On Fri, Nov 14, 2008 at 6:36 PM, Luis R. Rodriguez <lrodriguez@xxxxxxxxxxx> wrote: > On Fri, Nov 14, 2008 at 04:25:53PM -0800, Dan McGee wrote: >> On Fri, Nov 14, 2008 at 3:13 PM, Luis R. Rodriguez <mcgrof@xxxxxxxxx> wrote: >> > On Fri, Nov 14, 2008 at 12:41 PM, Luis R. Rodriguez <mcgrof@xxxxxxxxx> wrote: >> >> On Fri, Nov 14, 2008 at 12:33 PM, Bob Copeland <me@xxxxxxxxxxxxxxx> wrote: >> >>> On Fri, Nov 14, 2008 at 2:33 PM, Dan McGee <dpmcgee@xxxxxxxxx> wrote: >> >>>> On Fri, Nov 14, 2008 at 12:37 PM, Luis R. Rodriguez <mcgrof@xxxxxxxxx> wrote: >> >>>>> On Fri, Nov 14, 2008 at 10:33 AM, Sujith <m.sujith@xxxxxxxxx> wrote: >> >>>>>> Luis R. Rodriguez wrote: >> >>>>>>> > From the original log: >> >>>>>>> > ath5k_pci 0000:01:00.0: registered as '' >> >>>>>>> >> >>>>>>> That comes from >> >>>>>>> >> >>>>>>> dev_info(&pdev->dev, "registered as '%s'\n", wiphy_name(hw->wiphy)); >> >>>>>>> >> >>>>>>> which means that may be buggy too as wiphy_name() gets >> >>>>>>> >> >>>>>>> return wiphy->dev.bus_id; >> >>>>>>> >> >>>>>>> and we only have set at that point wiphy->dev.parent IIRC. >> >>>>>>> >> >>>>>> >> >>>>>> Nope, parent is set after that call in SET_IEEE80211_DEV() >> >>>>>> phy[x] gets assigned in alloc_hw(). >> >>>>>> >> >>>>>> So why is it NULL even if alloc_hw() succeeds ? >> >>>>> >> >>>>> Parent is wiphy->dev.parent not wiphy->dev though >> >>> >> >>> Sujith is right, wiphy->dev.bus_id should be set by the time that's printed, >> >>> since wiphy_new in alloc_hw fills it in. >> >> >> >> Ah, I see, yes, oh man. Well this is good :) >> > >> > OK then try the patch in the other thread. If that doesn't fix it then >> > it has to be another issue. >> >> Can someone point me to the "other thread"? I don't think I was CCed >> on it since I came late to the party. >> >> Someone also inquired about versions. I'm using stable kernel 2.6.27.6 >> and compat-wireless-2008-11-13 (10-28 was doing the same thing >> however). > > Bounced you a copy. Please give it a try. I gave it a shot. It worked (with some conditions attached). At first I built with just the patch, not enabling kernel debugging or anything, and I got a different NPE in some sysfs registration code deeper down in what I think was still the register_hw method. I can't fully remember, because it wasn't an easily savable stack trace and I immediately decided I would just rebuild everything with debug symbols. Of course, now that I have a kernel with full debug symbols, I can't get it to oops. Figures. I'll keep you updated if I see this error again. -Dan -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
