On Mon, Feb 13, 2006 at 10:43:17PM -0800, Steve Langasek wrote:
> Package: linux-2.6
> Version: 2.6.15-6
>
> Currently, the modules.pcimap for DEC tulip chips is as follows:
>
> tulip 0x00001011 0x00000009 0xffffffff 0xffffffff 0x00000000
> 0x00000000 0x0
> tulip 0x00001011 0x00000019 0xffffffff 0xffffffff 0x00000000
> 0x00000000 0x0
> de2104x 0x00001011 0x00000002 0xffffffff 0xffffffff 0x00000000
> 0x00000000 0x0
> de2104x 0x00001011 0x00000014 0xffffffff 0xffffffff 0x00000000
> 0x00000000 0x0
> lmc 0x00001011 0x00000009 0x00001376 0xffffffff 0x00000000
> 0x00000000 0x0
> lmc 0x00001011 0x00000009 0xffffffff 0x00001376 0x00000000
> 0x00000000 0x0
>
> Presently, neither de2104x nor lmc is being included in the installer
> images for etch; this is because discover1-data lists de4x5 as the module
> for *all* of these PCI IDs. With the switch to 2.6 in d-i, this means that
> a number of common NICs for alpha (0002, 0014, 0009) are not being
> auto-detected by udev. The de4x5 module *can* be loaded by hand on my own
> system (1011:0002), and appears to work correctly; this is the driver that
> I've been using under 2.6 on my installed system. I am also now testing
> de2104x; so far, it does appear to work.
>
> The apparent trade-off between de2104x is that de4x5 does not support
> full-duplex mode, whereas I have some vague impression that de2104x didn't
> work on my system in some earlier driver revision. However, the
> discover1-data changelog doesn't support this; it mentions that 1011:1002
> uses de4x5 because of bug #273265, which was about tulip, not de2104x -- and
> tulip doesn't detect my card at all, so there's no doubt that *that* is the
> wrong driver.
>
> Recent discussion of this issue on debian-alpha included the following
> response froman Alpha expert at HP:
>
> -------------------------------------------------------
> From: Jay Estabrook <jay.estabr...@hp.com>
> To: debian-al...@lists.debian.org
> Subject: Re: testing wanted: debian-installer, now with 2.6.15
> Date: Mon, 13 Feb 2006 10:53:43 -0500
>
> On Sun, Feb 12, 2006 at 03:41:24AM -0800, Steve Langasek wrote:
>
> > So should the 21040 and 21041 be switched from de2104x to de4x5, or should
> > de2104x be added to the installer? I seem to recall that I had problems
> > with de2104x on my 21040 as well; and at least in 2.6.15, the tulip driver
> > doesn't work at all for my card, I have to use de4x5.
>
> > The discover1-data package seems to have de4x5 listed for *all* of these PCI
> > IDs, but I think we may be using the kernel's map now with 2.6 (via udev).
>
> I don't know of any problems using "tulip" on the 2114x chipsets. And
> I don't think full duplex will significantly increase the throughput
> of a 2104x chipset, so, yes, I'd think that using de4x5 for any 2104x
> and tulip fo any 2114x might be the way to go...
> -------------------------------------------------------
>
> We probably shouldn't make that change in Debian without talking to
> whoever's responsible for these drivers upstream, though. In the meantime,
> I'm going to add de2104x into the d-i images. This bug report is opened to
> collect success/failure reports regarding the current driver mappings, so we
> can be sure we're making an informed decision about these
> historically-problematic PCI IDs.
>
> For the record, de4x5 is currently not referenced in modules.pcimap at all,
> so depending on the outcome of this bug report, it may make sense to drop it
> completely from the 2.6 build...
What's the status in Lenny? Is this bug still needed?
Cheers,
Moritz
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
