Re: IDE HDD control
From: wolfgang kern (nowhere_at_nevernet.at)
Date: 09/27/04
- Previous message: Scott J. McCaughrin: "Re: IDA beginner"
- In reply to: Cameron Gibbs: "Re: IDE HDD control"
- Next in thread: wolfgang kern: "Re: IDE HDD control"
- Reply: wolfgang kern: "Re: IDE HDD control"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Mon, 27 Sep 2004 11:09:59 +0000 (UTC)
Hi Cameron,
[]
| > | Primary IDE Controller, master drive:
| > | Address: 1F0-1F7h
| > | 8 Bytes
| > Same ports for 'Primary slave drive'.
| That basically answers my question.
| Thanks!
| > port 1f0 is four (two on older) bytes wide.
|
| I was thinking that some knowledgeable person would pick up on the 8
| byte reference.
| Speaking just about the older interface (negating newer improvements):
| I was aware that the register at 1F0 was actually 16 bits making the
| interface 9 bytes (I should perhaps say 8 registers).
|
| However I was not aware that one could have 32 bit registers mapped to
| the 16 bit I/O address space.
| Might the 16 bits for the I/O space refer only to the number of
| possible I/O addresses?
Right.
Data-bus width: 32 bits
Address-bus: 32(36/40) bits
I/O-address limit: 16 bits
I/O-space and MEMORY share data and address bus,
but use separated bus-control-lines (I/O vs. MEM RD/WR cycles)
| Otherwise maybe the HDD controller just handles the doubleword hit
| using the 16 bit register. Squeezing it through one clock strike or
| whatever?
32-bit ports allow faster HD-RD/WR (even not really double).
Most newer (~+1996) controllers have port[01F0] 32 bit wide,
without writing to ports [01f1,2,3] (but work also with 16 bit),
while the same chip may offer the ability to write
16 or 32 bits to [01f4&5 or 01f3..01f6] as this ports
are configured for byte access.
| > Some controllers allow dword writes to 01f3 (LBA 0..27)
| > bits 28..31 (high nibble of 01f6) are 'E' for master and 'F' for slave.
| Any idea how to determine what HDDs would accept such a write and what
| HDDs would not?
It is merely the IDE-controllers ability, even ATA-2 mentioned a
16/32-I/O bit in the identify device block (obsolete since long yet).
So I check this by trying first to read 32-bits from 01f0
(after any command which set the data-ready bit of course),
if the resulting high-word is identical to the low-word
then it is for sure that it supports only 16 bits.
Write (after wait_ready) a dword to 01f3 and read it back
as four single bytes, compare and you know.
| While I am asking that, how would one go about determining the
| presence or absence of a drive?
1. reset drives (clear any previous error condition),
here you can also decide to enable legacy IRQ(15/14) or not.
2. enable timer interrupt (at least..)
wait_ready or timeout (I still give it two seconds, just for oldies)
3. send a identify-device command (this works also if in sleep-mode)
4. wait_ready or timeout or error
5. check 'data-request bit' (data ready for read here yet)
6. if no_error,no_timeout then read the 512 bytes into a buffer.
if error or timeout then there is either no drive or no controller
or in some rare occasions a hardware or setup fault (like PCI-only).
7. optionally repeat 1..5 with a 'indentify packet' command for CD-drives.
ready: b7=0 b6=1 b0=0
error: b7=0 b6=? b0=1
data ready: b7=0 b6=1 b3=1 b0=0
timeout: b7=1 time elapsed
BUT: most controllers return b0=1 on all slave activity,
so the error indication is the 'masters job' only.
IIRC yet I used and 58h| cmp 58h |jnz error before all I/O-RD/WR.
| No further confusion at this time.
| Thanks wolfgang
We sure find some more confusing things in this wild grown circuits :)
Always welcome,
__
wolfgang
- Previous message: Scott J. McCaughrin: "Re: IDA beginner"
- In reply to: Cameron Gibbs: "Re: IDE HDD control"
- Next in thread: wolfgang kern: "Re: IDE HDD control"
- Reply: wolfgang kern: "Re: IDE HDD control"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
