[PATCH] arm64: initialize spinlock for init_mm's context

Will Deacon will.deacon at arm.com
Wed Apr 16 01:15:06 PDT 2014


On Tue, Apr 15, 2014 at 09:20:50AM +0100, Leo Yan wrote:
> ARM64 has defined the spinlock for init_mm's context, so need initialize
> the spinlock structure; otherwise during the suspend flow it will dump
> the info for spinlock's bad magic warning as below:
> 
> [   39.084394] Disabling non-boot CPUs ...
> [   39.092871] BUG: spinlock bad magic on CPU#1, swapper/1/0
> [   39.092896]  lock: init_mm+0x338/0x3e0, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0
> [   39.092907] CPU: 1 PID: 0 Comm: swapper/1 Tainted: G           O 3.10.33 #125
> [   39.092912] Call trace:
> [   39.092927] [<ffffffc000087e64>] dump_backtrace+0x0/0x16c
> [   39.092934] [<ffffffc000087fe0>] show_stack+0x10/0x1c
> [   39.092947] [<ffffffc000765334>] dump_stack+0x1c/0x28
> [   39.092953] [<ffffffc0007653b8>] spin_dump+0x78/0x88
> [   39.092960] [<ffffffc0007653ec>] spin_bug+0x24/0x34
> [   39.092971] [<ffffffc000300a28>] do_raw_spin_lock+0x98/0x17c
> [   39.092979] [<ffffffc00076cf08>] _raw_spin_lock_irqsave+0x4c/0x60
> [   39.092990] [<ffffffc000094044>] set_mm_context+0x1c/0x6c
> [   39.092996] [<ffffffc0000941c8>] __new_context+0x94/0x10c
> [   39.093007] [<ffffffc0000d63d4>] idle_task_exit+0x104/0x1b0
> [   39.093014] [<ffffffc00008d91c>] cpu_die+0x14/0x74
> [   39.093021] [<ffffffc000084f74>] arch_cpu_idle_dead+0x8/0x14
> [   39.093030] [<ffffffc0000e7f18>] cpu_startup_entry+0x1ec/0x258
> [   39.093036] [<ffffffc00008d810>] secondary_start_kernel+0x114/0x124
> 
> Signed-off-by: Leo Yan <leoy at marvell.com>
> ---
>  arch/arm64/include/asm/mmu.h |    3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/arch/arm64/include/asm/mmu.h b/arch/arm64/include/asm/mmu.h
> index f600d40..06a19bd 100644
> --- a/arch/arm64/include/asm/mmu.h
> +++ b/arch/arm64/include/asm/mmu.h
> @@ -22,6 +22,9 @@ typedef struct {
>  	void *vdso;
>  } mm_context_t;
>  
> +#define INIT_MM_CONTEXT(name) \
> +	.context.id_lock = __RAW_SPIN_LOCK_INITIALIZER(name.context.id_lock),

This should probably be __RAW_SPIN_LOCK_UNLOCKED instead.

Other than that, the patch looks fine to me. I should really get around to
porting the ASID allocator from arch/arm/, then we could kill this lock
altogether.

Will



More information about the linux-arm-kernel mailing list