Re: chk drive for errors?

On Mon, 23 Jul 2007 23:21:26 +0100, David Bolt wrote:

On Sat, 21 Jul 2007, imotgm wrote:-

On Sun, 22 Jul 2007 01:22:48 +0100, David Bolt wrote:

Whoever said it was playing the idiot? And what was stopping them from
building their own packages, using the TexStar source packages as the
basis?

That was TexStar's line. There in the repos, help yourself.

If you have a partition to spare, I'd suggest a full installation,

It's a VM. It'll have a virtual machine all to itself :-)

I run VMWare server also. I was just making the suggestion for checking
Beryl. I don't use it either. I like having 20 desktops to spread out on.

And I guess you need the pata_it821x driver. Now, do you really need to
eliminate every pata_* module, or just some of them?

No, I need the older it821x, which is what I have. The pata_it821x makes
the IDE drives appear as SCSI drives, the same as SATA drives, and limits
the number of partitions to 15.

That's an old limit that SCSI had. Moving SATA, and now IDE, under SCSI
means everything gets that 15 partition limit. LVM is one way round that
limit.

Also another layer of complexity, unneeded up until now. What happened to KISS?

I have some IDE/PATA drives with 18
partitions, so that's a no go from the start, even if the drivers worked
properly.

IIRC, there was discussion on the Factory list about this when the swap
was made, and the suggestion was to make the kernel use the older
modules.

But the process breaks the older modules. I can actually boot the last
updated default kernel with the included older module but my entire hde
drive still disappears. It's pretty hard to read or write to a drive that
the system can't see. My hard drives are all in removable caddies. It
doesn't matter which of ten drives I attach to hde, they are all invisable
with the default SUSE 10.2 kernels. All are properly seen by all other
distros, SUSE up through 10.1, and by my kernels in 10.2

IIRC, there was mention of patching libata[0] so it would
support more than 15 partitions. I don't recall exactly what the
conclusion was, or whether there was actually going to be a patch, but
I'm sure a search of the archive will show it up.

Last I heard was, "If you need more than 15 partitions, use LVM."

With drive size going up constantly, I have a hard time grasping
why it was OK to allow drives, less than 1GB in size, to be able to have
63 partitions, but a new 1TB drive should be limited to 15.

I think the idea is that, nowadays, people would be using LVM and
putting (almost) all the various partitions inside a logical volume.

I liken that to the $64.00 answer to a $2.00 question. Offered solution to
unnecessary problem !=KISS.


The logic here escapes me. Also hdparm automatically gets broken in the
process, and sdparm doesn't work right with PATA drives either.

You did tell it to use the ATA transport ( -t ATA ) ?

No one has mentioned that, when I posed the questions, and posted commands
given, and error messages received. What I got was, "We're still working
on that."

Because the drives had different physical properties, unix always
treated IDE, and SCSI, as separate entities. With SATA mostly replacing
SCSI drives, they were treated by Linux as SCSI, to separate them from
PATA, so you'd know which was which, even though the are in fact, still
IDE drives, with a serial interface.

Have you looked at the ATA commands? AFAIK they're almost a subset of
SCSI.

Yet how the commands are processed is enough different that they were
treated as seperate entities, and each works well as long as the
difference is addressed. If they lumped all drives as simply "hard drives"
and the new system actually worked, I would have no real objection with
that, but the system clearly is not working on my hardware, and it is
common hardware. There is nothing exotic about this hardware, and it has
worked fine up until this change.

My total argument is this;

When a system is composed of two parts and works without problems, through
the interaction of the two parts, and one of the two parts is changed,
while the other remains constant, if new problems suddenly arise, the
cause of the problems surely resides in the part that was changed, not the
part that remained constant, and those that made the changes need to
correct that which causes the problems before claiming the system to be
stable, and usable.

If I have no viable choice in the matter, as neither of the choices I am
given actually works, and must therefore abandon my otherwise fully
functional, two year old hardware, and also abandon the practices and
procedures that I have developed over years of use, in order to continue
using this changed software, how is this different from Vista?

Now the kernel boys, claiming the older IDE code is getting krufty,
decided all drives should be brought under libata, and treated as SCSI,
so they only have one drive source tree to worry about.

I can understand that. Less stuff to maintain, and less duplication of
efforts.

My take is, that's just lazy. If the code is krufty, fix it, don't break
the functionality of the drives, for coder convenience. Same goes for
USB.

In most common situations it doesn't the functionality. It's only when
you get someone creating a larger than normal number of partitions that
there's a problem.

