Update: the 2.6.28.10 kernel caused no lock-up with my NIC. Unless there is some oddball anomaly with my kernel configs, I'll conclude the problem was introduced in the 2.6.29 series. For now the problem remains, but I feel as though I'm getting closer at least to when the problem started.

I did a little looking around and found this http://patchwork.kernel.org/patch/16212/. Not sure if it will work but its worth a look.

and one more http://kerneltrap.org/mailarchive/li...ad#mid-5245584

Reading ChangeLog-2.6.29 seems to indicate when the problems started. Apparently somebody wanted the PHY device to power down when the interface goes down, which is exactly what I see happening on my box. That desire sort of makes sense but unfortunately, the device remains latched in that condition through a warm reset.

Seems the patch is already in the 2.6.29.4 forcedeth.c source file. I'll try compiling another copy of 2.6.29.4.

Hmm. I found CONFIG_FORCEDETH_NAPI=y in my config. Same option is set in the huge and generic kernels too. When I run/compile another pass I'll disable that option.

Update: Setting CONFIG_FORCEDETH_NAPI=n with the 2.6.29.4 kernel produced the same lock-up problem. The 2.6.28.10 kernel is clean.

I'm sad to report the 2.6.29.5 kernel causes the same lockup/latching problem.

Same result: From a cold boot the NIC initializes fine. If I perform a warm reboot then the NIC locks and won't reinitialize without a cold restart. I can avoid the lockup if I use the kernel magic keys (Alt-SysRq-R-E-I-S-U-B).

The problem also latches my WRT54GL router. That is, when the NIC shuts down the router latches into an unresponsive state too. Yes, I tried swapping router ports. I can power down and restart my computer, but the NIC LEDs never fire on. Therefore I can only conclude the router is latched and that prevents the NIC from initializing.

I seem to be left with only a few options: buy a new motherboard, buy a separate NIC (and consume more watts), or never update this box past 12.2.

Well, perhaps there is another option: update all of Slackware but keep using the 2.6.27.7 kernel or update to the 2.6.28.10 kernel. As I reported earlier, the 2.6.28.10 kernel does not cause the lockup.

Sigh.

This lock-up problem has been particularly frustrating. Down right discouraging.

Today I compiled and tested the 2.6.30 kernel. Upon rebooting, my NIC continued to function and also did not latch my router!

I compared the 2.6.29.5 forcedeth.c code to 2.6.30 and there are many differences. There is one section of code with a comment about the NIC "persisting after reboot." Apparently this lock-up problem finally has been fixed in 2.6.30.

I did not test anything else with 2.6.30, but I noticed both KDE and Xfce starting and responding much faster. Perhaps some of the ext3 issues really have been resolved with 2.6.30, which is what I use.

I have notified Pat. I'm hoping he'll consider updating Current to 2.6.30 before official release!