Re: [PATCH v11 1/9] arm64: Add HAVE_REGS_AND_STACK_ACCESS_API feature

[David Long]

On 03/11/2016 01:07 PM, James Morse wrote:
> Hi David,
>
> On 09/03/16 05:32, David Long wrote:
> > From: "David A. Long" <dave.long@xxxxxxxxxx>
> >
> > Add HAVE_REGS_AND_STACK_ACCESS_API feature for arm64.
> >
> > Signed-off-by: David A. Long <dave.long@xxxxxxxxxx>
>
> > diff --git a/arch/arm64/kernel/ptrace.c b/arch/arm64/kernel/ptrace.c
> > index ff7f132..efebf0f 100644
> > --- a/arch/arm64/kernel/ptrace.c
> > +++ b/arch/arm64/kernel/ptrace.c
>
> [ ... SNIP ... ]
>
> > +/**
> > + * regs_within_kernel_stack() - check the address in the stack
> > + * @regs:      pt_regs which contains kernel stack pointer.
> > + * @addr:      address which is checked.
> > + *
> > + * regs_within_kernel_stack() checks @addr is within the kernel stack page(s).
> > + * If @addr is within the kernel stack, it returns true. If not, returns false.
> > + */
> > +bool regs_within_kernel_stack(struct pt_regs *regs, unsigned long addr)
> > +{
> > +	return ((addr & ~(THREAD_SIZE - 1))  ==
> > +		(kernel_stack_pointer(regs) & ~(THREAD_SIZE - 1)));
>
> I'm not sure where this is called from, but if kernel_stack_pointer(regs) could
> ever point into an irq_stack you will get the wrong result.
>
> arch/arm64/include/asm/irq.h has 'on_irq_stack(sp, cpu)' which should help,
> although you will need to check the bounds of the irq_stack separately.
>
>
> The horrible details...
>
>  From arch/arm64/kernel/irq.c:20
> > /* irq stack only needs to be 16 byte aligned - not IRQ_STACK_SIZE aligned. */
> > DEFINE_PER_CPU(unsigned long [IRQ_STACK_SIZE/sizeof(long)], irq_stack)
> > __aligned(16);
>
> This was because per-cpu variables can be at-most page aligned.
> 6cdf9c7ca687 ("arm64: Store struct thread_info in sp_el0") changed
> current_thread_info() to work on these weirdly aligned irq_stacks.
>
>
> Thanks,
>
> James

It looks like this is ultimately used (currently) only by the 
arch-independent kprobes tracing code.  But it does seem like this will 
be recording the wrong data when stack contents are being traced from 
interrupt routine probes.  I will put a fix in for the next spin.

Thanks,
-dl