Messages in this thread |  | | From | Matthias Urlichs <> | Subject | Re: pre2.0.9 (was Re: CD-ROM Access Crashes) | Date | Sat, 1 Jun 1996 09:10:04 +0100 |
| |
In linux.dev.kernel, article <25301.833577365@drax.isi.edu>, Craig Milo Rogers <rogers@isi.edu> writes: > >No, we actually _have_ to call readpage regardless of any old > >error-status, because the error might go away (maybe the error was due to > >protection problems with the previous user, but now we're another person > >so..). > > I am confused by the statement above. Does it imply that its > converse might be true: that a user might successfully access a page > (read and cached by a prior user) that they would have been prevented > from reading had readpage been called? > No.
Well, yes, actually, but only if you're root and first try the access as a normal user. This is not a security hole because if the remote file is owned by root, this wouldn't help you, and if it isn't, there's nobody to prevent you from locally setuid()ing yourself to the file's owner in the first place.
The NFS no-root-allowed feature really doesn't prevent accessing files as root, it just makes root from A and B be a different user.
Unfortunately, Linux itself doesn't know anything about that, hence these problems.
Effectively, thus, no. ;-)
-- Small Earthquake in Chile; Not Many Killed -- Healing suggested for The Times of London by Claud Cockburn -- Matthias Urlichs
|  |