On Fri, Feb 03, 2017 at 09:10:41PM +0800, Tinker wrote: > On 2017-02-03 21:03, Stefan Sperling wrote: > > On Fri, Feb 03, 2017 at 08:03:09PM +0800, Tinker wrote: > > > Hi, > > > > > > I have a OpenBSD 6.0 GENERIC.MP system set up as follows: > > > > > > * sd0 is a physical harddrive. It has a "b" partition for swap, and > > > an "a" > > > partition for a softraid. The softraid is represented by sd1 . > > > > > > * sd1 is the softraid. It has some UFS partitions (a, d, etc.). > > > Importantly, it has no swap partitoin (which, if it would have > > > existed, > > > would have been named "b" by convention), as the system's swap is on > > > sd0 > > > already. > > > > > > To the best of my awareness this is a conventional and intended > > > OpenBSD > > > setup. > > > > No, the convention is to put swap onto the root disk. > > But.. isn't "sd0" the what you call the "root disk" in this case?
The "root disk" is the one where the kernel finds the root filsystem, for example: root on sd2a (6189a72b022271e5.a) swap on sd2b dump on sd2b > Also, if I would have put the swap on "sd1", then its contents would be > encrypted doubly. Isn't that a bit wasteful. Not at all. Swap encryption and disk encryption serve different purposes. Swap crypto keys are discarded when the system resets to make residual data in swap unrecoverable. > Anyhow, ok if this is the case then thank you very much for highlighting it. > > Is the fact that this is the convention declared or reflected anywhere else > (than in the fact that 'savecore' breaks if you not follow it)? I don't know if this is specifically written down anywhere.
