We have been seeing our powerpc based systems boot with only cpu0 being brought online, which seemed odd since these are dual core (P2020) systems. After some investigation and testing it appears the commit below introduced in v3.2.61 causes the problem:
commit 335a4d5ba599428c32e6bdf726cd7f20553220a9 Author: Michael Neuling <[email protected]> Date: Fri Jun 6 14:28:51 2014 +1000 powerpc: Don't setup CPUs with bad status commit 59a53afe70fd530040bdc69581f03d880157f15a upstream. The above commit doesn't take into account ePAPR spin-tables and was later fixed upstream, which appears to have been missed (see patch 4 change log). The result is any distribution using the upstream 3.2 series kernel including and after v3.2.61 prevents other cores from being brought online if their platform's bootloader follows ePAPR (f.e. U-Boot) and only enables a single core during bootloader boot. Another alternative is to revert the patch listed above, I made the assumption boot problems were seen in this LTS kernel which caused the backporting of this patch in the first place, hence the 4 series patch set. Patch 4 actually fixes the problem but depends on patch 1 and 2 for functionality. The only patch that is not absolutly needed is patch 3 ("powerpc: Make logical to real cpu mapping"). Please let me know if a re-spin without patch 3 is wanted. Anton Blanchard (1): powerpc: Make logical to real cpu mapping code endian safe Grant Likely (1): of: Add of_property_match_string() to find index into a string list Scott Wood (1): powerpc: Don't skip ePAPR spin-table CPUs Thierry Reding (1): dt: Add empty of_property_match_string() function arch/powerpc/kernel/setup-common.c | 16 ++++++++++++---- drivers/of/base.c | 36 ++++++++++++++++++++++++++++++++++++ include/linux/of.h | 10 ++++++++++ 3 files changed, 58 insertions(+), 4 deletions(-) -- 1.9.1 -- To unsubscribe from this list: send the line "unsubscribe stable" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
