On Mon, 2010-02-01 at 14:14 +0100, maximilian attems wrote: > On Mon, 26 Oct 2009, Tom Jägermeister wrote: > > > Hi Ben, > > > > this issue was almost a year ago but I can remember it as it was quiet > > annoying. > > It turned out that I had to use module 'ahci' instead of 'pata_marvell', I > > found the hint in the kernel config. However in the previous kernels > > (before > > 2.6.26-12 pata_marvell had worked alright). > > Unfortunately pata_marvell seems to be preferred over ahci for the > > mentioned > > controller, so I had blacklisted the module and built all later kernels > > without it. No more problems after ... > > closing as this indicates only a stable device names bug, > which is userspace related and should be fixed by newer d-i that uses > UUID and maybe soon on upgrade too.
This is not the problem. The problem is that these controllers use a single PCI function for all ports, implementing both AHCI and legacy HD interfaces. (Most PATA+SATA controllers have separate PCI functions.) So one driver must bind to this function and handle all ports, but instead we have two half-working libata drivers for this hardware: - pata_marvell uses the legacy HD interfaces to drive both PATA and SATA ports (despite its name), but apparently doesn't work with the SATA ports on the 6145. - ahci uses the AHCI, which is for SATA ports only. Since it doesn't support the PATA ports (bug #507432), it no longer binds to these controllers by default (since upstream version 2.6.27 and Debian version 2.6.26-12). In order to use it, you must blacklist pata_marvell and set the ahci module option marvell_enable=1. Since pata_marvell doesn't seem to work properly with the SATA ports on the 6145, users may currently be *required* to change module parameters just to make it work (and then without PATA functionality). Dan Alderman has also run into this problem, and pointed me at the vendor driver <http://jetwaycomputer.com/download/Drivers/ADPE4S/Marvell_M88SE6145_Linux.zip> which may be somewhat useful in understanding how to drive this hardware (no datasheet is available). But it is very different from a standard Linux ATA driver. Ben. -- Ben Hutchings Once a job is fouled up, anything done to improve it makes it worse.
signature.asc
Description: This is a digitally signed message part