Get slag.sh tonight!

root@slaglab:~# bash /root/slag.sh wipe --target /dev/vdb --allow-partial
[warn] not targeted (--allow-partial): /dev/vda - data on them is untouched

PROOF MODEL - read before trusting the result
---------------------------------------------
HOW (zeroize, then verify):
  * zeroize reads every block and writes zeros only where a block is not
    already zero: every block that held data is overwritten in place; blocks
    that read zero were not written (on thin storage they stay unallocated)
⋮
DOES NOT PROVE (no in-guest tool can):
  * hypervisor/provider SNAPSHOTS or BACKUPS are gone  -> delete them by hand
⋮
Mitigation for everything in the second list = the rotation checklist.
Rotate/revoke every credential so a recovered block is worthless.

TARGETS (every byte on these is destroyed):
  /dev/vdb         model=?  serial=slagtgt  size=512M
Type the device names to wipe, separated by spaces ("vdb"): vdb

[slag] re-exec from tmpfs: mounting /run/slag.20261105T083936Z
[slag] staging the wipe's tools into /run/slag.20261105T083936Z
[warn] not installed (steps needing them record 'unsupported'): hdparm sg_sanitize mdadm
[slag] self-test: running the staged bash, script and tools from /run/slag.20261105T083936Z
[slag] self-test: bash 5.2.15(1)-release running from /run/slag.20261105T083936Z
[slag] self-test: static busybox ok
[slag] self-test: busybox has httpd (--serve)
[slag] self-test: sha256sum ok
[slag] self-test passed: bash, dd, cmp, od, blockdev run from tmpfs; verify catches short reads and nonzero bytes; zeroize writes exactly the nonzero 64 KiB blocks; the strike zeroes exactly the blocks holding key signatures; overwrite patterns write every byte and read back by SHA-256; filefrag output maps to disk offsets
[slag] wipe running detached as systemd unit slag-wipe (pid 1306); it survives this SSH session

08:39:37 [slag] quiescing: swap off, stop writers on the targets, sync, emergency read-only remount, stop md/dm
⋮
08:39:37 [slag] strike /dev/vdb: scanning all 536870912 bytes for 11 key signatures; zeroing each match's block now
08:39:37 [slag]   [ok] /dev/vdb strike - 0 mapped file extents; scan: 0 signature hits; 0 ranges, 0 bytes struck and read back zero; 0 bytes written
08:39:37 [slag] === hardware erase: /dev/vdb (?, serial slagtgt, 536870912 bytes, rotational=yes) ===
08:39:37 [warn]   [unsupported] /dev/vdb hw-erase - no hardware erase path for vdb (virtual device?)
08:39:37 [slag] === zeroize + verify: /dev/vdb (?, serial slagtgt, 536870912 bytes) ===
08:39:37 [exec] wipefs -a -f /dev/vdb
08:39:37 [slag]   [ok] /dev/vdb wipefs - /dev/vdb: 2 bytes were erased at offset 0x00000438 (ext4): 53 ef
08:39:37 [slag] zeroize /dev/vdb: reading all 536870912 bytes in 16 MiB chunks (direct I/O), writing zeros only to 64 KiB blocks that are not already zero
08:39:42 [slag]   [ok] /dev/vdb zeroize - conditional zero pass: 536,870,912 of 536,870,912 bytes read; 520,093,696 bytes written where a block was not already zero (7936 blocks of 64 KiB in 32 writes, in 32 of 32 16 MiB chunks)
08:39:42 [slag] verify (full read-back, direct I/O): reading all 536870912 bytes of /dev/vdb
08:39:43 [slag]   [ok] /dev/vdb verify - full read-back: 536870912 of 536870912 bytes read, all zero
08:39:43 [slag] VERIFIED: /dev/vdb reads back all zero, 536870912 of 536870912 bytes
08:39:43 [slag]   [ok] /dev/vdb recheck - 1 signature blocks still zero

