On Sun, Dec 20, 2009 at 7:21 PM, Thiemo Nagel <thiemo.na...@googlemail.com> wrote: > Generally, I'm getting system freezes sporadically, maybe every couple of > days. Since I've dug up one other ath5k-related soft lockup from my logs > (2.6.32-rc8), to my mind the simplest explanation would be that all the > other freezes were also due to ath5k.
Probably... I'm trying to track down one of those myself. > kernels from the 2.6.31 line, therefore I think the described behaviour > constitutes a regression. This, however, is a bit of a mystery. There weren't a lot of changes to ath5k in 2.6.32. You can probably bisect them with only a handful of kernel builds. The soft lockup trace you posted is useful, however we have rewritten that code for 2.6.33. I don't see how with the old code you can wait for 60s, but there is in ath5k_hw_noise_floor_calibration: - ath5k_hw_register_timeout() - which can spin for 1/3s (or more if udelay is broken) - the calibration loop, which takes at least 1 ms and up to 20ms. Together that's about half a second in the worst case, which is bad, but I can't yet see any way for counters to go negative or interrupt storms to re-trigger the tasklet or anything like that to make it take 60s. You might try latest wireless-testing to see if the rewritten code helps. -- Bob Copeland %% www.bobcopeland.com _______________________________________________ ath5k-devel mailing list ath5k-devel@lists.ath5k.org https://lists.ath5k.org/mailman/listinfo/ath5k-devel