Bugzilla Bug 129216

Realtek r1000 driver adds support for RTL8111(B) and RTL8168(B) cards. It can
be downloaded from:

ftp://202.65.194.18/cn/nic/rtl8111brtl8168b/r1000-8111b(102).zip

It compiles and runs fine on my machine, I'm using gentoo-sources-2.6.15-r1 on
a x86 arch.

Thanks a lot for adding an ebuild for this driver :)

These looks like gigabit network cards to me, no reason to CC: mobile for
those.

I second the request for a maintainer... got it running on my machine as well
just now (gentoo-sources-2.6.16-r12)

If I'll find the time, I may deal with it myself... but my free time is spare
:(

can anyone confirm that these do not work with kernel 2.6.17?

refer to this thread for the build error on kernel 2.6.17:
http://forums.gentoo.org/viewtopic-p-3438854.html#3438854

Created an attachment (id=91661) [details]
ebuild for r1000 kernel module

This is a first ebuild for the r1000 driver from realtek
My ebuild is far from perfect though.
A few issues that you may help to resolve: 
-I couldn't figure out how to correctly escape the parentheses in the url, so I
copied the file to my server, for the time being (bad idea though, I know)
-I had some trouble with where the source was unpacked and where the compiled
modules finally where, if you know a better way, please feel free to improve my
ebuild

Created an attachment (id=91686) [details]
r1000-103.ebuild (fixed kernel version dependency)

I had a wrong kernel version dependency, so now it's fixed

Created an attachment (id=92012) [details]
realtek r1000 driver

added a dependency (app-arch/unzip)

concerning the kernel 2.6.17 incompatibility, this:
http://www.mail-archive.com/netdev@vger.kernel.org/msg13809.html
should answer that question.

Not that anyone has stepped up to maintain it, but I object to this being added
to the tree. The in-kernel r8169 driver will support this hardware very soon,
we just need some people to test it.

http://www.fr.zoreil.com/linux/kernel/2.6.x/2.6.18-rc2/r8169/0005-r8169-sync-with-vendor-s-driver.txt

Created an attachment (id=95235) [details]
ebuild for r1000-1.04

-new version 1.04.
-module_param_array is used instead of MODULE_PARM(obsolete for 2.6 kernels)
-added ~amd64 to the KEYWORDS

hi,
i just rewrote the ebuild so it works with the new version 1.04.
i also added a fix that module_param_array is used in the sources instead of
MODULE_PARM. this is deprecated for 2.6 kernels... my 2.6.17 kernel didn't
compile without this fix. now it works with 2.6.17. also added ~amd64 to the
KEYWORDS.

Oh great! Btw, the r8169 driver that's included in the kernel also does the job
if patched with Francois Romieu's changeset. (Just as Daniel pointed out in
comment #8)
(see http://forums.gentoo.org/viewtopic-p-3454719.html#3454719)

There is another ebuild here:
http://www.zugaina.org/gentoo/portage/net-misc/r1000/r1000-1.04.ebuild

8169 is supported in kernel 2.6.18, but this driver supports other chips too,
such as 8168B. 

gentoo-sources-2.6.18-r1 will include support for 8168 and other new chips

*** Bug 150637 has been marked as a duplicate of this bug. ***

2.6.18-r1 is now in portage

(In reply to comment #13)
> gentoo-sources-2.6.18-r1 will include support for 8168 and other new chips

it only supports 8169.

No, it supports 8169 plus the new hardware (8168, 8167, 8136)

$ grep 8168 drivers/net/r8169.c
        { PCI_DEVICE(PCI_VENDOR_ID_REALTEK,     0x8168), 0, 0, RTL_CFG_2 },

I confirm that it works. Maybe the description can be changed to note the
extended support?

I'm going to close this as an equivalent driver is now shipped in
gentoo-sources-2.6.18 (now stable) and the mainline kernel as of 2.6.19. If
there are bugs in that driver I'll happily tend to them.

