start_kernel+0x9e/0xa5
    Apr 11 14:03:37 x-www kernel: handlers:
    Apr 11 14:03:37 x-www kernel:  (sil_interrupt+0x0/0x209 )
    Apr 11 14:03:37 x-www kernel:  (io_irq_isr+0x0/0x3d )

     
  • Oliver Winker

    Oliver Winker - 2010-04-11

    Hi,

    Hubbitus, could you sent a full log from loading of driver until crash with following steps:

    - insmod
    - ifup
    - Several times: iwlist  scan
    --
    and if you feel like, the same with:

    - insmod
    - echo "0xff" >/proc/driver/acx0/acx_debug
    - ifup
    - Several times: iwlist  scan

    This will give a bigger log file.

    You can email the logs my latest email address in the git log  Signed-off-by.

    Make sure, that the interface is not automatically brought up before. E.g. disable existing network configs for the interface.

    It may still be interesting to see, if another card/driver/firmware combination shows the same behaviour.  The other firmware versions can be found in the tar.bz2 file mentioned on the firmware page.

    Let's see what this gives then … .

    Joachim, could you tell, which card you are using exactly ?

    I'll also try one time a newer fw with my DWL-G650+ for curiosity.

    Cheers, Oliver

     
  • Hubbitus

    Hubbitus - 2010-04-18

    Hello, oli1417.
    Firstly very thanks for help and work.

    Sorry for the delay with answer. This tests required many time and reboots and it was leaved to weekends.

    Now I perform many tests with all firmware from  http://acx100.erley.org/fw/
    For that wrote some scripts, included in archives.

    All results may be downloaded:
    http://ru.bir.ru/_/BUGS/acx/logs(nodelay).zip
    http://ru.bir.ru/_/BUGS/acx/logs(sleep).zip

    Some notes about it.
    1) I have no any kernel trace. I suppose it was from ndiswrapper conflict. Off course it was disabled and I do not known how. Now ndiswrapper was fully removed (renamed) for clear results.
    2) For clear results also after *each* test computer had been rebooted.
    3) All results in table set by best value. For example, if in 3 times do "iwlist wlan0 scan" at least once I seen 2 cell, it marked as "full". If 1-2 times appeared as "no scan results" and then 1 only self cell, it is marked as "self" and "no" otherwise. Similar with ping - if at least 1 ping successful - it marked as "yes".

    Some my general conclusions and why two test cases present:
    1) There no general differences (speed and similar things is not tested at all) how firmware used.
    2) Module unload is not work. What may be interesting, if I try "rmmod acx_mac80211" or "modprobe -r acx_mac80211" commands never returned. Furthermore, in this case reboot also not enough. Only hard reset helps. But if do not try remove it manually, just type reboot - machine rebooted successfully. It also was tested in each firmware.
    3) Accidentally found very interesting behaviour. In test script (see in archive with delay) I insert sleep commands before scan nets and before ping. I suppose there need time to establish connection… But real is absolutely different! If immediately after module insert I ifup interface it work!!! It work on almost all firmwares (see second test results). It does not work after some delay, it also does not work after interface restart!

    With hope it helps.
    Ehh, again step back to ndiswrapper…

     
  • Oliver Winker

    Oliver Winker - 2010-04-18

    Hi Hubbitus,

    Thanks a lot for the detailed logs. I'll have a look. I'm also running 1.2.1.34 in the meanwhile, without problems.

    G+, Oliver

     
  • Oliver Winker

    Oliver Winker - 2010-04-18

    Hello Hubbitus,

    You did really very detailed tests - great work ;)! Thanks a lot again. I'll put these logs in my own log reference ;).

    In fact it's strange, that the driver want's to update the beacon (acx: phy0: issue_cmd(cmd:ACX1xx_CMD_CONFIG_BEACON) FAILED).

    In my logs, that doesn't happen, and I think for STA mode a beacon is also not required.

    In fact, in my instructions I made a little mistake (sorry) … let's try the following.

    1) Check in /etc/network/interfaces (debian), that there is no entry for wlan0. Comment existing ones.

