Initialize the CPU interface registers also
for Cortex-A15 and Cortex-A7 CPU models, in
the same way as done for 64bit CPU models.
This fixes usage of GICv3 in virtualization
contexts in 32bit configurations.

Signed-off-by: Adam Lackorzynski <adam@l4re.org>
---
 target/arm/cpu.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/target/arm/cpu.c b/target/arm/cpu.c
index 32bec156f2..f525d45f6a 100644
--- a/target/arm/cpu.c
+++ b/target/arm/cpu.c
@@ -1972,6 +1972,9 @@ static void cortex_a7_initfn(Object *obj)
     cpu->ccsidr[0] = 0x701fe00a; /* 32K L1 dcache */
     cpu->ccsidr[1] = 0x201fe00a; /* 32K L1 icache */
     cpu->ccsidr[2] = 0x711fe07a; /* 4096K L2 unified cache */
+    cpu->gic_num_lrs = 4;
+    cpu->gic_vpribits = 5;
+    cpu->gic_vprebits = 5;
     define_arm_cp_regs(cpu, cortexa15_cp_reginfo); /* Same as A15 */
 }
 
@@ -2014,6 +2017,9 @@ static void cortex_a15_initfn(Object *obj)
     cpu->ccsidr[0] = 0x701fe00a; /* 32K L1 dcache */
     cpu->ccsidr[1] = 0x201fe00a; /* 32K L1 icache */
     cpu->ccsidr[2] = 0x711fe07a; /* 4096K L2 unified cache */
+    cpu->gic_num_lrs = 4;
+    cpu->gic_vpribits = 5;
+    cpu->gic_vprebits = 5;
     define_arm_cp_regs(cpu, cortexa15_cp_reginfo);
 }
 
-- 
2.27.0.rc2

On Sat, 30 May 2020 at 00:07, Adam Lackorzynski <adam@l4re.org> wrote:
>
> Initialize the CPU interface registers also
> for Cortex-A15 and Cortex-A7 CPU models, in
> the same way as done for 64bit CPU models.
> This fixes usage of GICv3 in virtualization
> contexts in 32bit configurations.
>
> Signed-off-by: Adam Lackorzynski <adam@l4re.org>

Hi; I'm confused by this patch. The Cortex-A15 and Cortex-A7
do not have or support the GICv3, so why would we need
to set GICv3-specific settings for them?

We're probably missing a sanity-check somewhere
to forbid user attempts to use non-GICv3 CPUs with
the GICv3.

thanks
-- PMM

On Mon Jun 01, 2020 at 13:36:13 +0100, Peter Maydell wrote:
> On Sat, 30 May 2020 at 00:07, Adam Lackorzynski <adam@l4re.org> wrote:
> >
> > Initialize the CPU interface registers also
> > for Cortex-A15 and Cortex-A7 CPU models, in
> > the same way as done for 64bit CPU models.
> > This fixes usage of GICv3 in virtualization
> > contexts in 32bit configurations.
> >
> > Signed-off-by: Adam Lackorzynski <adam@l4re.org>
> 
> Hi; I'm confused by this patch. The Cortex-A15 and Cortex-A7
> do not have or support the GICv3, so why would we need
> to set GICv3-specific settings for them?
> 
> We're probably missing a sanity-check somewhere
> to forbid user attempts to use non-GICv3 CPUs with
> the GICv3.

Indeed that's another option. Besides that A15+GICv3 currently just
works with this.
So I guess the alternative is to switch to 32bit from EL3 when using an A57?


Thanks,
Adam

On Mon, 1 Jun 2020 at 14:01, Adam Lackorzynski <adam@l4re.org> wrote:
> Indeed that's another option. Besides that A15+GICv3 currently just
> works with this.

Not sure how...

> So I guess the alternative is to switch to 32bit from EL3 when using an A57?

If you're using KVM then you can use -cpu host,aarch64=off

thanks
-- PMM