Please read this before reporting a bug:
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
FS#6212 - r1000 patch should be moved out of the kernel
Attached to Project:
Arch Linux
Opened by Papadopoulos, Mihail (scarecrow) - Sunday, 14 January 2007, 11:16 GMT
Last edited by Roman Kyrylych (Romashka) - Sunday, 14 January 2007, 12:11 GMT
Opened by Papadopoulos, Mihail (scarecrow) - Sunday, 14 January 2007, 11:16 GMT
Last edited by Roman Kyrylych (Romashka) - Sunday, 14 January 2007, 12:11 GMT
|
DetailsVersion 1.0.5 of the Realtek driver seems to be faulty with some network cards (including my onboard 8168 gigabit card, on an Asus P5B-VM mainboard.
The card worked fine with the previous driver revision (1.0.4), and is also working fine with the kernel module- just not with the current version. With it and HAL loaded, the network speed is horribly slow. Things gfo back to normal if hal is unloaded and the network is restarted a couple of times- although there are some stability issues. As it is now I have to rebuild myself every new kernel revision, omitting the r1000 patch. Like that, my RTL8168 is working just fine. Since that patch does good to some people and harm to some others, I think it would be better if it was moved outside the kernel, and offered as a separate module package instead. |
This task depends upon
But this is far from being a clean solution, especially since the unpatched kernel module just works (for me)... or not?
I can reproduce the problems you have with r1000, it gives 20% packetloss on my Asrock mainboard, while r8169 works fine. Blacklisting r1000 is enough in my situation.
However, I'm stalemated- here's why:
- If no module is blacklisted, an "lsmod" just shows r1869 and not r1000 (that's why I thought the patch just upgrades r8169). The network is running, but with severe packet loss (should be 80% or more!).
- With r8169 blacklisted, lsmod shows r1000, and the network is as slow as above.
- With r1000 blacklisted, lsmod shows just r1869, but the network does not start at all (network unreachable error message).
Maybe the problem is hardware-specific, but anyway, building the kernel with the r1000 patch omitted works fine- as I said before.
Is it strictly hardware specific, or a strange combo of hardware and the way mkinitcpio creates kernel modules? I'm really puzzled...