When I first checked my system today, there was a kernel call trace on the screen. The precesses running included kde with the clock screen saver, boinc, and at the time of the crash, mgaapplet. The only messages in /var/log/messages were ... Sep 9 07:16:53 hodgins mgaapplet[8650]: running: urpmi.update --update Sep 9 07:18:07 hodgins mgaapplet[8650]: Checking... Updates are available#012 Sep 9 07:34:35 hodgins kernel: BUG: Bad page state in process mgaapplet pfn:2566c Sep 9 07:34:35 hodgins kernel: page:f2382d80 count:0 mapcount:-15663104 mapping: (null) index:0xe566cfd8 Sep 9 07:34:35 hodgins kernel: page flags: 0x40000000() Sep 9 07:34:35 hodgins kernel: Pid: 22823, comm: mgaapplet Tainted: P 2.6.38.8-server-5.mga #1 Sep 9 07:34:35 hodgins kernel: Call Trace: Sep 9 07:34:35 hodgins kernel: [<c01dd0f2>] ? bad_page+0xa2/0xf0 Sep 9 07:34:35 hodgins kernel: [<c01dd56a>] ? free_pages_prepare+0x12a/0x140 Sep 9 07:34:35 hodgins kernel: [<c01de8cd>] ? free_hotSep 9 13:07:53 hodgins kernel: imklog 5.6.2, log source = /proc/kmsg started. When mgaapplet checks for updates, it's normally less than 1 minute. Sep 9 14:03:30 hodgins mgaapplet[12174]: Computing new updates... Sep 9 14:03:31 hodgins mgaapplet[12174]: running: urpmi.update --update Sep 9 14:04:13 hodgins mgaapplet[12174]: Packages are up to date I've been running 2.6.38.8-server-5.mga since Sept. 1st. This is the first problem I've seen. I don't know if this can be recreated, or what "Bad page state" means, but thought it should be reported. May be my 6+ year old hardware is starting to die.
Can you run memtest on that system to rule out hw and please run it for a long time, for example over night...
CC: (none) => tmb
I ran memtest for 14 hours with no errors detected, so it seems more likely to be a kernel problem, although I have no idea how to recreate it, as it happened while I was away from the system, and has only happened once in 8 days of using this kernel.
Is the bug still valid with the two updates (one in testing) ?
Yes, but superseded by bug 3186. Closing as old.
Status: NEW => RESOLVEDResolution: (none) => OLD