Re: 3.0 wishlist
Albert D. Cahalan (acahalan@cs.uml.edu)
Tue, 20 Jan 1998 17:02:50 -0500 (EST)
From: Dan Hollis <goemon@sasami.anime.net>
> On Tue, 20 Jan 1998, Albert D. Cahalan wrote:
>> Dan Hollis writes:
>>> On Mon, 19 Jan 1998, linux kernel account wrote:
>>>> I think you should point out the classic horror story:
>>>>
>>>> Imagine the panic results in the coruption of the disk driver or
>>>> the FS and it decides to dump core all over your important data..
>>> Thats why BSD and others do a CRC on the dump code to make sure
>>> everything is consistent enough to do a dump. It also only writes
>>> to reserved section of the disk and not the filesystem. Thus the
>>> horror story becomes vapor.
>>
>> Oddly however, NetBSD manages to trash partitions.
>
> I did a search through altavista and was unable to find any reports.
> Do you have any documented cases of this?
It is listed at their own website on a bug collection page.
Don't imagine Linux dumb code would be bug-free.
> I'm leaning toward the way IRIX does dumps now... which nullifies
> all arguments against 'confused hardware'. This will only be
> possible if a way can be found to prevent BIOS from stomping all
> over memory on a warmboot though.
Going over the network might be safe, since you could disable
disk access first. Maybe use a separate network stack with a
really simple non-standard (not IP) protocal. It could use
Ethernet broadcast and forget any streaming features.