[PATCH v17 01/13] arm64: Add back cpu reset routines

[geoff@infradead.org (Geoff Levand)]

On Thu, 2016-06-09 at 15:50 +0100, James Morse wrote:
> Hi Geoff,
> 
> On 03/06/16 19:13, Geoff Levand wrote:
> > Commit 68234df4ea7939f98431aa81113fbdce10c4a84b (arm64: kill flush_cache_all())
> > removed the global arm64 routines cpu_reset() and cpu_soft_restart() needed by
> > the arm64 kexec and kdump support.  Add simplified versions of those two
> > routines back with some changes needed for kexec in the new files cpu_reset.S,
> > and cpu_reset.h.
> > 
> > When a CPU is reset it needs to be put into the exception level it had when it
> > entered the kernel. Update cpu_soft_restart() to accept an argument which
> > signals if the reset address needs to be entered at EL1 or EL2, and add a
> > new hypercall HVC_SOFT_RESTART which is used for the EL2 switch.
> > 
> > Signed-off-by: Geoff Levand <geoff@infradead.org>
> 
> 
> > diff --git a/arch/arm64/kernel/cpu-reset.S b/arch/arm64/kernel/cpu-reset.S
> > new file mode 100644
> > index 0000000..c321957
> > --- /dev/null
> > +++ b/arch/arm64/kernel/cpu-reset.S
> 
> > +ENTRY(__cpu_soft_restart)
> > +> > 	> > /* Clear sctlr_el1 flags. */
> > +> > 	> > mrs> > 	> > x12, sctlr_el1
> > +> > 	> > ldr> > 	> > x13, =SCTLR_ELx_FLAGS
> > +> > 	> > bic> > 	> > x12, x12, x13
> > +> > 	> > msr> > 	> > sctlr_el1, x12
> > +> > 	> > isb
> > +
> > +> > 	> > cbz> > 	> > x0, 1f> > 	> > 	> > 	> > 	> > // el2_switch?
> > +> > 	> > mov> > 	> > x0, #HVC_SOFT_RESTART
> > +> > 	> > hvc> > 	> > #0> > 	> > 	> > 	> > 	> > // no return
> > +
> > +1:> > 	> > mov> > 	> > x18, x1> > 	> > 	> > 	> > 	> > // entry
> > +> > 	> > mov> > 	> > x0, x2> > 	> > 	> > 	> > 	> > // arg0
> > +> > 	> > mov> > 	> > x1, x3> > 	> > 	> > 	> > 	> > // arg1
> > +> > 	> > mov> > 	> > x2, x4> > 	> > 	> > 	> > 	> > // arg2
> > +> > 	> > ret> > 	> > x18
> 
> Why ret not br?

Sure.

> 
> 
> > +ENDPROC(__cpu_soft_restart)
> > +
> > +.popsection
> > diff --git a/arch/arm64/kernel/cpu-reset.h b/arch/arm64/kernel/cpu-reset.h
> > new file mode 100644
> > index 0000000..5a5ea0a
> > --- /dev/null
> > +++ b/arch/arm64/kernel/cpu-reset.h
> > @@ -0,0 +1,29 @@
> > +/*
> > + * CPU reset routines
> > + *
> > + * Copyright (C) 2015 Huawei Futurewei Technologies.
> > + *
> > + * This program is free software; you can redistribute it and/or modify
> > + * it under the terms of the GNU General Public License version 2 as
> > + * published by the Free Software Foundation.
> > + */
> > +
> > +#ifndef _ARM64_CPU_RESET_H
> > +#define _ARM64_CPU_RESET_H
> > +
> > +#include 
> > +
> > +void __cpu_soft_restart(unsigned long el2_switch, unsigned long entry,
> > +> > 	> > unsigned long arg0, unsigned long arg1, unsigned long arg2);
> > +
> > +static inline void __noreturn cpu_soft_restart(unsigned long el2_switch,
> > +> > 	> > unsigned long entry, unsigned long arg0, unsigned long arg1,
> > +> > 	> > unsigned long arg2)
> 
> What is the last arg for? machine_kexec() passes zero, but
> arm64_relocate_new_kernel() never reads this value..

cpu_soft_restart is a generic routine, and I thought 3 args would be
good.  It also allows for passing something extra to
arm64_relocate_new_kernel when debugging. 

> > +{
> > +> > 	> > typeof(__cpu_soft_restart) *restart;
> > +> > 	> > restart = (void *)virt_to_phys(__cpu_soft_restart);
> > +> > 	> > restart(el2_switch, entry, arg0, arg1, arg2);
> 
> This confuses me each time I see it, I think it would be clearer if the
> 'cpu_install_idmap()' call were moved into this function. Any other user of this
> function would need to do the same.

Sure.

> 
> By the end of the series, the caller of this has:
> > is_kernel_in_hyp_mode() ? 0 : (in_crash_kexec ? 0 : is_hyp_mode_available())
> which is difficult to read, I had to write out the values to work it out.
> 
> I thinks it makes more sense to move the hyp-aware logic into this
> cpu_soft_restart(), obviously kdump still needs a 'skip el2 jump' flag.
> 

I'll try it.

> +	unreachable();
> > +}
> > +
> > +#endif
> > diff --git a/arch/arm64/kernel/hyp-stub.S b/arch/arm64/kernel/hyp-stub.S
> > index 8727f44..a129e57 100644
> > --- a/arch/arm64/kernel/hyp-stub.S
> > +++ b/arch/arm64/kernel/hyp-stub.S
> > @@ -71,8 +71,17 @@ el1_sync:
> >  > > 	> > msr> > 	> > vbar_el2, x1
> >  > > 	> > b> > 	> > 9f
> >  
> > +2:> > 	> > cmp> > 	> > x0, #HVC_SOFT_RESTART
> > +> > 	> > b.ne> > 	> > 3f
> > +> > 	> > mov> > 	> > x0, x2
> > +> > 	> > mov> > 	> > x2, x4
> > +> > 	> > mov> > 	> > x4, x1
> > +> > 	> > mov> > 	> > x1, x3
> > +> > 	> > blr> > 	> > x4
> 
> blr not branch? If we ever did return from here, wouldn't we run the 'entry'
> function again at EL1?

Yes, this should not return.

> 
> > +> > 	> > b> > 	> > 9f
> > +
> >  > > 	> > /* Someone called kvm_call_hyp() against the hyp-stub... */
> > -2:> > 	> > mov     x0, #ARM_EXCEPTION_HYP_GONE
> > +3:> > 	> > mov> > 	> > x0, #ARM_EXCEPTION_HYP_GONE
> >  
> >  9:> > 	> > eret
> >  ENDPROC(el1_sync)
> > 
> 
> For what its worth:
> 
> Reviewed-by: James Morse <james.morse@arm.com>

Thanks for the comments.

-Geoff