Re: [PATCH v5 20/22] target/arm: Create tagged ram when MTE is enabled

[Peter Maydell <peter.maydell@linaro.org>]

On Fri, 11 Oct 2019 at 14:50, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
> v5: Assign cs->num_ases to the final value first.
>     Downgrade to ID_AA64PFR1.MTE=1 if tag memory is not available.
> v6: Add secure tag memory for EL3.
> ---
>  target/arm/cpu.h |  6 ++++++
>  hw/arm/virt.c    | 54 ++++++++++++++++++++++++++++++++++++++++++++++++
>  target/arm/cpu.c | 53 ++++++++++++++++++++++++++++++++++++++++++++---
>  3 files changed, 110 insertions(+), 3 deletions(-)
>
> diff --git a/target/arm/cpu.h b/target/arm/cpu.h
> index 93a362708b..faca43ea78 100644
> --- a/target/arm/cpu.h
> +++ b/target/arm/cpu.h
> @@ -765,6 +765,10 @@ struct ARMCPU {
>      /* MemoryRegion to use for secure physical accesses */
>      MemoryRegion *secure_memory;
>
> +    /* MemoryRegion to use for allocation tag accesses */
> +    MemoryRegion *tag_memory;
> +    MemoryRegion *secure_tag_memory;
> +
>      /* For v8M, pointer to the IDAU interface provided by board/SoC */
>      Object *idau;
>
> @@ -2956,6 +2960,8 @@ int cpu_mmu_index(CPUARMState *env, bool ifetch);
>  typedef enum ARMASIdx {
>      ARMASIdx_NS = 0,
>      ARMASIdx_S = 1,
> +    ARMASIdx_TagNS = 2,
> +    ARMASIdx_TagS = 3,
>  } ARMASIdx;
>
>  /* Return the Exception Level targeted by debug exceptions. */
> diff --git a/hw/arm/virt.c b/hw/arm/virt.c
> index d74538b021..573988ba4d 100644
> --- a/hw/arm/virt.c
> +++ b/hw/arm/virt.c
> @@ -1330,6 +1330,18 @@ static void create_secure_ram(VirtMachineState *vms,
>      g_free(nodename);
>  }
>
> +static void create_tag_ram(MemoryRegion *tag_sysmem,
> +                           hwaddr base, hwaddr size,
> +                           const char *name)
> +{
> +    MemoryRegion *tagram = g_new(MemoryRegion, 1);
> +
> +    memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
> +    memory_region_add_subregion(tag_sysmem, base / 32, tagram);
> +
> +    /* ??? Do we really need an fdt entry, or is MemTag enabled sufficient.  */

What's this '???' asking about? I would be surprised if the
kernel expected to have any kind of FDT for tag RAM
(with the exception that an implementation that puts tags
in a special part of normal-ram will want that not
to be described in the fdt as ram usable by the kernel), but
we should ask the kernel folks.

> +}
> +
>  static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
>  {
>      const VirtMachineState *board = container_of(binfo, VirtMachineState,
> @@ -1485,6 +1497,8 @@ static void machvirt_init(MachineState *machine)
>      qemu_irq pic[NUM_IRQS];
>      MemoryRegion *sysmem = get_system_memory();
>      MemoryRegion *secure_sysmem = NULL;
> +    MemoryRegion *tag_sysmem = NULL;
> +    MemoryRegion *secure_tag_sysmem = NULL;
>      int n, virt_max_cpus;
>      MemoryRegion *ram = g_new(MemoryRegion, 1);
>      bool firmware_loaded;
> @@ -1652,6 +1666,35 @@ static void machvirt_init(MachineState *machine)
>                                       "secure-memory", &error_abort);
>          }
>
> +        /*
> +         * The cpu adds the property iff MemTag is supported.

We've had confusion before from non-native-speakers and
non-maths-geeks about 'iff' in comments; better to expand to
'if and only if'.

> +         * If it is, we must allocate the ram to back that up.
> +         */
> +        if (object_property_find(cpuobj, "tag-memory", NULL)) {
> +            if (!tag_sysmem) {
> +                tag_sysmem = g_new(MemoryRegion, 1);
> +                memory_region_init(tag_sysmem, OBJECT(machine),
> +                                   "tag-memory", UINT64_MAX / 32);
> +
> +                if (vms->secure) {
> +                    secure_tag_sysmem = g_new(MemoryRegion, 1);
> +                    memory_region_init(secure_tag_sysmem, OBJECT(machine),
> +                                       "secure-tag-memory", UINT64_MAX / 32);
> +
> +                    /* As with ram, secure-tag takes precedence over tag.  */
> +                    memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
> +                                                        tag_sysmem, -1);
> +                }
> +            }

Are there really separate S and NS tag RAMs?

thanks
-- PMM