Latest working kernel version:2.6.28.7 + Latest wireless drivers Earliest failing kernel version: 2.6.28 Distribution: Archlinux Hardware Environment: Acer Aspire One Software Environment: ArchLinux Problem Description: When downloading files larger than 12MB I find that they are corrupted, error does not occur when using my ethernet port. I have tried both wireless at university and at home and corruption occurs. Does not occur with files smaller than 12MB though. I have confirmed that the source files are not corrupted as I can download and install them using ethernet. Steps to reproduce: Run pacman or powerpill - download and watch it fail.
Are you saying that the problem exists with 2.6.28, but not with 2.6.28 + compat-wireless?
sorry, when I thought it said, "latest working kernel' it was 'the latest kernel which you experienced problems with". The problem occurs with the latest stable kernel and with the latest wireless drivers as well.
Please attach the contents of /var/log/messages, in case that has something useful.
John, I can't attach it because Bugzilla states that my message file is too big. Having had a look through kernel.log there seems to be alot of: kernel.log:Mar 10 10:42:05 rachmaninov ath5k phy0: noise floor calibration timeout (2432MHz) and in message.log Mar 7 00:31:20 rachmaninov ath5k phy0: unsupported jumbo which seems to chime in when I'm downloading something large.
So the data being transferred actually gets corrupted, or you just have connection problems? Bugs 12068/12080 relate to unsupported jumbo but no one has yet reported data corruption.
RE: Bob Copeland 1) I uninstalled the Madwifi package and now only ath5k is loading. 2) Corruption isn't occurring, however, there are constant drop outs. Drop occur even when Enable Frame Bursting is disabled. 3) This dropping out occurs after suddenly 'drops' in connectivity - that is, it is still connected but I can't transfer anything. This occurs a couple of times then it disconnects, it fails to reconnect with the saved password, I end in the password again and it makes a connection. Rinse and repeat. 3) I have since disabled ath5k and built madwifi from madwifi-hal-0.10.5.6-r3942-20090205.tar.gz and it has been rock solid; no corruption, no drop outs, no problems. These problems occur when connected to the university network as well as my own one. 4) the corruption I have linked back to the ath_pci and ath_hal loading when ath5k loaded. When only ath5k loads the corruption stops but the connection is fragile 5) Before disabling it was unstable and corruption was experienced, now it is just unstable. I've gone back to using madwifi because the ath5k is too unstable and unreliable.(In reply to comment #5) > So the data being transferred actually gets corrupted, or you just have > connection problems? Bugs 12068/12080 relate to unsupported jumbo but no one > has yet reported data corruption. >
just to correct: Frame disabling is on my router at home These problems occur when connected to the university network as well as my own one is in regards to a problem using ath5k. The problem doesn't occur using the madwifi driver.
Can you try the patch from here: http://bugzilla.kernel.org/attachment.cgi?id=20240&action=view
(In reply to comment #8) > Can you try the patch from here: > http://bugzilla.kernel.org/attachment.cgi?id=20240&action=view > I've patched it and going to give it a test run; quick question, why hasn't this patch been accepted yet given that it is over 3 months old?
Because it hangs some machines.
(In reply to comment #10) > Because it hangs some machines. > I got so pissed off last night with Madwifi now corrupting large downloads that I've pulled Linux off the machine and put Windows XP back onto it. In all due respect, if this was occurring on Windows there would be hell to pay - yet nothing is being done in the 'great world' of opensource.
Ok, well this is not productive, so I'll mark it a duplicate of 12080.
*** This bug has been marked as a duplicate of bug 12080 ***