Using 7.8-stable/amd64, I'm seeing curious behavior when setting up a softraid-crypto filesystem on a vnd device (backed by a container file stored under /home).
Here's what I did: (as an ordinary user) // create the container file % mkdir /home/jonathan/tmp.x % chmod go-rwx /home/jonathan/tmp.x % dd if=/dev/random bs=1m count=1k of=/home/jonathan/tmp.x/data % chmod go-rwx /home/jonathan/tmp.x/data (then everything else is as root) // set up a vnd device backed by the container file # vnconfig -v vnd0 /home/jonathan/tmp.x/data vnd0: 1073741824 bytes on /home/jonathan/tmp.x/data # // partition the vnd device # fdisk -iy vnd0 Writing MBR at offset 0. # # disklabel -E vnd0 // add partition j: offset=256, fstype=RAID // ... larger-than-default offset is to try to ensure // we don't clobber the fdisk partition table // confirm the disklabel partitions # disklabel vnd0 # /dev/rvnd0c: type: vnd disk: vnd device label: fictitious duid: d994c4b4a2436716 flags: bytes/sector: 512 sectors/track: 100 tracks/cylinder: 1 sectors/cylinder: 100 cylinders: 20971 total sectors: 2097152 boundstart: 128 boundend: 2097152 16 partitions: # size offset fstype [fsize bsize cpg] c: 2097152 0 unused j: 2096896 256 RAID # // create a softraid-crypto disk backed by the just-created RAID partition # bioctl -c C -r 1024 -l /dev/vnd0j softraid0 New passphrase: Re-type passphrase: softraid0: CRYPTO volume attached as sd3 # // partition the just-created softraid-crypto disk # fdisk -iy sd3 Writing MBR at offset 0. # # disklabel -E sd3 // add partition a: offset=256, fstype=4.2BSD // ... again use larger-than-default offset to try to ensure // we don't clobber the fdisk partition table // confirm the partitioning # disklabel sd3 # /dev/rsd3c: type: SCSI disk: SCSI disk label: SR CRYPTO duid: 8e47c8fd3b1c430e flags: bytes/sector: 512 sectors/track: 63 tracks/cylinder: 255 sectors/cylinder: 16065 cylinders: 130 total sectors: 2096368 boundstart: 64 boundend: 2096368 16 partitions: # size offset fstype [fsize bsize cpg] a: 2096096 256 4.2BSD 2048 16384 1 c: 2096368 0 unused # // now create and mount a filesystem on the softraid disk # newfs -i 65536 sd3a /dev/rsd3a: 1023.5MB in 2096096 sectors of 512 bytes 5 cylinder groups of 245.47MB, 15710 blocks, 3968 inodes each super-block backups (for fsck -b #) at: 160, 502880, 1005600, 1508320, 2011040, # mount -o noatime,nodev,nosuid /dev/sd3a /home/jonathan/crypt/tmp # // the mount changed the mount-point ownership from jonathan:jonathan // to root:wheel, and the permissions from drx------ to drxr-xr-x, // so change these back # chown jonathan:jonathan /home/jonathan/crypt/tmp # chmod go-rwx /home/jonathan/crypt/tmp At this point everything seems fine: % mount|grep /home/jonathan/crypt/tmp /dev/sd3a on /home/jonathan/crypt/tmp type ffs (local, noatime, nodev, nosuid) % rsync -aHESvv --delete /home/jonathan/some/stuff/ /home/jonathan/crypt/tmp/ Now the problem: If I repeatedly unconfigure then reconfigure a few times, the softraid-disk filesystem shows up as uncleanly-unmounted (and needs fsck before mounting): // unconfigure # umount /home/jonathan/crypt/tmp # bioctl -d sd3 # vnconfig -u vnd0 // reconfigure # vnconfig -v vnd0 /home/jonathan/tmp.x/data vnd0: 1073741824 bytes on /home/jonathan/tmp.x/data # bioctl -c C -r 1024 -l /dev/vnd0j softraid0 Passphrase: softraid0: CRYPTO volume attached as sd3 # mount -o noatime,nodev,nosuid /dev/sd3a /home/jonathan/crypt/tmp mount_ffs: /dev/sd3a on /home/jonathan/crypt/tmp: filesystem must be mounted read-only; you may need to run fsck # # fsck -p /dev/rsd3a /dev/rsd3a: 794 files, 123804 used, 397619 free (595 frags, 49628 blocks, 0.1% fragmentation) /dev/rsd3a: MARKING FILE SYSTEM CLEAN # mount -o noatime,nodev,nosuid /dev/sd3a /home/jonathan/crypt/tmp # Question: why is the file system showing up as unclean, given that I explicitly unmounted it before unconfiguring the softraid disk? Finally, I see wierd timestamps on the mount point: If everything is unconfigured, the mount point has some timestamp from when I last used that directory before starting this project: # ls -lFgd /home/jonathan/crypt/tmp drwx------ 2 jonathan jonathan 512 Jun 2 2024 /home/jonathan/crypt/tmp/ # But if I re-configure, then the mount point has a wierd mtime: # vnconfig -v vnd0 /home/jonathan/tmp.x/data vnd0: 1073741824 bytes on /home/jonathan/tmp.x/data # bioctl -c C -l /dev/fnd0j softraid0 bioctl: could not open /dev/fnd0j: No such file or directory # bioctl -c C -l /dev/vnd0j softraid0 Passphrase: softraid0: CRYPTO volume attached as sd3 # mount -o noatime,nodev,nosuid /dev/sd3a /home/jonathan/crypt/tmp # # ls -lFgd /home/jonathan/crypt/tmp drwx------ 31 jonathan jonathan 1024 Jan 30 19:11 /home/jonathan/crypt/tmp/ # stat /home/jonathan/crypt/tmp 1072 2 drwx------ 31 jonathan jonathan 560 1024 "Apr 17 23:28:43 2026" "Jan 30 19:11:50 2026" "Apr 18 00:27:21 2026" 16384 4 0 /home/jonathan/crypt/tmp # What's going on here? (How) am I corrupting the filesystem? Thanks for any insights anyone can offer, -- -- "Jonathan Thornburg [remove -color to reply]" <[email protected]> (he/him; currently on the west coast of Canada) Alberto Moreira (comp.arch, Jan 1999): "Cache is, basically, a kluge generated by technological restrictions." Donald C. Lindsay (also comp.arch, Jan 1999): "No. It's a kluge generated by deeply fundamental restrictions, like the speed of light."
