This might be no news to some but here are some observations about
the Promise 20267 chipset fitted to a Gigabyte Titan mobo.
1. No success using 'harware' RAID, so I set the BIOS to make
the 20267 look like another 2 ATA devices. Used a suitable
"append=" line when booting to get ide2 and ide3. It's a
slippery little gadget sometimes appearing at IO address
starting at 0x9400 and other times at 0x9800, depending on
what features are enabled in the BIOS.
2. It doesn't like to share interrupts. On my mobo it got IRQ 11
which was also assigned to a USB device. Produced register
dumps when the kernel was booting. I disabled USB on the mobo,
the 20267 got IRQ11 all to itself and was happy to start.
3. No DMA. The IDE driver that comes with 7.3 can't talk to the 20267
in DMA mode. It runs in PIO mode only & seems to run reliably.
"hdparam -t /dev/hde" produces terrible figures. The same HDD
(60GB Maxtor) produces good figures when connected to ide0.
4. Geometry. The BIOS does understand more than two IDE controllers
so if you want Linux to use a specific geometry, its more
"append=" line work.
In summary, the 20267 chipset is bad news and not worth the trouble.
There seems to be some hostility to this chipset by Linux developers
because Promise are secretive about the chipset specs. The secrecy
could be that Promise are protecting their valuable IP, but could also
be that the chipset is not much good and Promise don't want the
tough judges to find out.
-------------------------------------------------
This mail sent through IMP: http://horde.org/imp/
