| From: Jeffrey Ross <jeff@xxxxxxxxxx> | I'm having system lockup problems with Fedora 7. This is running on a | machine that ran FC6 for long stretches of time without reboot. | | My questions are: | | 1) How to I track down what is causing the lockups | 2) When I find out what is causing the lockups how do I submit a bug? Good questions. Ones that I'm working through at the moment. Perhaps what I'm about to say is covered on the wiki but I don't see it. What do you mean by a lockup? I mean: computer no longer responds to external stimulae: mouse movement, keys struck (including Alt SysReq), pings from other computers, ... (If it isn't really locked up then the Magic System Request Key might give you useful information; see Documentation/sysrq.txt in the kernel source, or google.) The best case for a lockup is that the kernel has panicked. Alternatively, it could be locked up with NO signalling of what is wrong. Oh well. Assuming that the kernel has panicked, how can you get hints of what has gone wrong? A kernel panic dumps a bunch of diagnostic information on the "console" and then sits there with interrupts disabled, ignoring everything. So you need to have captured the console output. Too late to decide this after the crash. Unfortunately, a panic shuts everything down, including logging. If you could issue a dmesg, you would see the messages, but you cannot issue a dmesg. Similarly, the panic won't appear in /var/log/messages. A BUG message (sort of like a non-fatal panic) doesn't shut things down so either of these techniques would work. How can you see console output? By not running X and just using the text console. Pretty restrictive. Or by using a serial-console (more about this later). In my current testing, I'm leaving the text console up on the main screen and using second computer as an x-server for commands that I'm running on the first computer. (Remember, in X terminology, the X server is the machine with the screen and the client is the machine with the program.) It really helps to have more than one computer! The information presented upon a kernel panic is probably more than the 25 lines that a normal text console has. So you will lose some information (perhaps the most important) if you use a normal text console. I boot my kernel with the option vga=795 to give me 1280x1024 resolution on my text console; that gives me about 64 lines of text. Even this isn't enough (because I have a dual processor, there are two stackdumps in the panic message). If you don't use the console for a while, it will blank. Once a crash happens, you cannot induce the screen to become readable again. So I use setterm(1) to stop blanking. setterm -powerdown 0 -blank 0 If you don't want to tie up your main display while you wait for a panic's message, you might consider a serial console. A terminal (or other computer) connected to your computer by a built-in serial port can be used as the console. I've not tried this because my main machine doesn't have a serial port. You can even use a printer port as an output-only console but I think that this requires a printer driver that is built-in, not a module as is (I think) the case with fedora. See http://tldp.org/HOWTO/Remote-Serial-Console-HOWTO/ http://mbligh.org/linuxdocs/Kernel/SerialConsole Documentation/serial-console.txt in kernel source I see in you /var/log/messages that you are experiencing https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=240982 but that should not cause a lockup. This is an example of a BUG message. For an example of a panic message, see an earlier message I wrote to this list https://www.redhat.com/archives/fedora-list/2007-June/msg01076.html
