[Qemu-devel] Re: [PATCH] add VMSTATE_BOOL

[Michael S. Tsirkin]

On Tue, Nov 09, 2010 at 10:37:37AM +0100, Gerd Hoffmann wrote:
>   Hi,
> 
> >>+#define VMSTATE_BOOL_ARRAY(_f, _s, _n)                               \
> >>+    VMSTATE_BOOL_ARRAY_V(_f, _s, _n, 0)
> >>+
> >
> >Why don't we pack the bits?
> 
> Point being?  As long as we don't save *big* arrays of bools it
> simply isn't worth the effort IMHO.  And for big arrays we'll
> probably wouldn't use bool in the first place ...
> 
> >>+/* bool */
> >>+
> >>+static int get_bool(QEMUFile *f, void *pv, size_t size)
> >>+{
> >>+    bool *v = pv;
> >>+    *v = qemu_get_byte(f);
> >>+    return 0;
> >
> >We must really validate that the value is 0 or 1.
> >If it's not, we will get undefined behaviour.
> 
> I disagree.
> 
> You indeed have a bug in case your bool ends up with a value being
> neither 0 nor 1.  That is completely independant from savevm/loadvm
> though, it can trip you up even in case you don't save/load the VM
> at all.

I was wrong about undefined behaviour. Sorry.
What this implementation does is treat byte value '\0'
as boolean false, any other value as true.

I think we should verify that value is 0 or 1 and fail
migration otherwise, to make it more robust.


> >>+}
> >>+
> >>+static void put_bool(QEMUFile *f, void *pv, size_t size)
> >>+{
> >>+    bool *v = pv;
> >>+    qemu_put_byte(f, *v);
> >
> >Is there a guarantee that bool is a single byte, BTW?
> 
> No.  bool must be 0 or 1 though, and a single byte is big enough to
> keep that information.
> 
> cheers,
>   Gerd

Right.