On Monday 19 February 2001 07:17, you wrote:
> civileme wrote:
> > OK let's begin by comparing apples to apples.
> >
> > Most IDE setups, including Promise, are correctly detected and set up by
> > the installer and handled properly by the kernel. This includes ATA/100
> > or UDMA5.
>
> Sorry about this, but the 7.2 kernel just does not see IDE2 or IDE3,
> which is where Promise lives.
> This applies to both the installer and the installed runtime. I
> don't know where your erroneous information might be coming from.
It comes from actual, installed running systems.
Now what is _different_ about yours?
> Worse, neither lilo nor grub know about these IDE channels, so you
> MUST at all times have a disk on one of your 2 mobo IDE channels,
> otherwise there is no secondary boot loader (lilo gives no error
> message from its failure to write it to a non-existent hda) and all
> booting will stop at LI. Booting from floppy and re-executing
> /sbin/lilo makes no difference - still stops at LI.
OK, I'll try that. But if it works on my AZ11 with all drives on 2 and 3,
under their Promise ATA/100 controller, does that mean it should work on your
system too?
>
> > Some folks have more IDE channels than are autoprobed. Drive i
>
Again, your system, Ron. Not all. My drives were autodetected on channels 2
and 3.
> and e,f,g,h
>
> > on 7.2 will
> > never be autodetected... It has to be configured after boot, because
> > while major and minor numbers exist, it is not autoprobed (Run modprobe
> > -c for yourself).
> >
> > Some folks have streaming problems on CD burning. I know this doesn't
> > sound like the same problem, but it appears to be related to one or more
> > of the problems I will categorize here.
> >
> > 1. ASUS A7V -- to detect the embedded Promise controller you have to
> > boot the CD in rescue mode and
> >
> > lspci -vv | less
> >
> > and scroll through the output. The current advice on mandrakeuser.org is
> > to do
> >
> > linux ide2=0xa000 0x9802
>
> This is totally unacceptable, because there is no way to automate it
> that I am aware of since at that point there is no file system. To
> have to type in such gobbledigook at every boot is plainly
> unacceptable.
>
It is only for the boot--during bootloader installation we have to type the
ide2=0xa000 0x9802 into the Append blank of each linux system booting
(Usually linux linux-nonfb and failsafe) After that, the process is
automatic.
> > (which sets the controller up to run drives e and f) and the 0xa000 and
> > 0x9800 are the first of five hex numbers attached to the irq where the
> > drives set up. It is done for ide2 because the oters are already
> > assigned by autodetection and there is a need to keep things simple. Of
> > course strings to the install kernel could reassign all the drives if
> > desired.
>
> But the reassignment is only for the duration of the install.
>
Now that is a very valid point! The reassignment will not work for boot
because the append strings have to be loaded off the disk and it has to be a
specific disk with the then existing assignments--instant freeze on partition
check. With a flexible enough BIOS, this is not a problem. I have a
motherboard which has a LS120 on /dev/hda and a Maxtor on /dev/hde and LILO
boots from /hde with no problem, but if I tried to assign that to /dev/hda
--Yeech-- That could be interesting to try....
> > And the
> > bootloader is a followup--the append strings must be given to LILO or
> > Grub for transmission to the boot kernel.
> >
> > 2. LG CDRWs seem to have a problem with the ATA/100 backport for 7.2,
> > because kernels compiled without them run and burn CDRWs fine while those
> > with them balk at setting DMA on the CDRW which can, but not necessarily
> > does, break streaming. Results with burning are mixed, sometimes
> > working, sometimes not. This occurs whether or not there are any other
> > IDE devices on te system and with selected motherboards. Aditional
> > reports are welcome.
> >
> > 3. Setting DMA on some systems (and this happens on so few that we have
> > little data) simply does not work for one or more drives,
>
> This ws true of my old AIR 6ABX motherboard, and is one reason why I
> have just replaced it with a Gigabyte GA-7ZXR (and an AMD Athlon
> 900MHz CPU).
OK thanks for the data... That is something we can test. (and we did not
have that combination here).
>
> > and we have a
> > partition check that hangs the system. It might be related to a CDRW or
> > to a Promise ATA/66 Controller. More data is welcome. Particularly
> > welcome is any data of anyone able to reproduce this error _without_ WD
> > drives.
>
> I can on that old mobo and IBM drives, but I shifted my CDRW to the
> other IDE channel and all problems went away since I always master
> CDs to the same partition.
>
Yes, we have reports of streaming errors and DMA settings failures when
CDROMs and CDRWs share an ide channel. But reducing the variables to isolate
a cause or causes has been a nightmare, especially wih very sparse
occurrences and even sparser data.
> > Unfortunately, this brand seems to be a factor in the failures, and it
> > does have hardware differences from other drives. One empirical factor is
> > that setting idex=noautotune sometimes stops this behavior (where x is
> > the offending channel).
> >
> > Civileme
