4.8 - / out of space
dkelly at HiWAAY.net
Mon Jan 3 20:10:45 PST 2005
On Jan 3, 2005, at 7:07 PM, Gregor Mosheh wrote:
> --- David Kelly <dkelly at HiWAAY.net> wrote:
>> Another goof is for root to "write" to an unmounted filesystem.
>> Later when the filesystem is mounted the written files are hidden
>> yet still consume space on the fs containing the mount point
>> (usually /).
> Could you explain how this happens (or point me to a doc)? Do you
> mean something like "tar cvf /dev/ad0s1a"?
Guessing you quoted the wrong paragraph for the question.
As for tar, yes, "tar -cvf /dev/baddevicename myfiles" will happily
create a file (not device) named /dev/baddevicename and write the
contents of myfiles into it. Iff you have write permission on /dev/.
Its no different than "tar -cvf myfiles.tar myfiles" other than you
tried to hit a device but created a file instead.
As for what I was writing about consider the case where one is running
single user and /usr is not yet mounted. A directory named /usr exists.
Nothing is preventing root from writing in the /usr directory. This
will consume space on / but no longer exist in file namespace once a
filesystem is mounted on top of the /usr directory. Files are still
there but you can't get to them.
> Does that cause fs corruption? Would fsck reclaim that space?
No corruption. Fsck is perfectly happy with it and won't change a
thing. The stuff hidden under the mount point is still there. The only
way to get at it is to umount the fs on top and then the previous
contents of /usr reappear. Assuming of course that /usr actually had
Go read the manpage for mount. Search for "union". Its talking about
the same thing as above only a union mount appends the underlying
namespace at the end of the mounted filesystem's namespace. Without
"union" the underlying namespace is unreachable.
David Kelly N4HHE, dkelly at HiWAAY.net
Whom computers would destroy, they must first drive mad.
More information about the freebsd-questions