00:0a.0 Ethernet controller: Marvell Technology Group Ltd. 88E8001 Gigabit Ethernet Controller (rev 13) motherboard A8V Deluxe kernel-2.15.1-r1 module skge gives bad performances, while older sk98lin seems working better.
Created attachment 78711 [details] lspci
Created attachment 78712 [details] dmesg whith sk98lin driver
Created attachment 78713 [details] emerge info Pay attention. I'm using a 32-bit system.
Created attachment 78714 [details] sk98lin ebuild (local portage) I'm working on a machine used as ltsp server on a 100Mb lan. Module skge is slow too much and gives a lot of "nfs not working". sk98lin is compiled with the ebuild attached whith +napi useflag, from this other bug: http://bugs.gentoo.org/show_bug.cgi?id=93283 I'm sorry, because I'm not expert of this kind of problems, but if you need some other questions, may you ask.
Please attach dmesg from a skge kernel where you have been using the network adapter for a few minutes. Please describe the difference in performance that you are seeing plus the exact NFS issues.
(In reply to comment #5) > Please attach dmesg from a skge kernel where you have been using the network > adapter for a few minutes. Please describe the difference in performance that > you are seeing plus the exact NFS issues. > Thanks for your answer. The only error message I find is on the client: 192.16.200.99: server nont responding still trying It happens if more then three or four clients are booting, and makes impossible to work. In a correct behaviour (with sk98lin driver) I boot more then 10 clients in the same time, with minimal delay. Now I've recompiled the kernel without prempt otpion, but it did'nt help. So I attach this documentation: 1. New kernel config with no preemption (from /proc/config) 2. new emerge info 3. tail from /var/log/everything/current with skge driver. 4. tail from /var/log/kernel/current with skge driver. 5. dmesg with skge driver.
Created attachment 80080 [details] New kernel config with no preemption (from /proc/config)
Created attachment 80081 [details] tail from /var/log/everything/current with skge driver.
Created attachment 80083 [details] tail from /var/log/kernel/current with skge driver.
Created attachment 80084 [details] dmesg with skge driver.
Created attachment 80086 [details] new emerge info
It isn't the reason for slowdown, but I noticed that both skge and sk98lin are running at half-duplex on your setup. Is there any reason for this? Please could you test the latest development kernel (currently 2.6.16-rc5) - some skge fixes are included, hopefully that will help...
(In reply to comment #12) > It isn't the reason for slowdown, but I noticed that both skge and sk98lin are > running at half-duplex on your setup. Is there any reason for this? > Thank you for your answer. The half-duplex connection depends from the DFE-2624X hub. See here for a description: http://support.dlink.com/products/view.asp?productid=DFE%2D2624X . I don't know if it's a correct behaviour, but connecting directly two ethernet devices, I obtain a full duplex connection. About slowdown, I need more time to investigate.
no improvement with kernel 2.6.16-rc5
Sorry. I'm really stupid. Just now I read 16, instead of 15. So I've just lose a lot of time. I will try again later. Sorry again.
Created attachment 82360 [details] linux-2.6.16-rc5/.config I've try the new vanilla kernel with no lucky on a clone partition (with reiserfs instead reiser4). Nor skge nor sk98lin module runs well. If you want to suggest some other test, may I done.
Please retest with gentoo-sources-2.6.16-r1
Created attachment 83952 [details] linux-2.6.16-r1/.config new kernel, no improvements.
Marco, this is a known bug. Stephen, this is the bug I mentioned in person, which you already know about. 32bit system with 4GB RAM suffering from bad performance with skge. If you need any patches testing or any more info, feel free to throw things this way :)
(In reply to comment #19) > Marco, this is a known bug. > Well. But I didn't know that :) . May you please give me a link? For a now, I've not tested my system with no High Memory Support. Anyway, may I test anything you think be useful. Thank's for your support.
We've come to the conclusion that nForce4 does bad things with memory layout. On a 64bit setup you could probably work around this bug by booting with iommu=force but I don't think that applies to 32bit. Are there any BIOS updates available for your system?
(In reply to comment #21) > On a 64bit setup you could probably work around this bug by booting with > iommu=force but I don't think that applies to 32bit. > In this moment, I can't install a 64bit setup on my system. May I try later (may be in septembar?). > Are there any BIOS updates available for your system? > I'm using a bios version 1015 BETA 002 from here: http://support.asus.com/download/download_item_nn.aspx?model=A8V%20Deluxe&product=1&f_name=&type=Latest&SLanguage=en-us# Today, I find only a 1015 Bios version, without BETA. Anyway, I think it's the same BIOS. Some month ago, I tried to update to newer bios versions, But I had some boot trubbles and I was forced to downgrade. Other informations from my bios: AMIBIOS VERSION: 08.00.09 Build Date: 10/06/05 ID: A0036001
(In reply to comment #22) Marco, is the error still present with >=2.6.18 ?
(In reply to comment #23) > > Marco, is the error still present with >=2.6.18 ? > I'm sorry. Now I'm using an old eth card, with 8139too module, to avoid problems. I can't try with new kernel because I'm still using reiser4 filesystem, which does not support kernel 2.6.18. Anyway, I'm planning to change the filesystem and to make a check for you. Please wait for some time. Tanks for you support.
wonderfull! tested also with a 64-bit kernel on my 32-bit environment. Thanks a lot.