    2) Check, that you don't run hostapd

    3) insmod the module
    … apart from the driver starting, nothing more should happen now! Below the last lines that should be logged:

    Apr 18 17:43:46 lapix kernel: [19730.866173] acx: acx v0.5.0: net device phy2, driver compiled against wireless extensions 22 and Linux 2.6.33.2-ow1-bigmem
    Apr 18 17:43:46 lapix kernel: [19730.921608] udev: renamed network interface wlan0 to wlan3
    

    4) Now do: ifconfig wlan0 up

    5) Now do some scans again: iwlist wlan0 scan

    6) Start wpa_supplicant in standalone mode:

    wpa_supplicant -t -d -K -i wlan0 -D wext -c wpa0.conf
    

    with wpa0.conf containing something like this:

    network={
            ssid="Your SSID"
            scan_ssid=1
            key_mgmt=WPA-PSK
            psk="Your Passphrase"
    }
    

    7) That should connect to the wpa AP … and normally I think there should be no config of beacons. But I'll anyway note this TBC (To Be Checked)

    Regarding the two other findings of your tests:

    1) Module unload while ifup:

    That's a known problem, which I was just looking at this WE. rmmod while interface up, doesn't work. That's a bug. You need to ifdown first.

    2) ifup/ifdown/ifup is also known to not always work. That is also in my list.

    G+, Oliver

     
  • Oliver Winker

    Oliver Winker - 2010-04-18

    'Module unload while ifup' should be fixed now.

    G+, Oliver

     
  • Hubbitus

    Hubbitus - 2010-04-18

    1) I use Fedota, not Debian, but for insurance what there no wlan0 interface in script kernel module renamed!
    2) What is hostapd?
    3) Do you want I test it separately and separately write into log?
    4) "ifconfig wlan0 up" or "ifup wlan0"? You intentionally do not want apply some interface settings on this stage?
    6) No! There full open net now, until without ANY encryption.

    In fact it's strange, that the driver want's to update the beacon (acx: phy0: issue_cmd(cmd:ACX1xx_CMD_CONFIG_BEACON) FAILED).

    I even don't known what it is :)

    'Module unload while ifup' should be fixed now.

    If I correctly understand it fixed in git and I can update to try?

     
  • Oliver Winker

    Oliver Winker - 2010-04-19

    Hi Hubbitus,

    An AP sends regularly a radio beacon, like a light-house, to signalize that it's there (the details are probably in the IEEE 802.11 spec ;).

    Since in your logs the driver tries to configure beacons, I was wondering, whether your card tries maybe also running as AP, e.g. using hostapd ?

    "ifconfig wlan0 up": Indeed, that would brings the interface up, without any settings. And the driver should not do anything in the logs, until this cmd. If it does something before, then something else is operating the driver (some other daemon e.g.).

    Once the interface is up, you should be able to do scans.

    And you should be able to associate with an open net using "iwconfig wlan0 essid "My Network"", but that you need to test a bit (I mostly tested the WPA case, WEP I tested and worked).

    In general, what I would try, is to step by step bring up the link and approach the point, where the trouble starts.

    Cheers, Oliver

     
  • Anonymous - 2010-04-19

    Hey, has there been any progress made on the driver for the hx4700? Can i get the patch or source for it or the hx4700

     
  • Hubbitus

    Hubbitus - 2010-04-20

    > An AP sends regularly a radio beacon, like a light-house, to signalize that it's there (the details are probably in the IEEE 802.11 spec ;).
    I use Ad-Hoc configuration, no AP. Other side (192.168.0.2 in test) is my notebook.

    What is beacon? And as I say before, I do not use hostapd.

    And I try test it directly:
    1) ifconfig wlan0 up
    2) iwconfig wlan0 essid "My Network"

    What you say about "no-delay once work" phenomena? Have it worth to test with and without delay?
    If you say module unload now should work, is it enough to unload module before turn to next test? Or full reboot preferred?
    I plan test only 2 last firmwares in my list. Please say if you want see results with some additional too.

     
  • Hubbitus

    Hubbitus - 2010-04-20

    Sorry. I will try test it in day or two.

     
  • Oliver Winker

    Oliver Winker - 2010-04-20

    Hi https://me.yahoo.com/a/9JZM._sQ,

    I see you also hit the same SourceForge Openid bug as me … without trailing slash, they log you in as an URL !?

    For the hx4700:

    Since the start of this discussion thread, see beginning, the driver also supports the hx4700. And since then there have been also some other smaller improvements on this side. Just checkout the latest commit.

    However you'll need an >=2.6.32 kernel on the hx4700. I'm currently using an own constructed kernel for this.

