Looking for closure on: michael.goddard wrote: > > Hi Doug, > > This is Mike Goddard from AMD, we met on a phone call a month or two ago. > > We (AMD) has been doing some internal testing of RedHat 6.2 in our lab on > some of our yet to be released processors and found a CPUID "misuse" > problem in (I believe) the kernel. This causes a GPF on our new part. > RedHat 6.1 did not have this problem. > > The sequence of events: > > CPUID function 80000001 accesses the extended feature register, > returning value in EDX > if (bit 18 set) { > Push ptr to string "Disabling Pentium Serial Number" > do some rdmsr/wrmsr stuff > } > > Unlike the standard feature flags, the vendor feature flags are vendor > specific, i.e. either > "GenuineIntel" or "AuthenticAMD". Or new processor used bit 18 of the > 80000001 > extended feature flag to indicate ECC support not a serial number. > > If we use some internal tools to force bit 18 to zero, RedHat 6.2 boots > fine. > > Can you help? I'll take care of this in the next PIII patch that I put out (these patches make changes to that specific code and will therefore fix the problem). -- Doug Ledford <dledford> http://people.redhat.com/dledford Please check my web site for aic7xxx updates/answers before e-mailing me about problems
This was also resolved in the standard kernels.