cvs commit: src/sys/nfsclient bootp_subr.c nfs_diskless.c nfs_vfsops.c nfsdiskless.h

[Poul-Henning Kamp]

In message <20030816081235.GC74853 at cirb503493.alcatel.com.au>, Peter Jeremy wri
tes:
>On Sat, Aug 16, 2003 at 09:10:01AM +0200, Poul-Henning Kamp wrote:
>>In message <20030816010942.GC8274 at wantadilla.lemis.com>, "Greg 'groggy' Lehey" 
>>writes:
>>>>   Suggested replacement command sequence on the client:
>>>>
>>>>           dd if=/dev/zero of=/swapfile bs=1k count=1 oseek=100000
>>>>           swapon /swapfile
>>>>           rm -f /swapfile
>>>>
>>>>   For whatever value of 100000 you want.
>>>
>>>I'm confused.  Why was this necessary?  Which is the "magic" way?  Is
>>>the rm -f /swapfile really necessary, or just a foot guard?
>...
>>Removing the swapfile is means that the file dies on the server and
>>that nobody can write trash in it.
>
>Not if / is NFS mounted:  Removing the file on the client just renames
>it to a .nfsXXXXXXXX name.  It retains the original permissions and
>is still on both the client and server under this dummy name.

I thought we had magic which prevented client lookup of those names ?

Either way, if root wants to trash his swapspace doing so is no worse
than the many other things he could do.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.