My present drive hde has only four partitions, and the system can't even
see it with the current SUSE kernel using libata. It locks while booting
if it loads pata_it821x. It's blind to hde if it loads it821x. I consider
that a problem.

Windows does not treat USB as SCSI, but as an entity with it's own
identity.

That's probably because the Windows developers don't like the idea of
talking to each other and sharing ideas. On top of that, it's all been
bolted on, with additional bolt-ons with each new release.

There's nothing wrong with bolt-ons, when they work. Most of a Linux
kernel is now a series of bolt on modules. You bolt on the ones that work
for your hardware, and leave the rest in the toolbox. All of SUSE's USB is
compiled as bolt-on modules. I don't know of any distro that doesn't
compile USB as modules.


Most of the problems with USB on Linux come from insisting on
using the SCSI tree to control USB, and trying to work within the confines
of that format.

Well, so far, I've not had a problem with USB drives.

My problems are relatively minor, but they do include drives that suddenly
disappear, while I'm writing to the drive, then return with a different
drive designation, and different mount points. This happens to some
extent with all Linux distros. It does not happen under Win2K, on this
same machine, using the same drive, and I hate having to say that, because
I hate running Windows, and have no need to, except to test things like
this. Win2K is eight years old, but it's USB, and as an aside, also it's
UDF CD/DVD packet writing ability, through third party apps of the same
vintage, is more mature, and more stable than in present day Linux. That's
a long time to play catch up, and not yet arrive. I've never run XP, never
even seen it run, until a couple months ago. Whether there are any
improvements in these two items with XP, I have no way of knowing, but I
suspect there are.

I discount games in any Windows vs Linux, argument, because it's a
straw man. I dislike, intensely, having to admit that any basic process of
the OS is handled better on, any version of, Windows than Linux, in any
give contemporary time frame. I really hate it when an eight year old
Windows appears superior, on even trivial matters, than a current Linux.

I'm on the bugzilla mailing list and just did a search for your nick.
It's not like it's a common nick :-)

I must admit that I'm pretty easy to find on google group searches too. I
could have filed the bug reports under my real name. ;-)

I've got to be the only person out here that could use his real name, to
hide his real identity. My own mother has never called me by, or ever
referred to me by my given name. None of my friends do, nor either of my
wives. or any of my children, or grand children. I've had both personal
and business checking accounts that would not cash any checks that had a
signature that was any form of my real name, because that was a sure sign
that it was a forgery.

On that note, I once knew a woman, (still do, for that matter) whom I
visited in the hospital, on the occasion of her first son's birth. I asked
her what she named her son, and she said, "Mathursalees". This being
somewhat unusual, the obvious question to me was, "Why?" but her answer,
"So we can call him Jessie," left me almost speechless. (Imagine that)
After a moment's stunned silence, and seeking the logic that would make
sense of this, I asked, "If you want to call your son Jessie, why not name
him Jessie? Where does Mathursalees fit into this picture?" She looked at
me like I was the dumbest form of stupid on the planet, and answered, with
a quite superior tone, "My grandfather's name was Mathursalees, and
everyone always called HIM Jessie." Now, there must truly be some logic to
that, indeed, because Mathursalees is in his mid thirties now, and I have
never heard him called anything, but Jessie.


Have you tried the vanilla kernel.org sources and used the SUSE .config?

Not actually, as I'm still trying to work with the bugzilla boys, so want
as much of the SUSE kernel as possible, patches and fixes included, just
not the broken parts.

I'd start with the vanilla kernel and see if that works. If it does,
it's a SUSE patch that's at fault. If it doesn't, you need to be talking
to the kernel guys.

Good point. I hadn't thought of that aspect. I have some irons in the
fire, at the moment, but I'll try that as soon as time allows.

I'll wear mirrored glasses next time.

Naughty, naughty David. ;)

You been talking to my wife? She keeps saying that to me as well ;-)

Damn, now we're going to have to stop meeting like that. The old man is
getting suspicious.


[0] That speel chucker of mine still wants to change that to labia. I'm
sure it's trying to shift my train of thought :-)

My train was derailed a few minutes ago. I just got an e-mail from my
sister telling me her husband died last night. No details beyond he was in
hospital, about to be released, and died while dressing, to leave. Cause
of death, uncertain at the moment. I may be gone for a bit.

--
imotgm
"Lost? Lost? I've never been lost... Been a tad confused for a
month or two, but never lost."


.