Geometry

This layer has no counterpart in t10fs or s5fs, and it is the first place ITS turns out not to be TOPS-10 with different field names.

The arithmetic

A sector is 128 words on every drive ITS supported. An ITS block is SECBLK consecutive sectors, and SECBLK is 8 everywhere. So far, so ordinary.

The blocks are then numbered within a cylinder:

NBLKSC == NHEDS * NSECS / SECBLK        ; MIDAS integer division
NBLKS  == NCYLS * NBLKSC

MIDAS truncates. For an RP06, 19 * 20 / 8 is 47.5 and NBLKSC is 47: the cylinder has 380 sectors, ITS reaches 376, and the last four are addressable by the hardware and reachable by no block number. Over 815 cylinders that is 3,260 sectors — 3.2 MB of a 300 MB pack that the file system cannot name.

itsfs drives prints the remainder for each drive, because it is not the same everywhere:

  rp04   4 sectors per cylinder unused -- 1644 sectors, 1644 KB, on the pack
  rp06   4 sectors per cylinder unused -- 3260 sectors, 3260 KB, on the pack
  rp07   0 sectors per cylinder unused
  rm03   6 sectors per cylinder unused -- 4938 sectors, 4938 KB, on the pack
  rm80   4 sectors per cylinder unused -- 2236 sectors, 2236 KB, on the pack

The RP07 divides exactly (32 * 43 / 8 == 172), which is worth knowing: a reader tested only on an RP07 would never notice this layer was needed.

Block to sector

FSDEFS writes the conversion out for the one block whose address is a constant, the MFD:

MFDCYL==MFDBLK/NBLKSC
MFDSRF==<MFDBLK-MFDCYL*NBLKSC>*SECBLK/NSECS
MFDSEC==<MFDBLK-MFDCYL*NBLKSC>*SECBLK-MFDSRF*NSECS

A SIMH or KLH10 image stores sectors in (cylinder, surface, sector) order, so the linear sector is (cyl * NHEDS + srf) * NSECS + sec. That is its_blk_sector(), and it is the only place in the project where a block becomes a sector.

How this was found, and how it is checked

The MFD is at NBLKS/2-1, which for an RP06 is block 19081. Reading that as a linear block number — 19081 × 1024 words — lands in the middle of a mail file. Running it through the formula gives cylinder 405, surface 18, sector 8, which is linear sector 154268, and word 5 of that sector is SIXBIT M.F.D. — the check word MDCHK exists precisely so that a program can ask this question and get an answer.

The two numbers differ by 3,244 sectors. There is no way to be slightly wrong about this: either you have the geometry or you have somebody's mail.

The test suite computes block 19081's sector a second time, in shell, from the drive constants — deliberately not by asking itsfs, because a wrong mapping would otherwise agree with itself — and then checks that dumping the block and dumping that raw sector give the same words. itsfs dump -s exists for exactly that: it addresses sectors and skips this layer entirely.

A "track" in ITS source is a block

disk.1228 line 48:

;	TUT	TRACK (BLOCK) UTILIZATION TABLE

and the file system code uses the two words interchangeably from there on. The MFD-slot arithmetic in QFL2 is commented J <= TRACK ADDR OF USER DIR and produces a block number. It is not a surface track, and nothing in the file system addresses a surface track.

This confuses every reader of the ITS sources once. It is written down here, and in itsgeom.h, so that it confuses the next one less.

The drive table

Every constant is transcribed from that drive's own parameter file in the ITS source; the derived quantities are computed rather than transcribed, so a drive added later cannot get the truncation subtly wrong by hand.

drive	cyls	spare	surfaces	sectors/track	blocks/cyl	blocks	TUT blocks
rp04	406	5	19	20	47	19082	2
rp06	812	3	19	20	47	38164	4
rp07	627	3	32	43	172	107844	9
rm03	820	3	5	30	18	14760	2
rm80	556	3	14	30	52	28912	3

The spare cylinders are XCYLS, "extra cylinders for spares, hacks, etc." — they are inside the image and outside NBLKS, which is why info reports both totals and why the MFD, at NBLKS/2-1, is not in the middle of the file.

The size of a full image identifies the drive on its own: no two of these are the same number of sectors. That is how itsfs info names a drive with nothing but the file, and why -d exists for an image that is not a whole pack.