Bypassing block layer abstractions for true drive sanitization via raw kernel passthroughs (ioctl / SG_IO)

I’ve been digging heavily into the storage stack recently while working on some compliance tooling, and it’s frustrating how unreliable high-level tools can be when you need absolute data destruction.

Running user-space sequential zero-fills or legacy multi-pass overwrites (`shred`, `dd`) on modern NVMe or SATA SSDs doesn't guarantee you hit the over-provisioned or unmapped blocks managed by the Flash Translation Layer (FTL). Worse, it just kills the drive's lifespan.

To bypass the virtual file system entirely and force synchronous hardware-gated interlocks straight to the controller silicon, you have to leverage raw SCSI generic (`sg`) translation wrappers or low-level kernel passthrough structures (`ioctl` layouts like `SG_IO`). This allows you to force native NVMe Crypto Erase or ATA Block Erase commands via the controller ASIC in milliseconds.

It gets even hairier when managing multi-tenant enterprise hardware behind LSI MegaRAID controllers, where you have to automate proprietary binaries like `StorCLI` or flash to IT Mode just to see the raw disks.

https://redd.it/1u3qo1f
@r_linux