Dave Cotton wrote:
>
> I also have spent a frustrating week with UDMA problems including a
> totally unreadable disk (corrupted superblock).
> With Mandrake 7.0 and a UDMA 33 motherboard HDPARM reported 16MB/s with
> my WD205AA now with 7.1 this is down to less than 5MB/s. With the drive
> on my UDMA66 motherboard this drops to 3.5MB/s If I use DMA etc. I get a
> corrupted superblock. I cut the speed of the bus by 5% and that seems to
> have stopped the corruption.
> If you run hdparm -t /dev/hda with dma set there is a stream of seek and
> crc errors until the system resets the controller and disables dma.
>
> My question is what is your processor and motherboard. I have AMD 400
> and 500s and DFI P5BV3+ and K6XV3+/66 mbs.
>
> Dave Cotton
> Linux Autrement
> Avignon France
> +33 (0)4 90 16 07 89
If you search the archives for UDMA66 problems, you will find
most complaints involve WD drives. There is a reason. WD drives
have unusual timing requirements. 7.1 is the first MAndrake
distro to support UDMA66 out of the box, and the timing might be
a little too tight for the WDs. Also, WD does not actually use
Cyclic Redundancy Check (CRC) protocol but rather fakes it and
blows it off. The result is that there is nothing in the HDD
hardware/firmware to block a data stream corrupted by timing
chatter from being written to disk.
THere is a program that sets WD drives for UDMA66 available for
DL from WD, which might(tm) help.
In the mean time, on kernel traffic, the discussion crops up from
time to time that a possible solution is to restrict WD drives to
PIO upon recognition. No conclusion has been reached, but the
fact that such discussion is taking place should indicate
something to everyone.
Civileme
And of course Promise has actually made a driver available for
the HPT366 in source code. The driver may still need a few bugs
located, but the future looks brighter for stable UDMA66.
