Description of problem: The airo MPI350 in my IBM Thinkpad T41 laptop stopped working on my laptop with recent rawhide kernels and the 5.00.x and latest 5.30.x firmware versions. The last RH kernel tested to work properly was release 397. After that point the "RFMON" fixes went into the kernel. notting however said that it works on his firmware right now. Notting's Firmware ================== Firmware Version: 5b00.08 Warren's Laptop with kernel-2.6.6-1.397 ======================================= [root@ibmlaptop eth0]# pwd /proc/driver/aironet/eth0 [root@ibmlaptop eth0]# cat Status Status: CFG ACT SYN LNK PRIV KEY WEP Mode: 3bf Signal Strength: 87 Signal Quality: 26 SSID: spacedog AP: Freq: 0 BitRate: 11mbs Driver Version: airo.c 0.6 (Ben Reed & Javier Achirica) Device: 350 Series Manufacturer: Cisco Systems Firmware Version: 5.00.03 Radio type: 2 Country: 0 Hardware Version: ff Software Version: 500 Software Subversion: 3 Boot block version: 159 Note that this proc directory is empty on the 422 kernel for me. Attaching dmesg next.
Created attachment 100947 [details] airo-422.txt dmesg with 422 kernel showing airo failure.
Created attachment 100948 [details] airo-397-working.txt dmesg from 397 kernel showing working behavior prior to RFMON patch.
Created attachment 100972 [details] reenable card reset when module is loaded On my laptop (T40), it appeared that the initial card reset was not needed, but this was certainly a bad idea, if it breaks on other hardwares ? Does this patch helps ? I also removed the card reset in airo_pci_resume() too, invoked when coming back from APM/ACPI suspend, because I observed that the card could no longer associate when coming back from Monitor to Managed mode, after the reset command was issued (in airo_pci_resume()). Moreover, removing this reset didn't generate bad consequences in my case, but YMMV. So, if this patch works, does the card properly resume from ACPI and from APM suspend too ? (it should not be necessary to ifdown/ifup the interface in the apm scripts)
I made some tests, and discovered that this buggy behaviour is triggered by kudzu, and resolved by the above patch. If you don't see other problems, I'll submit it to Javier.
The bug should be fixed in 2.6.7-bk3.
Thanks. Our latest released kernel reverted to the old airo, but we will be sure that your latest airo driver is tested with our rawhide kernels.
kernel-2.6.7-1.448 seems to work on all firmware versions without problems. Great work Fabrice! =)
Hm, I have a Thinkpad R40 with a Mini PCI 350 (Firmware 5b00.08), which on Fedora Core 2, which works with kernel-2.6.5-1.358 and does not work with kernel-2.6.7-1.494.2.2. The latter will not even allow to unload the module again.
Oh, and I have kudzu disabled during startup. Regards Mirko
I tested with firmware 5.00.03 and kernel-2.6.7-1.494.2.2, and it works for me (it associates to the AP, with WEP enabled). Unfortunately, firmware 5b00.08 isn't available from Cisco web site, so I cannot use this exact version. Do you have any related errors in the logs ?