JJ "João Jerónimo" wrote:
>>> Wolfgang Kern wrote:
>>>> But if you write your own HD-driver, the physical CHS-limit is:
>>>> 16 heads, 255 sectors, 65536 tracks(cylinders), and this gives
>>>> us 267386880 sectors (= 127.5 GB).
>>>> So if the the hardware can handle it then it works almost like LBA28
>>>> (max= 128 GB), ...except for the unconvenient calculation needs.
>>> But is this a DOS or a BIOS level driver?
>> You will encounter I/O-permission issues if you try this on windoze
>> or in a winDOS-box.
> Do this, what?
> What does windows have to do with the discussion? Of course you cannot
call
> the BIOS if you want to write a disk driver for windows, because windows
> runs in 32-bit Protected Mode, but the BIOS runs in 16-bit Real Mode,
and
> relies on RM segment arithmetic, default operant/address sizes, stack
sizes
> and memory addressing limits. A PC running in Real Mode can only access
1MB
> of memory and have no sup****t for paging, so basically can't call the
BIOS
> unless you return to Real Mode, or use Virtual 8086 Mode.
That's what I mean. DOS-executables running in a windoze-DOS-box can
access I/O if the Os grants permission (ie:'LOCK', Admin-rights...).
I disassembled sblaster.com used in my DOS-box of win98se, it direct
hooks interrupt 66/67h and rd/wr from ****ts e000h... without problems.
> Read Intel docs for details on differences between RM and PM execution
> environments:
> http://www.intel.com/design/pentium4/do***entation.htm#manuals
thanks, I'm with AMD and know the difference since a while ... :)
>> My first disk-editor ran on DOS 3.0, and had its
>> own HD-routines, so bypassing all limits from DOS and BIOS int13.
> Yes, but regular DOS programs that use file I/O typically use DOS system
> calls. This means that it's limited by the routines DOS relies on for
disk
> access.
yes.
> For example, DOS-based operating systems usually rely on BIOS disk I/O
> routines. So if, for example, this routines are limited to 8GB, of
course
> DOS disk I/O will be limited to this size, and the same holds true for
DOS
> file I/O.
sure.
> Physical CHS addressing allows a limit of some Gigabytes. However, the
> original IBM PC BIOS uses some rather stupid bit fields to pack the CHS
> addresses. This stupid bit fields make CHS addressing unnecessarily
limited
> to 503MB when used through the BIOS. However, BIOS developers eventually
> created a custom CHS addressing that maps virtual addresses to physical
> ones. This virtual addressing (and mapping) allows up to 8GB.
yeah.
>> All it did:
>> tem****ary replace HD-IRQ vectors (just to not confuse DOS)
>> fill the HDC-registers as desired (incl. command-byte as the last)
>> wait for acknowledge
>> read or write sector(s)
>> [find all required details in RBIL-****ts (01F0../0170..)]
> Although User Mode programs in PM-based operating systems don't access
the
> hardware directly because they are forbidden to do so (there's a general
> protection fault that notifies the OS is they try to do so), proper DOS
> programs were supposed and encouraged to do DHA if they needed it.
Right, the topic says int13 ....
But utility programs like format,defrag,wipe-clean,diskedit,partit ...
all need to override the given limits from INT13/OS and usually got
their own routines to direct access the hardware.
> However, "normal"/"****table" programs did not do so. I'm not sure, but I
> think really no one was doing file I/O directly even on DOS (exception
> perhaps for defrag and so): they used kernel's filing calls instead.
Yes.
>> I kept this very rare now used CHS functions in my PM32 Os, just for
>> the op****tunity to read from old drives when they become replaced.
> If you don't want to run your PM32 OS on old PCs, you can even drop CHS
> sup****t completely, because LBA uses linear addresses to access the hard
> drive. For example, if you want to address the 654th sector of your hard
> drive, you can refer to LBA sector number 653 (because LBA is zero
based,
> that is, the very first sector of your hard drive has number 0, just
like
> your RAM memory). However, you you use CHS addressing, you'll have to
> compute three numbers to build the address. This addresses would depend
> upon the physical geometry of the disk.
Almost right, except that disk geometry is a fiction on modern HDs anyway
and old CHS-disks wont do much when accessed via LBA.
The reason for why I keep my CHS-funtions is just to be able for hardware
upgrades (copy users data from old CHS-drives to a new machine).
__
wolfgang
|
