skiboot
skiboot copied to clipboard
occ's are not initialized properly in fast-reboot path, for the disabled OCC's
On a fast-reboot enabled system if OCC's are disabled, then next reboot is not making up disabled OCC's
Disable OCC's by doing occ reset for more than 3 times [root@localhost ~]# opal-prd occ reset [root@localhost ~]# opal-prd occ reset [root@localhost ~]# opal-prd occ reset [root@localhost ~]# opal-prd occ reset [root@localhost ~]# opal-prd occ reset =feeling-luckyt ~]# nvram -p ibm,skiboot --update-config experimental-fast-reset [root@localhost ~]# reboot [ OK ] Started Show Plymouth Reboot Screen.
[root@localhost ~]# dmesg | grep -i powernv [ 0.000000] Using PowerNV machine description [ 0.750966] EEH: PowerNV platform initialized [ 4.971226] powernv-rng: Registering arch random hook. [ 6.938154] powernv-cpufreq: cpufreq pstate min -43 nominal -17 max 0 [ 6.954201] powernv-cpufreq: Pstate set to safe frequency [ 6.968361] powernv-cpufreq: Frequency Control disabled from OS [ 6.986228] powernv-cpufreq: PMSR = d5d5807fc8000000 [ 7.003352] powernv-cpufreq: CPU Frequency could be throttled [ 7.135968] powernv_idle_driver registered [ 7.136065] nx_compress_powernv: coprocessor found on chip 0, CT 3 CI 1 [ 19.451572] powernv_rng: Registered powernv hwrng. [ 19.476582] ipmi-powernv ibm,opal:ipmi: Found new BMC (man_id: 0x000000, prod_id: 0x0000, dev_id: 0x00) [root@localhost ~]# ipmitool sdr elist | grep -i occ OCC Active | 08h | ok | 210.0 | Device Disabled
After fast-reboot, still OCC's are disabled that means there is no power management, user need to do a full IPL
Hi Pridhiviraj,
OCC is loaded by Hostboot in OpenPOWER sytems. After a full IPL, OCC can be reset/started for a max three times. AFAIUC, this max_occ_reset_count is maintained by HBRT ot HTMGT and this needs to be cleared during the fast-reboot path inorder to get OCC reloaded.
Thanks and Regards, Shilpa
yes, we need to clear this count, and also currently in fast-reboot path, occ_pstates_init is just checking for only previous init path occ_pstates_initialized set to true.
Could you please retest with a recent version, and close if not reproduced?