===================== SLAG WIPE RECEIPT =====================
tool:     slag.sh 1.1.0
host:     slaglab
started:  2026-11-05T08:39:36Z
finished: 2026-11-05T08:39:43Z
run:      rescue (no target filesystem mounted), from tmpfs /run/slag.20261105T083936Z
zeroize:  conditional zero pass: every block read (direct I/O, 16 MiB chunks); zeros
          written only to 64 KiB blocks not already zero, in place; no discard before it
verify:   full read-back of every byte
reclaim:  not requested; nothing was discarded
⋮
device /dev/vdb
  model=?  serial=slagtgt  size=536870912 bytes
  signatures               ok           1 offsets noted (signatures and partition starts)
  strike                   ok           0 mapped file extents; scan: 0 signature hits; 0 ranges, 0 bytes struck and read back zero; 0 bytes written
  hw-erase                 unsupported  no hardware erase path for vdb (virtual device?)
  wipefs                   ok           /dev/vdb: 2 bytes were erased at offset 0x00000438 (ext4): 53 ef
  zeroize                  ok           conditional zero pass: 536,870,912 of 536,870,912 bytes read; 520,093,696 bytes written where a block was not already zero (7936 blocks of 64 KiB in 32 writes, in 32 of 32 16 MiB chunks)
  verify                   ok           full read-back: 536870912 of 536870912 bytes read, all zero
  file-extents             skipped      no mapped sensitive-file extents on this device
  recheck                  ok           1 signature blocks still zero
  zeroized: 536,870,912 bytes read, 520,093,696 bytes written where a block was not already zero; every block verified zero.
  verify:  full: 536870912 of 536870912 bytes read, all zero
  device verdict: PASS

slag summary, written 2026-11-05T09:20:19Z

mode:       wiped from the provider's rescue system
disks:      /dev/vda 3.0 GiB (disk 1), /dev/vdb 256.0 MiB (disk 2), /dev/vdc 64.0 MiB (disk 3)
⋮
verdict:    PASS
⋮
steps:
  push              ok            0.4s  slag.sh sha256 e5e80f203750a8fb646ddac393d28a8d1523214f2190421e9a44089219ae4fca on the target
  markers           ok            0.5s  SLAG-MARKER-02276c5e2bcb02ee792943f4 x3 (50.0 MiB each): /root/.slag-marker, /etc/.slag-marker, /var/.slag-marker
  inventory         ok            1.2s  inventory + checklist copied, tarball sha256 4c3482ed818def5cc0bf8e85c7f4b24587b2942bf1bb2a4cd317c4596805a034 matches
  rescue_boot       ok           10.5s  rescue system running; device map vda={"disk_id": 1}, vdb={"disk_id": 2}, vdc={"disk_id": 3}; rescue media loop0, sr0, zram0
⋮
  rescue_targets    ok            0.3s  wipe targets /dev/vda=disk 1 (3.0 GiB), /dev/vdb=disk 2 (256.0 MiB), /dev/vdc=disk 3 (64.0 MiB) match labqemu's device map and sizes; not targeted: nothing
  wipe              ok            2.1s  slag.sh wipe launched; receipt URL recorded (token not logged)
  verdict           ok           30.0s  VERDICT PASS; RECEIPT.txt sha256 05d93ef377c400a95dfe503f7c0edd4a60a6afd19c24854683fdcc179283bc67; served .sha256 matches; printed by slag.sh: matches
  proof             ok           11.4s  from the labqemu rescue system: vda ZERO-OK hits=0, vdb ZERO-OK hits=0, vdc ZERO-OK hits=0
⋮
  notify            ok            0.0s  notification sent (HTTP 200); SUMMARY.txt written

still yours to do:
  - rotate every credential in inventory/slag.out/ROTATION-CHECKLIST.md
  - provider: backups cancelled and the instance deleted by slag; check the account for volumes and images it did not touch

  > 2026-11-05T09:55:43Z  KEXEC: slagvps stops now and comes back as the slag appliance
⋮
  > 2026-11-05T09:56:22Z  appliance up after 5s:
  >   |   slag 1.1.0  ·  SSH key SHA256:ObcKVOLk+hN/321hq+DuDEr0q15PY5Pis79pJ+9xk3U burn-lab
  >   |   host key: SHA256:rBS4Fk/W+earMWB0BofoJFKvRZW/7hIKgpgtlI1NyIA
⋮
  >   | /dev/vda          3 GiB  -                    -                    ok
  >   | /dev/vdb        370 KiB  -                    -                    protected:read-only (cannot be wiped; e.g. a config-drive or ISO)
  > 2026-11-05T09:56:22Z  not wiped: /dev/vdb protected:read-only (cannot be wiped; e.g. a config-drive or ISO)
  > 2026-11-05T09:56:22Z  WIPE: slag --wipe-all  --confirm DESTROY --background (every disk slag offers)
⋮
  >   | 09:56:29 [slag] zeroize /dev/vda: reading all 3221225472 bytes in 16 MiB chunks (direct I/O), writing zeros only to 64 KiB blocks that are not already zero
  > 2026-11-05T09:57:02Z  /dev/vda  PASS  (zeroed and verified)
  > 2026-11-05T09:57:02Z  DONE. All selected disks were zeroed and VERIFIED.
⋮
  host: the VPS's root disk (3221225472 bytes, base image + overlay) reads back all zero