hw/misc/aspeed_scu: Correct minimum access size for AST2500 / AST2600

Guest code was performing a byte load to the SCU MMIO region, leading
to the guest code crashing (it should be using proper accessors, but
that is not Qemu's bug). Hardware and the documentation[1] both agree
that byte loads are okay, so change all of the aspeed SCU devices to
accept a minimum access size of 1.

[1] See the 'ARM Address Space Mapping' table in the ASPEED docs. This
is section 6.1 in the ast2400 and ast2700, and 7.1 in the ast2500 and
ast2600 datasheets.

Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2636
Signed-off-by: Joel Stanley <joel@jms.id.au>
Reviewed-by: Troy Lee <leetroy@gmail.com>
Message-ID: <20241118021820.4928-1-joel@jms.id.au>
[PMD: Rebased, only including SCU changes]
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Andrew Jeffery <andrew@codeconstruct.com.au>
Link: https://lore.kernel.org/qemu-devel/20250331230444.88295-3-philmd@linaro.org
Signed-off-by: Cédric Le Goater <clg@redhat.com>

hw/misc/aspeed_scu.c

@@ -443,7 +443,7 @@ static const MemoryRegionOps aspeed_ast2500_scu_ops = {
     .endianness = DEVICE_LITTLE_ENDIAN,
     .impl.min_access_size = 4,
     .impl.max_access_size = 4,
-    .valid.min_access_size = 4,
+    .valid.min_access_size = 1,
     .valid.max_access_size = 4,
     .valid.unaligned = false,
 };
@@ -787,7 +787,7 @@ static const MemoryRegionOps aspeed_ast2600_scu_ops = {
     .endianness = DEVICE_LITTLE_ENDIAN,
     .impl.min_access_size = 4,
     .impl.max_access_size = 4,
-    .valid.min_access_size = 4,
+    .valid.min_access_size = 1,
     .valid.max_access_size = 4,
     .valid.unaligned = false,
 };