| |||||||||||||
|
| |||||||||||||
Ubuntu 8.04.1 LTS releasedUbuntu 8.04.1 LTS releasedPosted Jul 4, 2008 17:41 UTC (Fri) by callegar (guest, #16148)Parent article: Ubuntu 8.04.1 LTS released
Sadly, the random kernel freeze that have afflicted hardy since its release and that are fixed upstream (see http://lwn.net/Articles/287840/) do not appear to be fixed or mentioned in the release notes of this maintenance release. Since those who are already users of hardy have most probably already received all the updates automatically, this release is probably most relevant to those who are not already hardy users. Having arrived at a maintenance release may give them a strong feeling of reliance and really encourage them to upgrade now. This is generally a good thing. However, it is unfortunate that the release notes do not mention the hard lockup problem cited above. Regardless of how infrequent it may be, it is now well documented on the ubuntu bug traker, acknowledged by the kernel developer and discussed on the forums (see the above link and comments therein for details). As a general recommendation, before upgrading I would suggest everybody to play with hardy from the live CD for at least a few (7-8) consecutive hours at high loads (including disk and network activity) to make sure that their platform is not affected by the hard lockup issue. Indeed, after having upgraded it is difficult to go back and the hard freeze problem may result in data loss. Typically, the random freeze manifests itself every few hours, so 7-8 hours without lockups should mean that your platform is unaffected and can safely receive a hardy installation. For those that might have installed hardy (either the original hardy or the maintenance release) discovering only too late that their platform is affected by the freeze, please consider that: 1) The bug appears to be fixed upstream, so any custom compilation of a 2.6.25 or higher kernel from kernel.org should fix the issue. 2) The -rt kernel comprised in ubuntu hardy (in opposition to the default -generic one) pays some performance penalty but appears not to be afflicted by the bug. I really hope that the above notes are not taken as an attack to the Ubuntu teams, but as constructive and helpful comments for users and as an incentive to finally bring to resolution an issue that I regard as maybe infrequent by very serious.
(Log in to post comments)
Ubuntu 8.04.1 LTS released Posted Jul 5, 2008 2:17 UTC (Sat) by madscientist (subscriber, #16861) [Link] I upgraded my father's laptop from Dapper to Hardy last week. After upgrading, just a few seconds after he logged in the laptop would simply turn itself off. If you turned it back on it would come right back up (didn't even need to fsck the disks, but it did really boot; it wasn't coming back from hibernation or anything). I just booted the Dapper kernel and that still worked fine with the Hardy userspace. Then I installed the Feisty and Gutsy kernels as well, and discovered that while Feisty worked, Gutsy also caused the system to turn off. I set the grub menu to always boot the Gutsy kernel and he was happy. That's a lot easier, IMO, than building your own kernel. Of course you don't have the newest kernel, but the older one worked fine for all those months so... Just an alternative idea.
Ubuntu 8.04.1 LTS released Posted Jul 5, 2008 9:18 UTC (Sat) by Cato (subscriber, #7643) [Link]
This doesn't always work though... I've had a different freeze bug ("IRQ nn, nobody cared" -
https://bugs.launchpad.net/ubuntu/+source/linux-source-2....) on Feisty, Gutsy
and Hardy kernels - the only cure is to run with "irqpoll".
Many non-techies would not be able to diagnose what was happening here, so I think this sort
of kernel bug will slow Linux and Ubuntu adoption on the desktop. Maybe the Ubuntu kernel
team is under resourced, but I get the feeling that stability bugs aren't always the high
priority they should be.
resource Posted Jul 10, 2008 13:38 UTC (Thu) by alex (subscriber, #1355) [Link] <em>"Maybe the Ubuntu kernel team is under resourced"</em> Certainly compared to a company like RedHat which has a fair number of high profile kernel hackers. I'm not sure how big the core distro team is but I get the sense a lot gets done on a shoestring, building as they are able to do on the groundwork done by Debian. I think RedHat also test their stuff on a wider range of hardware.
Ubuntu 8.04.1 LTS released Posted Jul 5, 2008 9:19 UTC (Sat) by Cato (subscriber, #7643) [Link]
This doesn't always work though... I've had a different freeze bug ("IRQ nn, nobody cared" -
https://bugs.launchpad.net/ubuntu/+source/linux-source-2....) on Feisty, Gutsy
and Hardy kernels - the only cure is to run with "irqpoll".
Many non-techies would not be able to diagnose what was happening here, so I think this sort
of kernel bug will slow Linux and Ubuntu adoption on the desktop. Maybe the Ubuntu kernel
team is under resourced, but I get the feeling that stability bugs aren't always the high
priority they should be.
