Status of ATA support

Status of ATA support

[Marco Gerards]

Hi,

A while ago I added support for ATA support.  It wasn't complete yet,
but I rather have this in CVS than rotting on my harddisk.  Besides
that, it isn't *that* bad.  I will describe the outstanding issues
below.

First of all, this is mainly for i386-linuxbios.  On i386-pc we have
to disable biosdisk support because ata.mod and biosdisk.mod do not
like eachother :-).  Perhaps disk access via the BIOS will not be
possible/safe anymore after loading ata.mod.

PCI devices are not supported *yet*.  Same for controllers 3 and 4.  I
will work on this.

The code to detect the type of controller (PATA, SATA, PATAPI, SATAPI)
really sucks.  I will have a look at this.

There is no code to detect the amount of sectors for a CDROM.  I will
write this eventually.  Patches are more than welcome!

The biggest problem is that grub_get_rtc is used to measure how long
we have to wait for the hardware to settle.  This function return the
amount of ticks (1/18th of a second), this resolution is not
acceptable.  It will result in large delays in ata.mod because it will
wait at least 1/18th of a second on i386-pc.

For i386-linuxbios we still need grub_get_rtc.

More about this on my blog:
http://www.mgerards.net/blog/?p=34

--
Marco

Re: Status of ATA support

[Vesa Jääskeläinen]

Marco Gerards wrote:
> First of all, this is mainly for i386-linuxbios.  On i386-pc we have
> to disable biosdisk support because ata.mod and biosdisk.mod do not
> like eachother :-).  Perhaps disk access via the BIOS will not be
> possible/safe anymore after loading ata.mod.

We have to make sure we are not using BIOS for accessing disk after that
point. And ATA controller should be at stable state after resuming
normal operations OR we install custom int13h handler. Actually you
could try to determine what devices are handled by biosdisk and what are
not. EDD supports some queries that report on what controller disk comes
from.

Re: Status of ATA support

[Robert Millan]

On Sat, Nov 10, 2007 at 09:21:38PM +0200, Vesa Jääskeläinen wrote:
> Marco Gerards wrote:
> > First of all, this is mainly for i386-linuxbios.  On i386-pc we have
> > to disable biosdisk support because ata.mod and biosdisk.mod do not
> > like eachother :-).  Perhaps disk access via the BIOS will not be
> > possible/safe anymore after loading ata.mod.
> 
> We have to make sure we are not using BIOS for accessing disk after that
> point.

We already do (check for grub_disk_firmware_fini() function and
grub_disk_firmware_is_tainted variable).

But it has no effect on loaded OSes of course :-/

> And ATA controller should be at stable state after resuming
> normal operations OR we install custom int13h handler.

An int handler sounds like the sane thing to do..

-- 
Robert Millan

<GPLv2> I know my rights; I want my phone call!
<DRM> What use is a phone call, if you are unable to speak?
(as seen on /.)