Determining maximum partition size on a hard disk

I am trying to calculate the maximum size a partition can be on a hard drive and I ran into some problems I don't fully understand.

First I found that the maximum size of the drive Linux reports is not the maximum size I get when I calculate it from the drives geometry. Secondly, the total drive space reported by linux is not the amount available for the maximum partition.

For example, I have a 4.3Gb disk. The drives geometry is 525 cylinders, 255 heads and 63 sectors (525 * 255 * 63 * 512 = 4318272000 or 4.318Gb).

This is an IDE disk so I found in /proc/ide/hdx/capacity a block size 8439184, which when divided by 2048 is 4120.7, ~200Mb less than what I calculated as the disk size.

Finally, using a program such as fdisk or sfdisk I create a partition manually to be the maximum amount allowed, which turns out to be 4217041 blocks, or, when divided by 1024 is 4118.2.

I assume that the difference between the maximum size that linux reports and the maximum partition size is due to linux leaving room for a MBR and such. If so how do I go about calculating this? Also why does the size of the drive I calculate out by using the drives geometry differ so much from the amount Linux reports? I thought maybe Linux set 1MB=1000k but that doesn't seem to case.

Any information or a push in the right direction would be appreciated, thanks!

Regards,
-Nick DeClario

I don't know why you would want to divide by 2048. Multiply by 512 and find 512*8439184 = 4320862208 bytes. Since that is more than you thought you had, be happy.

No. There are rounding differences. The disk capacity is not an integral number of cylinders and you lose if you insist on alignment.

Find a lot of details in the Large Disk Howto.

Well, 1 M = 1000 k by definition of the SI system of units. This has nothing to do with Linux.
But if you are confused about units, just compute in bytes.

While it is true that M = 10^6 and k = 10^3, surely that doesn't apply to byte quantities?!? At least I have always interpreted 1 Megabyte = 1024 kilobytes = 1024*1024 bytes, and I think the poster meant the same when writing 1MB = 1000k...

If we are nitpicking, he probably should have written 1MiB = 1000kiB. [Feel free to correct me if I am wrong, but IIRC, Mi and ki are the abbreviations for 2^10 multiples rather than 10^3...]

Anton

  "Nothing succeeds like success." - Alexandre Dumas
-- 
Anton Altaparmakov <aia21 at cam.ac.uk> (replace at with @)
Linux NTFS Maintainer / WWW: http://linux-ntfs.sf.net/
ICQ: 8561279 / WWW: http://www-stu.christs.cam.ac.uk/~aia21/

Ah, an old discussion - I imagine it must have occurred here a few times.
Anyway: k is the prefix for 1000, M is the prefix for 1000000. Always.

But there is a natural sloppiness, where people round numbers when either the precise size is unimportant, or is clear from the context. If xyzzies come in boxes of 27 and I say that I bought a hundred xyzzies then that probably means that in fact I bought 108.

Thus, our old PDP8 with 4k of memory had 4096 12-bit words, that is, 6 KiB in modern terminology.

You see two very different situations here: "precise size is unimportant" and "precise size is clear from the context". When someone says that she bought a 10 GB disk I will assume that the precise size is somewhere between 9.5 and 10.5 GB. This is the "rounded" case.
When someone says that she bought a 32 MB SIMM I will assume that this was 32 MiB, that is, 2^25 bytes, since, unlike disks, memory modules tend to come in power-of-two units. This is the "clear from the context" case.

Andries