Messages in this thread | | | From | Keith Owens <> | Subject | Re: Oops assist... | Date | Fri, 07 May 1999 19:54:40 +1000 |
| |
On Thu, 06 May 1999 23:20:18 -0500, Nicholas Henke <henken@seas.upenn.edu> wrote: >What if this were done to a "who-cares" filesystem, such as that on a >floppy disk.
That only solves half the problem. If the kernel is still running after an oops then you can write to a floppy/printer/disk/whatever. But if the kernel is still running, then writing to syslog is almost as good. The other half of the problem (the hard bit) is how to dump kernel data when it is so badly hosed that interrupts do not work and you will have to reboot to get the system running again.
When I say "interrupts do not work", this can be anything from a single masked IRQ, through bottom half handling problems all the way up to all interrupts being masked. Basically anything that causes a hard system hang. Linux runs fast because we catch an interrupt, grab the data and (most of the time) schedule a bottom half to process the data. If bh does not run, the interrupt is effectively dead.
Right now, the only way we have of capturing oops data when interrupts are not working is via a serial console. The serial console runs in polling mode, no interrupts, no bottom half handling. Any other suggestion for capturing kernel data really needs to have the same characteristics. So forget about dumping oops to any file system unless you can do all the I/O without interrupts or bh.
- 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/
| |