> On May 28, 2026, at 6:01 AM, Thor Lancelot Simon <[email protected]> wrote: > > It's concievable the Xylogics controllers and their drivers would work > (if the Connector Conspiracy did not intervene) not just with the CDC 97xx > drives that were commonly used on larger minicomputers (on VAXen, usually > via Emulex controllers on Massbus) but with their OEM variants, including > several from DEC and, I think, some from other CPU manufacturers too. > > These have removable packs and are not an implausible > target for historical data-recovery. I actually think that's an important > use we should support. However, I also think _booting_ from such a drive > actually is one of the higher-risk practices if one is concerned about > its data forever descending into the memory hole, and I don't think we > need direct support in the device drivers for any reason except to boot.
I mean, if that’s where the OS is installed and you don’t happen to have a SCSI3 board, that’s what you’re booting from. (Ok, sure, you can net-boot, but *really*?) > This pushes me further towards the conclusion that if any reorganization > of how the kernel handles bad144 is in order, a pseudodevice would make > the most sense. I think everyone is over-thinking this, here. What I intend to do is quite minimally-invasive and leaves the logic in the individual drivers that consume it essentially intact (a flag-in-the-disklabel check will be replaced by a pointer-in-the-softc check for “maybe we have to re-map a sector”). -- thorpej
