| Message ID | 20221216115930.2667772-1-npiggin@gmail.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | powerpc: Misc early boot fixes | expand |
On Fri, 16 Dec 2022 21:59:26 +1000, Nicholas Piggin wrote: > The first patch is a fix for a commit that's in powerpc next which > is a pretty harmless dump_stack(). Except that we had some bugs with > doing stack unwinding that early when the boot CPU is not zero so > that ended up crashing badly. > > First patch should be relatively safe and solve that problem, but > only needed if you're testing with stress_hpt. Second patch is a > relatively minimal fix that should close most of the large window > where we're running with the wrong task_cpu, but I would not call > it trivial because it could cause some weird breakage or expose > another bug. Third patch closes the rest of the window and makes > the paca allocation a lot better, but it has more possibility for > regressions. Last patch is independent of the rest and should be > quite straightforward. > > [...] Applied to powerpc/next. [1/4] powerpc/64s: Fix stress_hpt memblock alloc alignment https://git.kernel.org/powerpc/c/dea18da459922d70692ae1e5736d1c760040cb07 [2/4] powerpc/64: Fix task_cpu in early boot when booting non-zero cpuid https://git.kernel.org/powerpc/c/9fa24404f5044967753a6cd3e5e36f57686bec6e [3/4] powerpc/64: Move paca allocation to early_setup() https://git.kernel.org/powerpc/c/dc222fa7737212fe0da513e5b8937c156d02225d [4/4] powerpc: Skip stack validation checking alternate stacks if they are not allocated https://git.kernel.org/powerpc/c/1ee4e35076e36724a62ba2fa29b722fb53eb68f2 cheers
The first patch is a fix for a commit that's in powerpc next which is a pretty harmless dump_stack(). Except that we had some bugs with doing stack unwinding that early when the boot CPU is not zero so that ended up crashing badly. First patch should be relatively safe and solve that problem, but only needed if you're testing with stress_hpt. Second patch is a relatively minimal fix that should close most of the large window where we're running with the wrong task_cpu, but I would not call it trivial because it could cause some weird breakage or expose another bug. Third patch closes the rest of the window and makes the paca allocation a lot better, but it has more possibility for regressions. Last patch is independent of the rest and should be quite straightforward. Thanks, Nick Nicholas Piggin (4): powerpc/64s: Fix stress_hpt memblock alloc alignment powerpc/64: Fix task_cpu in early boot when booting non-zero cpuid powerpc/64: Move paca allocation to early_setup() powerpc: Skip stack validation checking alternate stacks if they are not allocated arch/powerpc/include/asm/paca.h | 1 - arch/powerpc/include/asm/smp.h | 1 + arch/powerpc/kernel/process.c | 11 +++++++++++ arch/powerpc/kernel/prom.c | 12 ++---------- arch/powerpc/kernel/setup-common.c | 4 ++++ arch/powerpc/kernel/setup_64.c | 14 ++++++++------ arch/powerpc/mm/book3s64/hash_utils.c | 3 ++- 7 files changed, 28 insertions(+), 18 deletions(-)