On Sat, May 27, 2000 at 03:30:24PM -0400, Paul Gortmaker wrote:
> The following small patch is needed if ide is to continue
> to support stone-age ST-506 MFM/RLL disks. The question is
> whether we want to support them or not. If not, then the whole
> probe_cmos_for_drives() cruft should be deleted entirely.
I am going to give a much too long answer.
The answer is in 5 parts:
a) Why don't we throw out probe_cmos_for_drives() for IDE disks?
b) Do we want to support ST-506?
c) Must ide.c support ST-506?
d) Do we want to keep hd.c?
e) Does hd.c belong in the ide subdirectory?
In between I give some comments on your patch.
a) Why don't we throw out probe_cmos_for_drives() for IDE disks?
(i) As the comments in probe_cmos_for_drives() already indicate,
this routine has to go, we would be happy if it were gone.
(ii) However, throwing it out will give many people a different
disk geometry, and they might be unable to boot or have other
problems. Thus, each time something is broken and we have to
touch that code anyway, I try to touch the code in such a way
that it gets smaller. And if nothing is broken I try to avoid
changing anything.
Consequently, in my opinion your patch
> + if (sect == 17 || sect == 26)
> + drive->present = 1;
goes in the wrong direction.
[And concerning this patch, my very old docs all say that the
number of sectors/track is 17. But my old docs know about quite
some variety - I see 25, 26, 33, 38 at first sight, there may be more.
So, given only my docs, the choice of 26 seems a bit arbitrary.]
b) Do we want to support ST-506?
Another matter is whether we want to support ST-506 MFM/RLL disks.
My answer is a qualified yes. These drives still exist, and are
being used both in obscure countries and by people who love old
hardware - it would be a pity to scrap support, especially since
almost nothing is needed to continue support. On the other hand,
I see no objections against requiring someone who wants to use
2.4.0 on a 10 MB MFM disk to supply boot parameters.
In other words, I removed the "drive->present = 1" since it was
wrong. You want to reintroduce it arguing that "sect == 17"
protects us. But it does not. The main problem here is that
in this routine we only know that the BIOS sees a disk. But this
disk may well be a SCSI disk. And also a SCSI disk can have
17 sectors/track.
It is better not to add obscure heuristics and just say that
people with a drive that would not otherwise be detected
have to give explicit boot parameters.
c) Must ide.c support ST-506?
You write (effectively):
if (id == NULL) { /* Old, non-IDE drive */
drive->capacity = drive->cyl * drive->head * drive->sect;
drive->select.b.lba = 0;
return;
}
If that is all required to maintain support for old drives in ide.c
then I do not object.
d) Do we want to keep hd.c?
Of course. It is so much smaller than ide.c. Regardless of whether
one has ST-506 disks - hd.c remains useful.
e) Does hd.c belong in the ide subdirectory?
These days there is some confusion about the right CONFIG_*
parameters required in partitions/msdos.c and ide/ide-geometry.c.
The setup was wrong, and has been changed but is still wrong.
I think there would be less confusion if we move hd.c back to
drivers/block. It has nothing to do with ide.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed May 31 2000 - 21:00:18 EST