hi. i'm using 2.2.18 with ext3-0.0.5e on a dual p3-550 with 1.5 gigs of ram. the system is redhat 6.2-based but with lots of modifications and a clean kernel build. i've been using 2.2.18 and 2.2.17 with ext3 patches on other machines with much success and no problems, up until last night, when the machine mentioned above kernel panicked on me. unfortunately i don't have much of the info. the console was completely locked and all i could do was cycle power, but the oops message on the screen was in a jfs_something() call within the kupdate process. this machine is often under considerable load and memory use, altho probably not too much disk activity. i'm just wondering if any errors along these lines have been reported / if this is a known issue at all, or if i've stumbled across something unique to my setup or a potential unknown bug. the system has a / ext3 with journal partition, a small ext2 /boot partition, and 2 swap partitions, and that's it. if that makes any difference. any info would be appreciated. -tcl.
TC Lewis writes:> hi. i'm using 2.2.18 with ext3-0.0.5e on a dual p3-550 with 1.5 gigs of > ram. the system is redhat 6.2-based but with lots of modifications and a > clean kernel build. > > i've been using 2.2.18 and 2.2.17 with ext3 patches on other machines with > much success and no problems, up until last night, when the machine > mentioned above kernel panicked on me. > > unfortunately i don't have much of the info. the console was completely > locked and all i could do was cycle power, but the oops message on the > screen was in a jfs_something() call within the kupdate process.Possibly a JASSERT? Pretty slim info to go on. Stephen will be able to fix it, I imagine, because everyone knows the major Linux hackers have ESP. Getting the function name and a few of the details would be a good start. If you can compile the kernel with KDB (patches included with ext3), then you can do a stack trace (the "bt" command), to help out a bit more.> i'm just wondering if any errors along these lines have been reported / if > this is a known issue at all, or if i've stumbled across something unique > to my setup or a potential unknown bug.Haven't heard much in this regard. Have you checked bugzilla? It may be because of highmem or something. Because it is basically your whole installation on a single filesystem, it is also harder to narrow it down by the type of I/O that is being done. Cheers, Andreas -- Andreas Dilger \ "If a man ate a pound of pasta and a pound of antipasto, \ would they cancel out, leaving him still hungry?" http://www-mddsp.enel.ucalgary.ca/People/adilger/ -- Dogbert
Hi, On Mon, Jan 29, 2001 at 07:22:28PM -0500, tc lewis wrote:> > unfortunately i don't have much of the info. the console was completely > locked and all i could do was cycle power, but the oops message on the > screen was in a jfs_something() call within the kupdate process.There's one possible explanation for this which will be fixed in ext3-0.0.6. However, a detailed record of the oops is always valuable in this sort of situation.> this machine is often under considerable load and memory use, altho > probably not too much disk activity.Memory pressure is exactly the trigger for the race I know of (which basically involves the journaling code sleeping due to out-of-memory while trying to journal a previously non-journaled buffer.) Cheers, Stephen