The inbuilt kernel driver does not work with the 8101E chip in my Toshiba
Satellite M110 laptop, kernel 2.6.19. Realtek has released version 1.05 of the
r1000 driver, it would be nice if this ebuild remained under development for
those users of the more obscure hardware supported by this driver that is not
yet in the kernel.

(In reply to comment #20)
> The inbuilt kernel driver does not work with the 8101E chip in my Toshiba
> Satellite M110 laptop, kernel 2.6.19. Realtek has released version 1.05 of the
> r1000 driver, it would be nice if this ebuild remained under development for
> those users of the more obscure hardware supported by this driver that is not
> yet in the kernel.
> 

Do you want that I reopen this bug?

(In reply to comment #21)
I am using the 1.05 version of the r1000 driver with a 2.6.19 kernel, and it
creates a soft-lock in the kernel when modprobing the module, however it only
causes a ~1 second pause and everything functions correctly once loaded. I am
unsure whether it would be better to continue with the r1000 ebuild or to file
a new bug instead asking for 8101E support in the inbuilt kernel. What is your
recommendation? 

Created an attachment (id=116656) [details]
r1000-fastcall.patch

patch to make 1.05 build on a 2.6.18-hardened-r6 kernel. I removed the sed
stuff (MODULE_PARM fix) in ebuild.

I think this bug could be reopened. Its nice to have r1000 as an alternative
since there people have experienced problems with r8169. Some chipsets are not
detected others have problems with bonding. If r1000 driver were available it
would be possible to use the driver that works best for you.

Thanks.

(In reply to comment #23)
> Created an attachment (id=116656) [edit] [details]
> r1000-fastcall.patch
> 
> patch to make 1.05 build on a 2.6.18-hardened-r6 kernel. I removed the sed
> stuff (MODULE_PARM fix) in ebuild.
> 
> I think this bug could be reopened. Its nice to have r1000 as an alternative
> since there people have experienced problems with r8169. Some chipsets are not
> detected others have problems with bonding. If r1000 driver were available it
> would be possible to use the driver that works best for you.
> 
> Thanks.
> 

Could you please provide a link relating some of these problems?

r1000 drivers works in kernel 2.6.20?

http://bugzilla.kernel.org/show_bug.cgi?id=6032(In reply to comment #24)
> (In reply to comment #23)
> > Created an attachment (id=116656) [edit] [details]
> > r1000-fastcall.patch
> > 
> > patch to make 1.05 build on a 2.6.18-hardened-r6 kernel. I removed the sed
> > stuff (MODULE_PARM fix) in ebuild.
> > 
> > I think this bug could be reopened. Its nice to have r1000 as an alternative
> > since there people have experienced problems with r8169. Some chipsets are not
> > detected others have problems with bonding. If r1000 driver were available it
> > would be possible to use the driver that works best for you.
> > 
> > Thanks.
> > 
> 
> Could you please provide a link relating some of these problems?

Here is a quote from the email traffic with the user that reported those
problems:

>> problem on ubuntu-6.10 too (IIRC its a 2.6.17.x kernel):
>> http://www.howtoforge.com/network_bonding_ubuntu_6.10#comment-3401
>> 
>> Suse have problems with ifdown on r8169 (afaik 2.6.18.2 kernel):
>> http://lists.opensuse.org/opensuse-bugs/2007-03/msg00503.html
>> 
>> The changelog for 2.6.20-rc3 does mention r8169 changes.
>> http://kerneltrap.org/node/7529
>> So there is hope that it might have get fixed.
>
> Looks like a driver bug:
>
> http://bugzilla.kernel.org/show_bug.cgi?id=6032
>
> If mac address can't be changed on the fly, bonding won't work ;-(
>
> There are patches available for various kernels, haven't yet made it
> into mainline due to regressions apparently.


> r1000 drivers works in kernel 2.6.20?

I don't know, sorry.

(I just got no rtl gigabit nics. I will try get some time next week to do some
testing)

This is solved for me with current kernel (well, since ayes) and kernel drivers
Also ftp://202.65.194.18/cn/nic/rtl8111brtl8168b/r1000-8111b(102).zip seems
died