[PATCH 2/3] arm: introduce IRQ stacks

[Russell King - ARM Linux admin]

On Wed, Oct 21, 2020 at 01:45:42PM +0100, Russell King - ARM Linux admin wrote:
> On Wed, Oct 21, 2020 at 02:42:48PM +0200, Arnd Bergmann wrote:
> > (also resending this reply from @kernel.org)
> > 
> > On Fri, Oct 16, 2020 at 12:19 PM Arnd Bergmann <arnd at arndb.de> wrote:
> > > On Thu, Oct 8, 2020 at 9:20 AM Maninder Singh <maninder1.s at samsung.com> wrote:
> > > >
> > > > This patch adds code for switching to IRQ stack.
> > > > IRQ stack and Kernel SVC stack have below design.
> > > >
> > > > IRQ STACK:-
> > > >                     ------------ IRQ stack top
> > > >                     |          |
> > > >                     ------------
> > > >                     .          .
> > > >                     .          .
> > > >                     .          .
> > > >                     ------------
> > > >                     |    sp    | <- irq_stack_base + 0x8
> > > >                     ------------
> > > >                     |    fp    | <- irq_stack_base + 0x4
> > > >                     ------------
> > > >                     |tinfo_ptr | /* pointer to thread info */
> > > > irq_stack_ptr -->   ------------ IRQ stack base
> > > >
> > > > Kernel SVC stack:-
> > > >                     ------------  Kernel stack top
> > > >                     |          |
> > > >                     ------------
> > > >                     .          .
> > > >                     .          .
> > > >                     .          .
> > > >                     ------------
> > > >                     |          |
> > > >                     |          |
> > > >                     ------------
> > > >                     |tinfo_ptr |  /* pointer to thread info */
> > > >                     ------------ Kernel stack base
> > >
> > > The extra indirection doesn't look great, and I don't see any of the
> > > other architectures need that. Since we can access percpu data
> > > without going through thread_info, maybe doing the same as
> > > x86 would work here:
> > >
> > > - define 'current' as 'this_cpu_read_stable(current_task);'
> > > - convert to CONFIG_THREAD_INFO_IN_TASK
> 
> That means we need to also code that up in assembly - remember, we
> need to access thread_info from assembly code.

Note also that there is a circular dependency involved. If you make
thread_info accessible via per-cpu, then:

#ifndef __my_cpu_offset
#define __my_cpu_offset per_cpu_offset(raw_smp_processor_id())
#endif
#ifdef CONFIG_DEBUG_PREEMPT
#define my_cpu_offset per_cpu_offset(smp_processor_id())
#else
#define my_cpu_offset __my_cpu_offset
#endif

smp_processor_id() ultimately ends up as raw_smp_processor_id() which
is:

#define raw_smp_processor_id() (current_thread_info()->cpu)

and if current_thread_info() itself involves reading from per-cpu data,
we end up recursing... infinitely.

This is why I said in the other thread:

"We don't do it because we don't have a separate register to be able
to store the thread_info pointer, and copying that lump between the
SVC and IRQ stack will add massively to IRQ latency, especially for
older machines."

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!