Ubuntu 8.04.1 LTS released Posted Jul 6, 2008 22:05 UTC (Sun) by gavinmc (guest, #46316) [Link] that sounds rather like this issue: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/75704 Is it an L400? Did you try disabling ACPI? Gavin
Ubuntu 8.04.1 LTS released Posted Jul 7, 2008 13:06 UTC (Mon) by madscientist (subscriber, #16861) [Link] No, it's a Toshiba and it doesn't go into suspend: it actually turns off completely. See: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/242400 Also, it actually works with 6.06, 6.10, and 7.04. It fails with 7.10 and 8.04. I didn't try disabling ACPI, which is a good idea in any case, but it'll have to wait for a month until my father comes back into town since I don't have access to the system. My dad's a long-time computer user/etc. but... I don't feel comfortable trying something like this long-distance.
Ubuntu 8.04.1 LTS released Posted Jul 5, 2008 9:32 UTC (Sat) by tajyrink (subscriber, #2750) [Link] > However, it is unfortunate that the release notes do not mention the > hard lockup problem cited above. Regardless of how infrequent it > may be, it is now well documented on the ubuntu bug traker, > acknowledged by the kernel developer and discussed on the > forums (see the above link and comments therein for details). Which specific bug was the one that was well documented and understood? You're linking to ca. 10 bugs in those linked comments, most of which are quite vague and have no clear idea about what's going wrong. If the bug is well understood, there is probably a patch for 2.6.24 doable for it, but so far the only "fix" was to upgrade to 2.6.25, which is just a shot in the dark.
Ubuntu 8.04.1 LTS released Posted Jul 5, 2008 13:57 UTC (Sat) by callegar (guest, #16148) [Link] Please, don't make me say what I am not actually saying: I have never suggested that the bug is well understood. Indeed if it was, it would be fixed already. Suggesting the contrary would mean believing that it is left there on purpose, that I surely don't. I am only saying that: 1) the bug /existence/ is well documented as proved by the many (often duplicate) bug reports. And although these do not allow one to count the statistical incidence of affected systems, surely it says that it is > 0%. 2) the bug is acknowledged by the ubuntu kernel team. Otherwise, they would not be the first saying that it is fixed in the intrepid kernel, and there wouldn't have been some of them suggesting that something should (should have) possibly be done for 8.04.1 3) the bug appears to be fixed (maybe even incidentally, as the side effect of other changes) in more recent versions of the kernel. In my opinion (but I think that we may agree on it), this - together with the fact that the bug is marked at a high priority, and that frequent lockups can easily result in data loss - should have made this problem mentioned in the release notes, together with notices on how to identify if a platform is affected and with workarounds. Also in my opinion (and here evidently we do not agree), using a 2.6.25 "point 10" is not such a shot in the dark. In my opinion it would be wise to provide it /in addition/ to 2.6.24 /for those/ experiencing this showstopping bug on ubuntu's 2.6.24. In the end 2.6.25 is what everybody is using from Fedora to Opensuse. Indeed, (but again this just "pour parler"), I think that one of the reasons why some kernel bugs have not yet been fixed in Hardy is not only to be sought in the lack of resources of the kernel team (which surely plays a role), but also in the very initial choice of 2.6.24 that prevents from "re-using" the work of other distros kernel teams (thus relaxing the resource scarcity issue).
Ubuntu 8.04.1 LTS released Posted Jul 5, 2008 18:27 UTC (Sat) by tajyrink (subscriber, #2750) [Link] Right, the word "understood" was made up by me, from the "documented". I believe there are hang reports on every kernel, but possibly these hardy ones are now more common than with some of the earlier Ubuntu kernels (2.6.22, 2.6.20, 2.6.17, 2.6.15...). I'm not sure if the release notes could have something more than disclaimer-like "it may be that not all hardware combinations continue to work flawlessly in 8.04", and the same is true with every release. I agree to disagree with the 2.6.25 stableness vs. shot in the dark, and agree to agree with the fact that Ubuntu kernel team could use more resources. I've heard before there are no resources to support more kernels at the same time, which is why .25 cannot be offered in addition to .24. In the gutsy times people would have wanted the kernel team to offer differently compiled kernel flavors because ATI's binary drivers didn't resume from suspend with newer kernels with the new allocator. Thanks for elaborating the issue and answering my questions. I do think the issue is real, any regression is, and it's just details that I'm checking and also wondering how realistic any one solution is (if no clear patch can be found). It's not just a problem of Ubuntu 8.04, it's a problem with the generic problem of regressions that always happen.
|
|
||||||||||||