Le mardi 16 juillet 2013 14:06:36 Florian Fainelli a écrit : > Hello Jonas, > > 2013/6/29 Jonas Gorski <jogo@xxxxxxxxxxx>: > > This patchset aims at unifying the different BMIPS support code to allow > > building a kernel that runs on multiple BCM63XX SoCs which might have > > different BMIPS flavours on them, regardless of SMP support enabled in > > the kernel. > > > > The first few patches clean up BMIPS itself and prepare it for multi-cpu > > support, while the latter add support to BCM63XX for running a SMP kernel > > with support for all SoCs, even those that do not have a SMP capable > > CPU. > > > > This patchset is runtime tested on BCM6348, BCM6328 and BCM6368, to > > verify that it actually does what it claims it does. > > > > Lacking hardware, it is only build tested for BMIPS4380 and BMIPS5000. > > > > Jonas Gorski (10): > > MIPS: bmips: fix compilation for BMIPS5000 > > MIPS: allow asm/cpu.h to be included from assembly > > MIPS: bmips: add macros for testing the current bmips CPU > > MIPS: bmips: change compile time checks to runtime checks > > MIPS: bmips: merge CPU options into one option > > MIPS: BCM63XX: let the individual SoCs select the appropriate CPUs > > MIPS: bmips: add a helper function for registering smp ops > > MIPS: BCM63XX: always register bmips smp ops > > MIPS: BCM63XX: change the guard to a BMIPS4350 check > > MIPS: BCM63XX: disable SMP also on BCM3368 > > After fixing the typo on BMIPS4350 vs BMIPS4380 and fixing the > following (which I will submit just in a few minutes) I just gave this patchset a try on a BMIPS5000 system, but unfortunately the kernel crashes early on due to a change in smp-bmips.c. It crashed in mem_init(), and adding a printk in bmips_ebase_setup() in the final else clause before the return did no longer make the kernel crash. This could be some nasty cache issue and simply changing the kernel size did help. Will investigate this more. -- Florian
