Description of problem: please see: http://kernel.org/git/?p=linux/kernel/git/stable/linux-2.6.15.y.git;a=commitdiff;h=bcf2887b1416a506e3461c504642a1b7fad52ddc Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info:
Created attachment 129118 [details] intel_early_workaround() moved to generic_identify() as early_detect_cpus() only run at BP. Geoff, please test this patch, please. I can submit it to rhkernel-list soon after.
I sanity-checked on one box. Oddly generic_identify gets run twice on each proc, but same upstream. Looks fine.
We've made it this far without anyone tearing their hair out, I'm tempted to think this couldn't possibly matter. But I'll check with the mothership before giving up on it.
Suresh pointed out, and I confirmed in the latest EL4 source, that the x86_cache_alignment field this function sets is only ever read with the macro cache_line_size() which reads it from the boot cpu. So the patch isn't really needed. The one exception to this is reporting to /proc/cpuinfo, so on some machine I guess it will show 128 for boot cpu and incorrectly show 64 for others. But I'm not sure that makes it worth posting.
From Comment 4, it doesn't seem like this is a bug. Or at least not worth posting a patch. Please re-open if disagree.