--- Ron Stodden <[EMAIL PROTECTED]> wrote:
> Eugenio Diaz wrote:
> >
> > Well, despite all that effort in documenting the problem, no action has
> been
> > taken to solve it.
>
> But action has been taken, in the ongoing development sense, but not
> the field support sense. Maybe by the kernel team rather than
> Mandrake. For example I installed the latest Cooker yesterday
> (kernel 2.4.1-12mdk) and it happily installs and runs through the
> Promise Mass Storage chip.
>
> > As I test new kernels, I see that some times they work, and
> > when the next update comes, it does not work again; which leads me to
> believe
> > there is a real problem with the ide patches. And under no circumstance is
> this
> > an intermittent hw problem, since 2.2.16 work every time, and the ones that
> > don't fail every time. I normally just update my rpms, and have an install
> that
> > went from RH 5.2 to Rawhide to Cooker; and for a while I thought the
> problem
> > was because of that, but when I tried doing a boot from the install image,
> I
> > got the same problem, which means that anybody that has a Promise (may be
> with
> > WD drives) based ATA66/100 will fail to install Mandrake. Isn't that
> important
> > enough to get some attention?
>
> Yes, it certainly is. I for one complained loudly and got this reply
> from Pixel:
...
> QA? Probably all it would take is a little note attached to all
> unsold copies with the append string to enable the Promise IDE
> channels as IDE 2 & 3 so that the installer can find these drives.
> Or showing how to disable both the existing IDE channels and with the
> required lilo append string to make the Promise IDE channels appear
> as IDE 0 & 1?
Ron, you are confused. The problem I described in bug #525, is not about kernel
parameters sent with the lilo "append" string. It is about the kernel hanging
on a DMA timeout during the "partition check", and that occurs way *after* lilo
gave it the correct parameters, and it already detected the controller and
drives.
> Muse: It is strange that Promise chose that way of doing it. On
> Windows it is acceptable because shifting your hard drive farm from
> IDE0 & 1 to IDE 2 & 3 will cause the same drive letters to be
> assigned to the same partitions. But on Linux that concept does not
> apply, and Promise should have known that. Changing your disk farm
> from hda, hdb, hdc and hdd to hde, hdf, hdg & hdh is not trivial
> since none of those Linuxes will run any more and it takes work from
> another running Linux located somewhere else to fix them up. IMHO,
> Promise should be taken out and shot! Their mass storage chip
> should have taken over all 4 IDE channels at ATA100.
Promise has nothing to do with this. It is a software problem. Linux, or any
other OS for that matters, deals with hardware by:
1. detecting it
2. detemining what it is, and applying a correct driver for it
3. the driver must map the device to an apropriate common name, like device
files: /dev/hda etc.
4. the OS uses the driver to talk to the device.
And in the problem we have, here is what happens:
(1) Linux detects it fine, it is a PCI card what did you expect.
(2) Linux does not really know what it is, so you must tell it to use the IDE
driver, by providing the lilo append string. It is you who is making the
mistake of telling the ide driver you want your addon IDE controller on IDE2
and IDE3, when instead you wanted it to be IDE0 and IDE1.
(4) Linux is able to identify the drives on the channels correctly, but when it
tries to use DMA to talk to them for the partition check, it hangs (in fact,
this is not completely correct, it hangs on just one drive, and no it is not a
HW problem).
So how is this bad design on a Promise's part? They don't write the linux IDE
drivers, or do they? Of course, they could help more by helping the kernel
developers provide a better driver.
Anyway, read my bug #525 report, and you will se that I have it configured to
use the addon Promise card as IDE0 and IDE1, and the motherboard IDE controller
as IDE2 and IDE3. In fact I have my CDROM and CDRW as hde and hdf on the mobo
controller, and four more drives as hda, hdb, hdc, and hdd on the Promise card;
you just need to switch the addresses and labels in the append script. I did
not reinstalled my system after I bought the Promise card, I just figured out
how to tell Linux which controller to use as what, and it is as simple as a
kernel module parameter, the "append" string.
=====
________________________
Eugenio Diaz, BSEE/BSCE
Linux Engineer
[EMAIL PROTECTED]
__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail - only $35
a year! http://personal.mail.yahoo.com/
