The GICv3 has some registers that support byte accesses, and some
that support 8-byte accesses. Our TCG implementation implements all
of this, switching on the 'size' argument and handling the registers
that must support reads of that size while logging an error for
attempted accesses to registers that do not support that size access.
However we forgot to tell the core memory subsystem about this by
specifying the .impl and .valid fields in the MemoryRegionOps struct,
so the core was happily simulating 8 byte accesses by combining two 4
byte accesses. This doesn't have much guest-visible effect, since
there aren't many 8 byte registers and they all support being written
in two 4 byte parts.
Set the .impl and .valid fields to say that all sizes from 1 to 8
bytes are both valid and implemented by the device.
Signed-off-by: Peter Maydell
---
hw/intc/arm_gicv3.c | 8
1 file changed, 8 insertions(+)
diff --git a/hw/intc/arm_gicv3.c b/hw/intc/arm_gicv3.c
index 6d3c8ee231c..0b8f79a1227 100644
--- a/hw/intc/arm_gicv3.c
+++ b/hw/intc/arm_gicv3.c
@@ -369,11 +369,19 @@ static const MemoryRegionOps gic_ops[] = {
.read_with_attrs = gicv3_dist_read,
.write_with_attrs = gicv3_dist_write,
.endianness = DEVICE_NATIVE_ENDIAN,
+.valid.min_access_size = 1,
+.valid.max_access_size = 8,
+.impl.min_access_size = 1,
+.impl.max_access_size = 8,
},
{
.read_with_attrs = gicv3_redist_read,
.write_with_attrs = gicv3_redist_write,
.endianness = DEVICE_NATIVE_ENDIAN,
+.valid.min_access_size = 1,
+.valid.max_access_size = 8,
+.impl.min_access_size = 1,
+.impl.max_access_size = 8,
}
};
--
2.25.1
