Bug#394344: another Marvell Technology 11ab:6101 PATA Controller (support all-generic-ide)
Geert Stappers wrote:
>Op 31-01-2007 om 16:27 schreef Jacob L. Anawalt:
>
>
>
>>I have been using it and it works well enough for my needs post-install.
>>It looks like drivers/ata/pata_marvell.c is in 2.6.20-rc7.
>>
>><snip/>
>>
>>
I mean that I have been using pata_marvell.c in linux-2.6.19-rc1-mm1 in
the system I first tested with in October 2006. The pata_marvell.c
driver has been working for me, so I didn't need to wait for Marvell to
do the honours. Actually, I didn't even need to wait for that driver if
I used a kernel with ide built in so that the kernel option
all-generic-ide actually did something for this chipset, like the one in
Knoppix. Running the device in generic mode is fine for installing from
a CD-ROM.
>What follows is from the type of
> "I don't understand the remark, but I think this helps"
>
>
>
>>I looked at the source for linux-2.6-2.6.18.dfsg.1. I will try it when
>>it is released, but there seems to be no driver to claim the Marvell 6101.
>>
>>
>
>
>The file /lib/modules/$(uname -r)/modules.pcimap should contain an entry
>that for PCI-ID 11ab:6101 the module pata_marvell should be claimed.
>
>
Do you mean "If the file /lib/modules/$(uname -r)/modules.pcimap
contains an entry for PCI-ID 11ab:6101 because pata_marvell.c is
compiled for that kernel and had a MODULE_DEVICE_TABLE(pci entry for
modutils to find, the module pata_marvell should be claimed"?
If so, then yes, I understand that.
If that is not what you mean, then please explain how pata_marvell will
be in linux-2.6-2.6.18.dfsg.1's modules.pcimap when pata_marvell.c is
not included in the source.
You said that when my hardware has an unknown PCI IDE controller the
wrong or no module will be loaded. I have shown that this is what is
happening to me.
You also said that you seriously doubt that an IDE controller driver
_only can work_ in a _monolithic kernel_. Then you have said that you
leave the honour to Marvell Technology Group.
Do you mean "if your hardware won't work as a module, it won't work in
the debian-installer because we will not support all-generic-ide as long
as it requires that ide is built-in" or are you saying "you are up in
the night about all-generic-ide being able to work without an entry in
modules.pcimap?"
I don't understand why ide/pci/generic.c only enables all-generic-ide
when not built as a module, but it's been that way for the last few 2.6
kernels at least.
What I do know is that I could have installed from the amd64
debian-installer using an IDE CD-ROM drive on the Marvell PATA interface
in October of 2006 when I first got the Intel DG965RY main board if I
could have gotten the installer to support the device in generic mode.
I consider lack of generic ide support a bug in the way Debian x86/amd64
kernels are built, or, if everything must be modular, a bug due to the
lack of a patch enabling generic ide support when CONFIG_BLK_DEV_IDE is
built as a module.
This is not just "another Marvell Technology 11ab:6101 PATA Controller"
issue. The issue is the lack of having the option to fall back to
generic support for an ide device without recompiling the kernel when a
more specialized driver is not available.
--
Jacob
Reply to: