found 630593 linux-2.6/3.1.1-1 quit Leopold Palomo-Avellaneda wrote:
> I don't have any hope that it would be solved in the linux kernel, but at > least in the debian kernel it could be applied. So, it seems that "the Intel bug" is that Intel's 82091 is not advertised to implement (and does not implement?) the timeout bit in the ECR register. There are various possible approaches to work around that. The current code in mainline that tries to detect these chipsets is nonsense. Even so, the suggested patch that rips it out would have a chance of introducing regressions and probably should not be applied unless someone finds a way to test it widely. There were two alternative approaches suggested in the upstream discussion, intended not to suffer from the same problem[1][2] (caveat: I have not sanity-checked the second one). In other words, we are between steps 1 and 2 of the short story described earlier in this bug log. The next step is to take the feedback received into account and propose a new patch. My only additional advice is that [3] seems to be a nice reference for getting accustomed to Linux internals. If I were writing it, I would rip out the code marked "Check for Intel bug" and use a kernel parameter to decide whether to assume broken EPP support, defaulting to true. Simpler, more robust, and hard to object to. Thanks again and hope that helps, Jonathan [1] http://thread.gmane.org/gmane.linux.kernel/1191731 [2] http://thread.gmane.org/gmane.linux.kernel/1192024/focus=1746 [3] http://lwn.net/Kernel/LDD3/ -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected] Archive: http://lists.debian.org/[email protected]
