On Thu, Sep 14, 2017 at 4:27 PM, Fred Cisin via cctalk < cctalk@classiccmp.org> wrote:
> On Thu, 14 Sep 2017, Warner Losh via cctalk wrote: > >> And then there's Venix disks, which have both an interlace, as well as a >> skew from track to track (as well as doing one side in ascending track >> order followed by the other side in descending track order), but I >> digress... >> > > The software that you use must deal with interlace and skew. > Yea, as with the DOS disk, this was all logical. The physical sectors on the disk weren't skewed (they all started at the index mark) and were numbered 1..10 in order. It's the logical swap done in the drivers for speed that the upper layers didn't know (or care) about. So if you looked at the headers, you'd never know that these sorts of things were going on. For the record, Venix's f0 device skipped track 0 and used the 2:1 interleave on the rest of the tracks with a span of 7. so the first track was 1 3 5 7 9 2 4 6 8 10, but the next track was skewed by 7, so track 2 ran 6 8 10 1 3 5 7 9 2 4 etc. All the 'restore' disks and 'tar' disks for Venix were like this and it took some trial and error to discover this to recover the raw tar files from the Rainbow Venix disks... These values are slightly different for 1.2MB disks for the IBM PC version of Venix, but the same for 360k disks (well, with fewer sectors, obviously). You can read through my trial and error sequence starting at http://bsdimp.blogspot.com/2017/04/rainbow-100-venix86r-disks-found.html if anybody cares. There's quite a few mis-steps along the way... > (typically with a list/table of sector sequences, and variables such as > index gap and inter-sector gaps) > Yea. It depends on the controller. WDC765 (or was that a clone of the NEC uP765) and the Intel 8287A and their ilk need these details since they try to cope with reading and they need some help guessing where the sector started. For some reason, the older WDC-179x chips didn't care, but on those chips to format a track you had to send it the sequence of bytes which would ultimately make it onto the drive (so there was no format track command, just a write track command). > Thank you for the details. > You bet. > There are an amazing variety of possibilities. > A physical skew and/or interlace/arrangement of sectors will sometimes > work in the wrong order, but with a performance penalty (or, in some cases, > improvement). > The speed improvement is quite noticeable. It was 2-3x faster with the interleave for a mostly sequential work loads, and the skew that venix did added about a 20% improvement on top of that. Random workloads sucked no matter what you did :( > I have run into quite a few badly maintained drives. The innermost > (higher numbered) tracks are the most sensitive to problems. > Yes. I had the most errors on those tracks when I was reading my big box of floppies back. My wife thought I was crazy for doing that... I'm not sure she was wrong :) As you might tell from this thread, floppies, reading them, and their formatting stir way too much passion in me (for which if it was too much I apologize).... Warner
