On Sun, 1 Apr 2001, Philip Stoev wrote:

> So, my question is: can a non-privileged user cause a small filesystem
> corruption on non-privileged data or portion of the drive, however
> significant enough to cause the fsck to fail and the boot process to
> enter single user mode? I assume that filesystem manipulation
> functions will not create corruption unless being forced to somehow.
> So, can you create a set of circumstances so that any of those
> functions fail? Like making a nasty race-condition or killing your own
> program at the wrong place?

depends on how well maintained the server is. ext2 is, in my experience, a
somewhat fragile filesystem when compared to more resistant ones like FFS,
UFS, XFS or JFS.

anyhow, its pretty straight forward: a lot of disk activity. a whole bunch
of mv's, cp's, makes, etc .. get those running simultaneously. just abuse
the filesystem. then locally DoS the machine (take your pick of methods).
and if itreboots automagically, then you get what you anticipate.

there are easier ways to hang a system though.

one workaround is synchronous mounts, at least on user writable
filesystems. that helps keep metadata up to date, and hence fscks do less
work when they occur. but you get a performance hit ...

____________________________
jose nazario josecwru.edu
                           PGP: 89 B0 81 DA 5B FD 7E 00 99 C3 B2 CD 48 A0 07 80
                                       PGP key ID 0xFD37F4E5 (pgp.mit.edu)

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]