Re: [Xen-devel] [[PATCH v3] 4/4] xen/arm: Replace call_smc with arm_smccc_smc

[Julien Grall <julien.grall@xxxxxxx>]

Hi Andrew,

On 10/01/2018 02:11 PM, Andrew Cooper wrote:
> On 01/10/18 13:46, Julien Grall wrote:
> > call_smc is a subset of arm_smccc_smc. Rather than having 2 methods to
> > do SMCCC call, replace all call to the former by the later.
> >
> > Signed-off-by: Julien Grall <julien.grall@xxxxxxx>
> >
> > ---
> >
> >      Changes in v3:
> >          - Use PSCI_RET where needed
> > ---
> >   xen/arch/arm/Makefile            |  1 -
> >   xen/arch/arm/platforms/exynos5.c |  3 ++-
> >   xen/arch/arm/platforms/seattle.c |  4 ++--
> >   xen/arch/arm/psci.c              | 37 +++++++++++++++++++++++++------------
> >   xen/arch/arm/smc.S               | 21 ---------------------
> >   xen/include/asm-arm/processor.h  |  3 ---
> >   6 files changed, 29 insertions(+), 40 deletions(-)
> >   delete mode 100644 xen/arch/arm/smc.S
> >
> > diff --git a/xen/arch/arm/Makefile b/xen/arch/arm/Makefile
> > index b9b141dc84..37fa8268b3 100644
> > --- a/xen/arch/arm/Makefile
> > +++ b/xen/arch/arm/Makefile
> > @@ -39,7 +39,6 @@ obj-y += processor.o
> >   obj-y += psci.o
> >   obj-y += setup.o
> >   obj-y += shutdown.o
> > -obj-y += smc.o
> >   obj-y += smp.o
> >   obj-y += smpboot.o
> >   obj-y += sysctl.o
> > diff --git a/xen/arch/arm/platforms/exynos5.c b/xen/arch/arm/platforms/exynos5.c
> > index c15ecf80f5..e2c0b7b878 100644
> > --- a/xen/arch/arm/platforms/exynos5.c
> > +++ b/xen/arch/arm/platforms/exynos5.c
> > @@ -26,6 +26,7 @@
> >   #include <asm/platforms/exynos5.h>
> >   #include <asm/platform.h>
> >   #include <asm/io.h>
> > +#include <asm/smccc.h>
> >   
> >   static bool secure_firmware;
> >   
> > @@ -249,7 +250,7 @@ static int exynos5_cpu_up(int cpu)
> >       iounmap(power);
> >   
> >       if ( secure_firmware )
> > -        call_smc(SMC_CMD_CPU1BOOT, cpu, 0, 0);
> > +        arm_smccc_smc(SMC_CMD_CPU1BOOT, cpu, NULL);
> >   
> >       return cpu_up_send_sgi(cpu);
> >   }
> > diff --git a/xen/arch/arm/platforms/seattle.c b/xen/arch/arm/platforms/seattle.c
> > index 893cc17972..64cc1868c2 100644
> > --- a/xen/arch/arm/platforms/seattle.c
> > +++ b/xen/arch/arm/platforms/seattle.c
> > @@ -33,12 +33,12 @@ static const char * const seattle_dt_compat[] __initconst =
> >    */
> >   static void seattle_system_reset(void)
> >   {
> > -    call_smc(PSCI_0_2_FN32_SYSTEM_RESET, 0, 0, 0);
> > +    arm_smccc_smc(PSCI_0_2_FN32_SYSTEM_RESET, NULL);
> >   }
> >   
> >   static void seattle_system_off(void)
> >   {
> > -    call_smc(PSCI_0_2_FN32_SYSTEM_OFF, 0, 0, 0);
> > +    arm_smccc_smc(PSCI_0_2_FN32_SYSTEM_OFF, NULL);
> >   }
> >   
> >   PLATFORM_START(seattle, "SEATTLE")
> > diff --git a/xen/arch/arm/psci.c b/xen/arch/arm/psci.c
> > index 941eec921b..4ae6504f3e 100644
> > --- a/xen/arch/arm/psci.c
> > +++ b/xen/arch/arm/psci.c
> > @@ -42,42 +42,53 @@ uint32_t smccc_ver;
> >   
> >   static uint32_t psci_cpu_on_nr;
> >   
> > +#define PSCI_RET(res)   ((int32_t)(res).a0)
> > +
> >   int call_psci_cpu_on(int cpu)
> >   {
> > -    return call_smc(psci_cpu_on_nr, cpu_logical_map(cpu), 
> > __pa(init_secondary), 0);
> > +    struct arm_smccc_res res;
> > +
> > +    arm_smccc_smc(psci_cpu_on_nr, cpu_logical_map(cpu), __pa(init_secondary),
> > +                  &res);
> > +
> > +    return PSCI_RET(res.a0);
> >   }
> Sorry if I'm jumping into the middle of a conversation, but why force
> all callers to manually extract the return value when it is a fixed
> register?
The SMCCC allows up to 4 return value (a0 ... a3). At the moment, most 
of the caller in Xen only care about one result value. But this will 
change soon (see OP-TEE support in Xen).
> Wouldn't it be far easier to do this:
>
> #define arcm_smccc_smc(...)                         \
>      ({                                              \
>          struct arm_smccc_res res;                   \
>                                                      \
>          if ( cpus_have_const_cap(ARM_SMCCC_1_1) )   \
>              res = arm_smccc_1_1_smc(__VA_ARGS__);   \
>          else                                        \
>              res = arm_smccc_1_0_smc(__VA_ARGS__);   \
>                                                      \
>          (int)res.a0;                                \
We can't possibly cast here. The interpretation of a0 is different from 
one call to another.
>      })
>
> Which also allows the compiler to optimise out the structure if it isn't
> read, and also avoids the caller needing to pass a NULL pointer for "I
> don't want the result".
While I understand the value, I don't think this would be correct if we 
want to implement an interface with the SMCCC.
Cheers,

--
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel