What number is a cyl boundry

On Mon, 21 Aug 2006 06:17:40 +0000, ANC

It isn't about* new *ideas, as much as about the historical reasons for why things are the way they are with PC architecture, and how we got here. For one thing, disk capacity has been growing faster than Moore's Law. We like extra disk capacity, so we'll find a way to use it- whatever it takes. I'll explain different methods for disk addressing below.

This is the problem as I understand it with this bug. In simplest terms, your OS needs to know where it is located on your disk. It reads the parbreastion table to find the parbreastion where the loader and the rest of the OS is located. When multiple OSs are used, they can be using different buttumptions about the disk geometry and different methods of addressing. The start of a parbreastion can be in different places, depending on what buttumptions are used. That is a big problem, especially if an OS is using a "mixed mode" (#3, below.)

Disk addressing method 1: Historical (C,H,S) Problem: Find a given sector on a hard disk. Solution 1: Use BIOS Int 13h to retrieve sector. This was the method used by MS-DOS. In this system, disk geometry (and offsets to sectors) can be expressed using an ordered triple of three integers:

(Cylinder,Head,Sector).

The original IBM BIOS imposed limits for each value. This imposed an overall limit for the capacity of hard disks. When each value is at its maximum, there is a disk limit of 8G (+--)

Cylinder: (10 bits) (0-1023) Head: (8 bits) (0-255) Sector: (6 bits) (1-63)

Disk capacity limit: 1024*256*63*512= 8455716864 bytes

Actual disk capacity exceeded this limit in the 1990s. As the disk capacity starting approaching the limit, fictional disc geometry was introduced just to maximize each value in the ordered triple. For example, the drive would report that it has 256 heads, even though that it had nowhere close to that many heads. In this era, hard disks had a physical geometry and a "logical" geometry. The fictional geometry was called the logical geometry.

Disk capacity is now much greater than 8G. Physical disk geometry can still be expressed using (C,H,S) with no numerical limits on the values. Typically, the last cylinder is now much larger than 1023.

The difference-conflict arises when different OS use different fictional drive geometry. Typically, one OS will be using 16 heads and the other will be using 256 heads. The location of a "cylinder boundary" will not be consistent in that case. The error message you are getting is symptomatic of this problem.

Addressing Method 2: LBA LBA is a system where disk sectors can be retrieved from the beginning of the disk using a simple value for the offset sector. The size of the disk is limited by the maximum offset. IIRC, LBA can use 48-bit addresses now.

That gives a maximum disk limit of 2^57 bytes.

Because LBA does not require (or use) the (C,H,S) ordered triple, and does not use the cylinder to compute the address, a "cylinder boundary" is an artifact and does not appear to be relevent. However, the loaders for OSs would like to see that the parbreastion table matches the buttumed drive geometry in use. So, by convention, disk parbreastions must still start on a cylinder boundary. Even though parbreastions could be described as an range of sectors beginning at an arbitrary sector offset, they are constrained (by convention) to starting on a cylinder boundary.

Addressing Method 3: Mixed mode.

Some loaders (in the past) have used (C,H,S) addressing, while the OS used LBA. lilo used to have a limit like that, but the Linux kernel would be using LBA to address the disc. In any case, this mixed mode increases the chance for a conflict somewhere. The NT-2k-XP loaders may use mixed mode, I don't know for sure.

Note: comments inline.

The wikipedia has some good basic explanations about disks, too.

-- Douglas Mayne