On Fri, Oct 16, 2015 at 21:02:46 -0700, Colin Percival wrote: > On 10/16/15 01:06, Christoph Borsbach wrote: > > On Thu, Oct 15, 2015 at 13:33:30 -0700, Colin Percival wrote: > >> The relevant limits value is the datasize, not the stacksize. But to > >> answer > >> your question, scrypt uses a large amount of memory -- the larger the > >> better > >> -- to convert your passphrase into the key used for encryption; so this is > >> entirely independent of the amount of data you're processing. > > > > I have access to another OpenBSD System with 6 GB Ram. I cranked the > > datasize > > limit to ~5GB and still I get the same error: > > > > $ ulimit -a > > time(cpu-seconds) unlimited > > file(blocks) unlimited > > coredump(blocks) unlimited > > data(kbytes) 5316608 > > stack(kbytes) 4096 > > lockedmem(kbytes) 2026642 > > memory(kbytes) 6078248 > > nofiles(descriptors) 512 > > processes 256 > > Hmm, I wonder why scrypt thinks you don't have enough memory then. > > > Anyway, do I understand the -M option to scrypt correctly that if I use, > > say, > > -M 536870912 (512MB), on all of my machines, it should decrypt on all > > machines > > (that have a ulimit of 512MB or more) without error? > > You'll need a slightly higher ulimit than that (since there is a small amount > of memory used by other things), but yes, if you use -M to set a reasonably > low limit everywhere then you should be fine. If that doesn't work on the > systems you've mentioned above, I definitely want to know.
Hi, with the -M option it works in all directions. Thanks for the pointers! All the best, Christoph
