On 9/8/25 08:57, Peter Maydell wrote:
+static int get_pstate64_1(QEMUFile *f, void *opaque, size_t size,
+ const VMStateField *field)
Why the _1 suffix ?
Symbol conflict:
static const VMStateInfo vmstate_pstate64_1 = {
...
static const VMStateDescription vmstate_pstate64 = {
The get/set_pstate64_1 symbols are named to match usage.
Naming suggestions welcome. :-)
I note that there is actually a defined bit above 32
in the SPSR_ELx format for exceptions taken from
AArch32 to AArch64: PPEND, used with FEAT_SEBEP. That
suggests we should probably at least consider handling
64-bit AArch32 "CPSR" values, though FEAT_SEBEP in
particular may be out of scope for us.
Incidentally I think we are not correctly migrating
PSTATE.SS when in AArch32 -- we will migrate the CPSR
via cpsr_read() / cpsr_write(), but that view doesn't have
the PSTATE.SS bit in it. Possibly these things could
be addressed at the same time, so we have a subsection
for 64-bit pstate/cpsr, and its save/load uses
cpsr_read_for_spsr_elx() (and a corresponding _write_
that we don't have yet) when AArch32, and the .needed
function is "if top 32 bits not all zero, or PSTATE_SS bit
is set".
Good catch. I'll have a look at this.
r~