    G+, Oliver

     
  • Oliver Winker

    Oliver Winker - 2010-04-20

    Hi Hubbitus,

    Ahh Ok - you are setting up an adhoc network! That could explain the beacon and related problems. I have not yet tested adhoc myself … that's then maybe something to add on the TODO list. 

    G+, Oliver

     
  • Oliver Winker

    Oliver Winker - 2010-04-20

    … and for the firmware: just go continue with the 1.2.1.34 . That doesn't make a difference then it seems.

     
  • Hubbitus

    Hubbitus - 2010-05-08

    Ok.
    I again there with new test results. It is today git build.
    Firstly - very very thanks, at least now module unload work and I shouldn't restart computer each time.

    But other think seams same. If I do scan network just after modprobe - it works. I cant' say exactly time frame, but if I sleep at least 5 seconds there no more chance see it worked.
    Meantime, there one enhancement too - if first run just after insmod, then it possible then ifdown wlan0; ifup wlan0 and it works. It wasn't work in previous test.

    There one more issue. Even if I fast start it and it start working it work not many time. After some minutes of downloading it became unavailable and only module unload/load may help. For this case i include logs with "continuous" in suffix.

    All logs (with 1.2.1.34 firmware all) you can download there: http://ru.bir.ru/_/BUGS/acx/acx.logs.20100508.tar.xz

    Warning: continuous logs with debug is very big, in uncompressed state it is around 1.5Gb. If you wishm I can repack it and upload separately.

     
  • Oliver Winker

    Oliver Winker - 2010-05-08

    Hi Hubbitus,

    Thanks for the feedback. Indeed the module unload was improved.

    Are you still running in Ad-hoc mode ?

    If yes, then there two news: 1: I can confirm, right now Ad-hoc mode is not working - or at least not as it should. For Ad-Hoc mode the for card needs template and beacon support. And that code currently is badly broken. 2: I'm busy fixing this part. And I think there are good chances, that if it works for AP mode, Ad-hoc will benefit as well.

    I'll keep updated. Cheers, Oliver

     
  • Hubbitus

    Hubbitus - 2010-05-08

    Yes, it computer act as router (AP), so only Ad-Hoc used.

    Thank you, I wait finishing work.

     
  • Anonymous - 2010-05-15

    I'm really glad there's some development going on here.

    I've got a D-Link DSL-G604T access point running OpenWRT. I'd really like to get the wireless working on it.

    I've been trying with the acx and acx-mac80211 drivers, and I haven't had any luck.

    I'd really like to connect to a WEP network, and it seems to associate, but I never get anything through it.

    I can provide logs if you're interested.

    cheers

     
  • vkchk

    vkchk - 2010-05-16

    Hi everybody!
    Happy to see some development activity at acx100. Thanks a lot, Oliver.

    Couple of month ago i've obtained two D-Link DWL-G520+ cards and now i'm setting up a router using one of this cards, old AMD K7 box and Archlinux. So, i need to make DWL-G520+ function in Master mode (preferred) or Ad-hoc mode (less preferred) with WPA (or WPA2, if possible) encryption.

    I have tested several configurations and none of them worked properly.
    1. Old acx-20080210 driver and 2.6.27 kernel. hostapd seems to have no support for the driver. wpa_supplicant crashes.
    2. Old acx-20080210 driver and 2.6.33 kernel. Driver compiles against 2.6.33 with some set of patches but crushes the kernel upon 'iwconfig'.
    3. ndiswraper and  2.6.33 kernel. Works perfectly in ad-hoc mode with WEP encryption, but when i'm starting wpa_supplicant (to enable WPA) no data could pass through the wifi link. I guess debugging windows driver is pointless.

    It seems that my last chance to use make use of those wifi cards is the new acx-mac80211 driver :)
    It's not a problem to switch to 2.6.32 kernel, so i wonder when the driver would be ready for testing with Ad-hoc / WPA-NONE configuration?

     
  • Oliver Winker

    Oliver Winker - 2010-05-16

    Hi Andy,

    We are currently working with OpenWRT on support for the driver, and it's also making progress. WEP I didn't check, but the good news is experimental WPA AP with an WAG54G already worked (but setup is unstable still), and WEP should then also be feasible. So it's basically work in progress.

    I'm currently on vacation and don't have regular internet access. So follow up will take a bit longer.

