Description of problem: The NMI doesn't work on some Intel boxes. Version-Release number of selected component (if applicable): 2.6.21-54.el5rt How reproducible: 100% Actual results: No NMI events are seen in /proc/interrupts Expected results: NMI events should be occurring. Additional info: Westford RHTS system hp-dl380g5-01.rhts experiences this problem. Patch against 54.el5rt is attached.
Created attachment 271691 [details] RT fix for this issue
Parit, With this patch in the 2.6.21-55.el5rt kernel, The i386 Intel systems look good I see NMI iterrupts on the systems that did not have them before. As a unwanted side affect we have lost NMI on some x86_64 not all. For example: ********** System Information ********** Hostname = ibm-morrison.lab.boston.redhat.com Kernel Version = 2.6.21-55.el5rt Machine Hardware Name = x86_64 Processor Type = x86_64 Number of Processors = 4 System NMI Interrupts = NMI: 0 0 0 0 Link to dmidecode data for the above system. http://rhts.lab.boston.redhat.com/cgi-bin/rhts/test_log.cgi?id=1282028 ********** System Information ********** Hostname = ibm-ls21-7972-01.lab.boston.redhat.com Kernel Version = 2.6.21-55.el5rt Machine Hardware Name = x86_64 Processor Type = x86_64 Number of Processors = 4 System NMI Interrupts = NMI: 0 0 0 0 Link to dmidecode data for the above system. http://rhts.lab.boston.redhat.com/cgi-bin/rhts/test_log.cgi?id=1282010 ********** System Information ********** Hostname = ibm-wildhorse-01.rhts.boston.redhat.com Kernel Version = 2.6.21-55.el5rt Machine Hardware Name = x86_64 Processor Type = x86_64 Number of Processors = 4 System NMI Interrupts = NMI: 0 0 0 0 Link to dmidecode data for the above system. http://rhts.lab.boston.redhat.com/cgi-bin/rhts/test_log.cgi?id=1282046 Thanks, Jeff
I just gave you a backport but didn't give you the fix for 391741 (which was just confirmed as a valid fix yesterday evening by AMD). After POSTing the fix for 391741 to RHKL, I will open up a new BZ for the RT kernel. ie) the patch above is valid -- it backports the current RHEL5 fixes into the RT kernel. 391741 (and therefore the backport into RT) is a "new" bug. P.
This patch was committed? Moving to POST. P.