Messages in this thread | | | Date | Mon, 21 Dec 1998 16:59:50 +0000 ( ) | From | Aaron Lehmann <> | Subject | Linux 2.0.35 crash and weirdness |
| |
Hey all,
I did my most sucessfull GNU/Linux install yesterday on my friend's computer. He's a windows user that I had convinced with a few screenshots to install linux on his old Pentium 133.
As he did the insatllation of RedHat and I directed him over the telephone, things seemed to work pretty well. X worked on the first try, networking took a few minutes, and we insatlled all kinds of eye candy like enlightenment and gnome without any trouble. After he had used it for about 90 minutes, it crashed. I'm not sure how, but I would assume X freezed up since this is a windows user by nature and probably remembers that when the cursor freezes, you're dead.
Perhaps just X windows itself locked up but left the kernel running behind it...
So the machine dropped him into single-user mode to fix filesystem problems. Scared the hell out of him. I taught him how to use fdisk and fix his errors.
Sorry for saying that X windows is likely to crash the system when of course we all know that, but maybe the next story will be a little better.
Here's an exercise: start x, start netscape, kill x (lets say its segfaulted or you need to kill it with c-a-backspace becuase it's not responding). Now look at your load average. Netscape has gone crazy. It's probably in an infinite loop.
The same thing happened to my friend _for_no_apparent_reason_. Suddenly you would notice that the load average was high and I would have to tell him to kill netscape and gmc. (I think it happened 4 or 5 times with netscape and once with gmc.) The load average gets pretty high when 3 processes have gone into infinite loops!
I suppose that this is a bug in a few applications. Still, it is not acceptable in a server setting to come in a few days latter and find that 3 netscape processes that i thought were long gone had gone into infinite loops.
Perhaps the kernel should kill processes that go into infinite loops. I am not sure how that would be possible, especially when it is perfectly accectable to use 100% of the cpu (examples: rc5des, crack). For some reason, these phantom processes usually do not end up using the full 100% of the cpu. its usually more like 50%, but if 3 are running tahn thats an attempted 150% load on a poor little box. I've had similar thinks happen on my machine too, but usually only when X is killed.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |