=Comment: #0================================================= KEITH MANNTHEY <mannthey.com> - 2008-03-17 16:09 EDT Problem description: We would like to push all the SMI free bits to RedHat for MRG. This includes EDAC kernel bits (AMD driver for the most part), BIOS driver kernel bits and Userspace bits. This code will be a port of the code from our SR3 release. This is a place marker bug at this point. There are RFC patches on the IBM RedHat list at this time.
------- Comment From mannthey.com 2008-03-24 13:41 EDT------- Just wanted to update: There is a changed in the EDAC driver that is causing a change in topology on the LS21. We are working to find a good solution for the situation. Once things are properly sorted we will post the patches here. We expect this to happen 1/2 of this week.
Created attachment 299808 [details] Userspace SMI bits untar and run "make install" is our current install method. Contains all the userspace bits. The /etc/ service and the deamon that reports the ecc errors to the BMC.
Created attachment 299809 [details] Kernel: IBM RTL driver This driver interacts with a bios table in the EDBA labeled _RTL_ . It creates a sysfs interface for users to use to enter/exit the BIOS SMI free state (/sys/devices/system. From below it is a pretty decent description of what this patch does. + Enable support for IBM Preimium Real Time Mode (PRTM). + This module will allow you the enter and exit PRTM in the BIOS via + sysfs on platforms that support this feature. System in PRTM will + not recive cpu generated SMIs for recoverable errors. Use of this + feature without proper support may void your hardware warranty. + + If the proper bios support is found the driver will load and create + /sys/devices/system/ibm_rtl/. The "state" variable will indicate + weather or not the BIOS is in PRTM. + state = 0 (BIOS SMI's on) + state = 1 (BIOS SMI's off) On system that do not support the feature the module fails to load. This driver is written by me, and thoroughly tested on our 2 platforms that support this feature. I would not call it hardened just well tested. Any further platforms from IBM will use this same interface. This has been updated to use the readb and writeb interface from the patch on the list. Signed-off-by: Keith Mannthey<kmannth.com> ---
------- Comment From mannthey.com 2008-03-31 22:44 EDT------- (From update of attachment 35807) This is a port of the current SVN Patch "drivers-edac-new-k8-rev-f.patch" It adds the AMD rev F (and all earlier) support to the EDAC frame work. My LS21s I use for testing are not booting 2.6.24 kernels that I build from the source rpm. The binary MRG kernel is working just fine but any other 2.6.24.3 kernel is having issues (I will be posting another email about this issue shortly) so please take the patch as RFC only, I am working to get this sorted out ASAP. We used the AMD64 driver with our 2.6.21 build so this EDAC driver has not been thoroughly tested yet. Signed-off-by: Keith Mannthey <kmannth.com> ------- Comment From mannthey.com 2008-03-31 22:46 EDT------- Ok I think everthing is here Userspace bits, RTL kernel driver and the 3 EDAC patches. It has been tested on LS21 and HS21 with the MRG kernel as a base. Please send me any feedback you have.
Created attachment 299810 [details] EDAC refactor This is a port of a current SVN EDAC patched called "drivers-edac-refactor-opstate_init.patch" to current MRG. It is a consolidation of code related to EDAC drivers polling or waiting for NMIs. The AMD driver (next patch) uses this interface. Signed-off-by: Keith Mannthey <kmannth.com> ---
Created attachment 299811 [details] AMD EDAC driver (needs the refactor patch)
Created attachment 299812 [details] EDAC 5000 TMID bit fix This is a bit of a hack to turn off a bit in the EDAC check on the i5000 controller. In some situations the HS21XM may raise this bit and EDAC will spam the console with error messages to the point where the local console is unusable. The error involves memory trying to throttle itself due to temperature reasons. Our BIOS folks said the feature is not supported, that they don't trigger on that bit being set and to ignore the error. I don't have any option but to ignore the bit at this time. This is has been well tested to fix the issue. Signed-off-by: Keith Mannthey<kmannth.com>
------- Comment From dvhltc.com 2008-06-10 14:20 EDT------- Still pending the final package from Clark, Should become available this week.
ibm-prtm-1.0-1.el5rt pushed to beta repository
Are we good on this one?
Delivered with MRG 1.0 GA Closing