    Cheers, Oliver

     
  • Oliver Winker

    Oliver Winker - 2010-05-16

    Hi vkchk,

    The latest version of the mac80211-mem of the repo I'm maintaining has master mode enabled for pci cards.

    But it's still experimental and there are problems with the WPA connection setup with one card I used for tests (wpa_supplicant had to be restarted several times before successful assoc). But it was already possible to setup a WPA network with two STAs.  … so basically AP is currently under dev ;).

    As already mentioned above, I'm currently on vacation and don't have regular internet access. So follow up will take a bit longer.

    Cheers, Oliver  

     
  • Anonymous - 2010-05-17

    Thanks Oliver!

    If you need any logs or anything, let me know.

    cheers

     
  • Oliver Winker

    Oliver Winker - 2010-06-01

    Hi everyone,

    Small update on the oli1417-clone/mac80211-mem branch, current commit is 74b15d:

    * AP should work with ACX111/100 Pci and Vlynq cards. I tested it on

    - PCI: DWL-650+, DWL-G650+, 2.6.34, 32bit
    - Vlynq: WAG54G with OpenWRT (see the OpenWRT svn trunk acx-mac80211 package, see commit inside Makefile)

    … I observed, that with my ath5k card EAPOL handshake is sometimes a bit flaky (don't know why yet), but it usually comes to an end after some retries. All my other STAs associate quite fast and reliably.

    * ADHOC should also work now with ACX111/100 Pci cards, 32bit - … however tested only unencrypted.

    - I tried ADHOC WPA with wpa_supplicant, and it also seems to setup and configure encryption etc, but still I couldn't get a packet though !? No crash however, just no transmission.

    I specified 32bit, because we observe issues on 64bit. That's currently under investigation. Awaiting a DWL-G520+for this, to do some own tests under amd64.

    In general data rates are currently only b-rates. I'm currently busy with some infrastructure work on the driver, which was on the todo list since quite some time. As soon as that's done, also the rates will be tackled. 

    Cheers, Ol

     
  • Hubbitus

    Hubbitus - 2010-06-03

    Awesome! It seams work in Ad-Hoc mode!!
    Very-very thank you Oliver!

    By the way, in log indefinetly repeated errors like:
    Jun  3 22:32:49 x-www kernel: acx: acx_e_after_interrupt_task: Schedule CMD_RADIO_RECALIB
    Jun  3 22:32:49 x-www kernel: acx: acx_s_after_interrupt_recalib: phy1: less than 5 minutes since last radio recalibration, not recalibrating (maybe the card is too hot?)
    Jun  3 22:32:51 x-www kernel: acx: phy1: tx error 0x20, buf 15! (excessive Tx retries due to either distance too high or unable to Tx or Tx frame error - try changing 'iwconfig txpower XXX' or 'sens'itivity or 'retry')                                                                                                                                                                     
    Jun  3 22:32:53 x-www kernel: acx: phy1: tx error 0x20, buf 09! (excessive Tx retries due to either distance too high or unable to Tx or Tx frame error - try changing 'iwconfig txpower XXX' or 'sens'itivity or 'retry')                                                                                                                                                                     
    Jun  3 22:32:53 x-www kernel: acx: phy1: tx error 0x20, buf 03! (excessive Tx retries due to either distance too high or unable to Tx or Tx frame error - try changing 'iwconfig txpower XXX' or 'sens'itivity or 'retry')                                                                                                                                                                     
    Jun  3 22:32:53 x-www kernel: acx: acxpci_handle_tx_error: phy1: several excessive Tx retry errors occurred, attempting to recalibrate radio. Radio drift might be caused by increasing card temperature, please check the card before it's too late!                                                                                                                                          
    Jun  3 22:32:53 x-www kernel: acx: acxpci_handle_tx_error: disabling above message
    Jun  3 22:32:53 x-www kernel: acx: phy1: tx error 0x20, buf 08! (excessive Tx retries due to either distance too high or unable to Tx or Tx frame error - try changing 'iwconfig txpower XXX' or 'sens'itivity or 'retry')

    how it majority?

    And as usual, if someone interested, I pack it (today git checkout) in RPM for Fedora - http://hubbitus.net.ru/rpm/Fedora11/dkms-tiacx/noarch/

     
<< < 1 2 3 > >> (Page 2 of 3)

Get latest updates about Open Source Projects, Conferences and News.

Sign up for the SourceForge newsletter:





No, thanks