raw disk copying?

On Tue, 2006-04-25 23:15:14 -0400, Doc Shipley <doc at mdrconsult.com> wrote: These days are gone. Long gone. The C/H/S layout of a disk isn't meaningful since years. In early days, these represented physical characteristics of a disk; operating systems had been programmed to use these directly. As a result, eg. the well-known DOSsish partition table evolved, that encoded the partition layout in C/H/S terms. Disks got larger, which typically involved the cylinder count to rise up. Unfortunately, the partition table as well as the ABI to the disk controllers at that time only had 9 bits (1024 values, 0..1023) to encode cylinder count, so something new had to be implemented. That was when actually two things were done: the controllers _internally_ started to not use C/H/S geometry to access the drive, but linear addresses. (So all C/H/S values are converted to linear addresses irrelevant of the operating system's access mode.) Additional to that, fake plattern were simulated: there were 6 bits (64 values, 0..63) reserved for heads, but no real disk would ever use 64 heads (famous last words...) So "newer" drives scaled their size by claiming to have more hads than there are in reality, with the result that the C/H/S addressing a "legacy" operating system is using doesn't match the physical layout anymore. Linux doesn't use the C/H/S values internally anymore, except for reporting these legacy settings to the user once they're read out from the drive. (Though, these values don't make much sense anyway.) There is, unfortunately, one point in the boot-up phase where they are still used: that's when the BIOS reads the partition sector (which also contains the very first modifyable boot code). For legacy issues, the partition table still contains C/H/S addresses and the BIOS honors them. This is why, on "real" PeeCees, you'd keep your boot partition below 8GB: 1024 Cylinders * 64 Heads * 256 Sectors * 512 Bytes. [Note that there are bugs in many early BIOSes that will crash with accesses beyond typically at limits beyond 2GB, 1GB or even 512MB.) This basically means that all data the PC needs to start up (kernel, boot code, extended partition table, initial ramdisk) needs to be within the 8GB addressable limit. (To make things even worse, many today's BIOSes implement non-standard extensions to allow access to the full size of today's disks. This means that a disk which used to work (as in: boot) in one machine won't neccessarily boot in a different box.) The cylinder boundary is only a legacy thing. With today's linear addresses, that's no problem. Indeed, it's only needed due to DOS bugs. A PeeCee'ish partition table allows partition boundaries not matching a cylinder boundary, but DOS won't probably be happy with it. Eg. if you use some Linuxish fdisk variant, they'll probably allow you to create such partitions and possibly warn about them. A different C/H/S layout (for disks of otherwise the same size) will only matter in the boot-up phase, IFF the BIOS doesn't implement extensions to just ignore the whole C/H/S stuff and use linear addresses. For a real operating system (which uses it's own drivers and not the BIOS callbacks), this doesn't matter at all. And finally, people realized that the C/H/S layout is overcome: There's a new standard (though not yet really implemented beyond the scope of some Itanium or i386 Apple machines) called EFI (Extended Firmware Interface) that fixes a lot of old time's shortcomings. MfG, JBG -- Jan-Benedict Glaw jbglaw at lug-owl.de . +49-172-7608481 _ O _ "Eine Freie Meinung in einem Freien Kopf | Gegen Zensur | Gegen Krieg _ _ O f?r einen Freien Staat voll Freier B?rger" | im Internet! | im Irak! O O O ret = do_actions((curr | FREE_SPEECH) & ~(NEW_COPYRIGHT_LAW | DRM | TCPA));