Switching the root filesystem to writeback mode (instead of the default ordered mode) will work with EXT3 or EXT4. It is not a solution, but it is possibly a method of gathering further data. In my case, I have a system that is functioning (almost) perfectly with kernel 2.6.34 and freezing with kernel 2.6.32-22. The freeze now affects a single process, all the inodes attached to that process and subsequently all processes depending on those inodes - yes, the system does fully freeze after some time, but I have no difficulty accessing logs after recognizing that a process has frozen. (See http://www.iol.ie/~stuartneilson/Bootup_fsck.html)
I have not located the version where this is merged into the Ubuntu kernel, or indeed if it is directly relevant. I am going to test Paul.Drover's suggested test for affected systems from http://ubuntuforums.org/showpost.php?p=9471373&postcount=496
Igor Wojnicki wrote "Take a lookat StuartN's post: http:// ubuntuforums. org/showpost. php?p=9468596& postcount= 487
This might be a hint what causes the freezes."
Switching the root filesystem to writeback mode (instead of the default ordered mode) will work with EXT3 or EXT4. It is not a solution, but it is possibly a method of gathering further data. In my case, I have a system that is functioning (almost) perfectly with kernel 2.6.34 and freezing with kernel 2.6.32-22. The freeze now affects a single process, all the inodes attached to that process and subsequently all processes depending on those inodes - yes, the system does fully freeze after some time, but I have no difficulty accessing logs after recognizing that a process has frozen. (See http:// www.iol. ie/~stuartneils on/Bootup_ fsck.html)
Following the first message, "EXT4-fs warning (device sda4): dx_probe: dx entry: limit != root limit" (which would equally apply to EXT3-fs), leads to a commit 3d82abae9523c33 d4a16fdfdfd2bdd e316d7b56a with the title "dir_index: error out instead of BUG on corrupt dx dirs" - http:// mirror. celinuxforum. org/gitstat/ /commit- detail. php?commit= 3d82abae9523c33 d4a16fdfdfd2bdd e316d7b56a - http:// lkml.indiana. edu/hypermail/ linux/kernel/ 0709.3/ 0208.html
I have not located the version where this is merged into the Ubuntu kernel, or indeed if it is directly relevant. I am going to test Paul.Drover's suggested test for affected systems from http:// ubuntuforums. org/showpost. php?p=9471373& postcount= 496