home *** CD-ROM | disk | FTP | other *** search
- From: itschere@techfak.uni-bielefeld.de
- Subject: Re: XATTR structure for BIOS filesystem #2
- Date: Mon, 22 Nov 93 12:32:27 MET
- In-Reply-To: <9311220909.AA25289@hpbeo79.bbn.hp.com>; from "Claus Brod" at Nov 22, 93 10:09:22 am
-
- > Not too bad an idea. I'd suggest the following "calling convention":
- >
- > - call datime in mode 0, store return value
- > - call datime in mode 2
- > - call datime in mode 0; compare with previous result; if the results
- > differ, call datime in mode 1 with the old value
- >
- > This way, you make sure that even old drivers will not permanently
- > change access times because they don't know about mode 2 and 3.
-
- That's right, but:
-
- 1) You would then force "old" drivers to keep their creation time, which
- might be a good idea, but may be useless 'cause MiNT now returns the
- current time/date for it, at least if you call Fxattr() and not Fdatime(),
- so I doubt any driver itself, or literally any program uses this info so
- far, so you needn't expect any troubles if you don't do so, I think.
-
- 2) (more general) Setting up an auxiliary FILEPTR in getxattr and filling it
- with some fake data already needs some time (and so far I've forgotten to
- take care of device drivers being responsable for more than one device, so
- this would involve yet more ops) plus 3 or 4 calls to fdatime... Well, it's
- not that I particulary dislike it, but I'm a _bit_ fanatic on speed, so I'd
- like to be at least sure there isn't a faster solution with the same
- properties.
-
- The completely other idea (as discussed some days ago) could change the data
- itself by gaining access to its bios_file over (struct bios_file *)f->fc.index,
- only that for now it doesn't know an XATTR field exists. To be very clear:
- This mustn't be extra implemented, this is already there, so I may very well
- consider my own security thoughts being totally wasted and unfullfillable :-(
-
- But this one would be quite faster for both sides.
-
- But either method would take advantage if the kernelptr would not only
- provide time manipulating functions, but time values itself... :-)
-
- Shall we vote which idea is better? ;-)
-
- bye,
- TeSche
- --
- PS: If the above written looks weird, than that's because it probably IS.
- WhoDunnIt: Torsten Scherer (Schiller, TeSche...)
- Technical Faculty, University of Bielefeld, Germany (52'5"N 8'35"E)
- EMail: itschere@techfak.uni-bielefeld.de / tesche@dave.hrz.uni-bielefeld.de
-