Date: Wed, 06 Apr 2005 18:35:11 +1000 From: Stephen McKay <smckay@internode.on.net> To: Poul-Henning Kamp <phk@phk.freebsd.dk> Cc: Stephen McKay <smckay@internode.on.net> Subject: Re: Can't change partition table anymore Message-ID: <200504060835.j368ZB2i009939@dungeon.home> In-Reply-To: <6570.1112773884@critter.freebsd.dk> from "Poul-Henning Kamp" at "Wed, 06 Apr 2005 09:51:24 %2B0200" References: <6570.1112773884@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wednesday, 6th April 2005, "Poul-Henning Kamp" wrote: >We will not by default allow people, be it novice or selfdescribed >wizards, to write to disk sectors which a filesystem is currently >in possesion of, without a deliberate disabling of the protection >mechanism. As a "self described wizard", I know which sectors I can write safely. Protect the novices all you like, but don't prevent me from doing interesting/extraordinary things. >Enabling foot-shooting is in the category of open-heart surgery: >it is not something we want people to try "just to see if that >happens to solve my problem". > >So the sysctl knob is here to stay, one way or another. OK, if you want it that way, then make it a feature, not a debug flag. >It is far less obvious where the documentation of features like >this belong than most people think. This is not something that >belongs in the dd(1) or ata(4) manual pages, although they could >and probably should cross-reference it. In the geom man page, perhaps? (With the cross-references too, of course.) >It has been suggested that the kernel issue a printf when this >happens, but that is 100% precisely the wrong response: that would >introduce an effective DoS against any machine with a serial console. I agree that a kernel printf is not a solution. >And I hate to say this, but this "horribly undocumented sysctl" is >in company of about 200 other equally undocumented sysctls in the >system, many of which have equally profound impact on how the system >works. Having bad documentation in one area is not a good reason to introduce poorly documented features in another. >So for all I care, this discussion is over until somebody comes up >with a patch we can all agree on. "Harrumph!" he said, and stalked from the room. :-) I'm not yet convinced that this whole thing is not just a bug. I'm off to read some manuals and some code... Stephen.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200504060835.j368ZB2i009939>