fsck & 3b1 continuous power up

[Alex Crain]

In article <331 at heurikon.UUCP> dklann at heurikon.UUCP (David Klann) writes:
>In article <1084 at adds.newyork.NCR.COM> tanya at adds.newyork.NCR.COM (Tanya Katz) writes:

>It's too bad the UNIXpc kernel doesn't have/make use of the s_state field in
>the super-block.  That would increase the certainty about the state of
>the root file system when booting.  I'd love to set the file system
>state to "FsOKAY" as the last thing before shutting the system down.  In
>fact I think I'll look into that very thing.  That's it!  All we need is
>a pair of utilities, one to set it, and a version of fsstat to check it.
>I'll post them if/when they're working...
>
>Comments?  Flames?

	I don't think that this will work, because the disk never
gets unmounted. In order to unmount the disk, you would have to
flush all the kernals buffers and guarentee that nothing will get
written to disk after you mark the disk.

	This works on other systems because they have a root 
partition and a user partition that can be umounted separately. 
When the kernal reboots, it kills all the user processes and 
unmounts /usr, marking it clean. Then it shuts down, leaving /
in an undetermined state (which is almost always ok). When the
system starts up, it only has / to wory about, so it boots right 
quick.

	The problem I envision is having the kernal crash in 
between the marking of the disk and the reboot. The system would
then come up, find the disk clean, and propagate the disk trash
everywhere (or just over the inode table). Not likely, but likely
enough that I won't do it to my machine.

	It is probably possible to partition the disk into / and
/usr, but it more trouble then its worth, me thinks. Nice thought
though.





					:alex
Alex Crain
Systems Programmer			alex at umbc3.umbc.edu
Univ Md Baltimore County		umbc3.umbc.edu!nerwin!alex