Latest working kernel version: 2.6.27-7 Earliest failing kernel version: Distribution: Ubuntu Intrepid Hardware Environment: Apple Macbook Pro 2G (late 2006, Atheros AR5008, Core 2 Duo) Software Environment: Ubuntu Intrepid x86 32 bits Problem Description: Please have a look at the problem description on the madwifi web site, I submitted a bug here on the behalf of Madwifi team: http://madwifi-project.org/ticket/2178 Summary: On hardy and a madwifi early 2008 driver, my wifi connection speed on Internet is around 10 Mb/s, with ath9k provided "out of the box" by Intrepid it is less than 1 Mb/s, so my connection is pretty unusable and I downgraded back to Hardy. Steps to reproduce: Just use Intrepid on Macbook Pro 2G
This is for the upstream kernel -- please use the madwifi bug tracking system for madwifi driver problems.
(In reply to comment #1) > This is for the upstream kernel -- please use the madwifi bug tracking system > for madwifi driver problems. > Oh guys, please my your guidelines on the same line !! I first created a bug in madwifi and they told me to create it here because ath9k is part of the kernel. So what should I do to have some hope to be able to properly use Ubuntu 9.04? Where do I have to create the bug?
Sorry, I misread your report. Are you using an upstream kernel? Or one from Ubuntu? Did you report the problem in Ubuntu's bug tracking system? Can you recreate the issue using an upstream kernel?
I have the same issue a similar machine (Macbook Pro 3.1, Atheros AR5418) with Fedora 10, if that helps convince you it's a mainline kernel problem. I get slow throughput and frequently-dropped connections, especially under load (i.e., a download of more than a couple megabytes), on several access points, and I don't have these problems when using ath_pci instead. Any info I can provide that would help? If you still think it will help, I can probably build a vanilla kernel and give that a shot.
Please try these 3 patches: http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/ath9k/2008-11-22/27-IOMMU-01/ This may be the same issue seen on this bug: http://bugzilla.kernel.org/show_bug.cgi?id=11811
I'll give that a try later today.
(In reply to comment #3) > Sorry, I misread your report. Are you using an upstream kernel? Or one from > Ubuntu? Did you report the problem in Ubuntu's bug tracking system? Can you > recreate the issue using an upstream kernel? > No problem, yes I guess it's an upstream kernel, the on provided by Ubuntu Intrepid, no change, just used it "out of the box".
> This may be the same issue seen on this bug: > > http://bugzilla.kernel.org/show_bug.cgi?id=11811 > To me, no, Luis. I didn't experience a single kernel crash, just ssssllllllloooooooooowwwwwwwwwwwwww connection speed, about 10% of the normal speed.
It's not an upstream kernel if it is the one "out of the box" from Ubuntu.
Well to be fair as far as I can tell ath9k and mac80211 in Ubuntu 8.10 (Intrepid) are intact from the stock kernel. Even if you have not seen any crashes please try the patches supplied, it fixes some serious DMA issues seen only on MacBook Pros.
> Well to be fair as far as I can tell ath9k and mac80211 in Ubuntu 8.10 > (Intrepid) are intact from the stock kernel. Not to mention the fact that I'm getting it on Fedora as well. No crashes, though. I grabbed the latest vanilla sources and built ath9k with those DMA patches. It loaded and seems to be working fine for now. I'll use it for a few days and report how well it worked. Interesting how it only seems to occur on Macbook Pros.
(In reply to comment #10) > Well to be fair as far as I can tell ath9k and mac80211 in Ubuntu 8.10 > (Intrepid) are intact from the stock kernel. > > Even if you have not seen any crashes please try the patches supplied, it > fixes > some serious DMA issues seen only on MacBook Pros. > Luis, I need to use my computer, so to me, it's just not possible to use Intrepid if I can't properly use my internet connection. That's why I downgraded back to Hardy and I won't reinstall my whole computer just to verify if this fixes the bug or not. I'm sorry, to me it's a completely unacceptable situation, the madwifi driver was give up to a clearly not fonctional driver, not understandable in my opinion. > It's not an upstream kernel if it is the one "out of the box" from Ubuntu. That is the illustration it is pretty impossible for an average user to rise a problem on Linux, because upstream is not part of the language of such people. So what do I have to do with this issue? > It loaded and seems to be working fine for now. Garrett: is your connection correct?
Well, thanks for reporting the problem. Of course, refusing to test potential solutions limits the usefulness of your report. Thankfully Garrett was available in this case for testing. Garrett's report suggests that the issue will be fixed in 2.6.28. Perhaps you can persuade the Ubuntu kernel team to take those patches now so that you can go back to Intrepid, or you can wait until Intrepid has a 2.6.28 kernel -- I have no idea how Ubuntu's release cycles work, so I don't know when that will be. As for the "clearly not functional driver" garbage, please give it a rest. Madwifi was never in the kernel anyway for a variety of reasons. Besides, just because there is a problem on the box you happen to have doesn't mean that the driver doesn't work well for thousands of other users.
Just 1 additional comment for people who find this bug: The slow performance is not just a Macbook Pro issue. I have an HP G60-123CL with an Atheros AR928x running XUbuntu Intrepid. According to the connection information, I too am running at only 1Mb/s. I have not tried the patches because I haven't had time to learn how to patch a stock kernel from XUbuntu. As for the kernel version, again just FYI, it's running 2.6.27-9. If I have time to try the patches or XUbuntu releases a 2.6.28 kernel, I'll post an additional comment with results.
Thanks for your feedback David, but don't rely on the Network-Manager speed display, do test your connection speed with one of these site: http://test-debit.free.fr/debit/ http://www.degrouptest.com/test-debit.php They are in French, but easy to understand, click on "Démarrer" or "Submit". A question for people who "know": Let's say the bug is fixed, when will the fix be published in the upstream kernel? can I hope a new kernel upgrade in my favorite distro (Ubuntu)? Thanks.
Same performance problems on gentoo-sources-2.6.27-r7 with ath9k from the kernel (iwconfig shows 1MB/s and ofer that links http://test-debit.free.fr/debit/ 8MB/s) mac80211 was build in the kernel and ath9k was built as a module. Channel 12 and 13 is not available. After that i tried to build both as a module (mac80211 and ath9k) (/etc/modporbe.d/options to mac80211--possible solution to get channel 12 an 13 to work) mac80211 failed. I have also tried these patches but the doesnt change the situation. ???? Please describe how can i use these patches or ??? should i use the module from the git or madwifi ??? Thanks for help
If you want bleeding edge drivers you can try compat-wireless: http://wireless.kernel.org/en/users/Download
Dear all, This little message to say I've successfully upgraded my Mac to Intrepid, including the last kernel updates and the wifi seems to work pretty well, with acceptable performances. I'll post additionnal information about wifi bandwidth as soon as I can make some tests, but Intrepid is now my default environment thanks to your efforts. Thanks for all.
Dear all, I do confirm the bug is fixed, but I'm going to create a new bug report about the bad stability of the wireless connection using this improved ath9k. Thanks to all.
Dear all, I'm sorry to reopen this bug report, but I've upgraded my Mac to Jaunty and the situation is pretty similar as the previous one on Intrepid. The kernel is 2.6.28-11 On my wife computer (Dell Latitude D620 with a Broadcom Wifi chipset, Jaunty 32 bits) I can achieve a download rate around 12800 kb/s. On my Mac (Jaunty 64 bits), I can not get more than 4900 kb/s. So please, make this driver work well.
This is obviously my mistake, my apologizes. The bug is well fixed.
Can this please be closed?