On Wed, Jun 21, 2017 at 03:10:36PM +0200, Paolo Bonzini wrote:
>
>
> On 21/06/2017 12:19, Yang Zhong wrote:
> > @@ -3738,8 +3738,13 @@ int main(int argc, char **argv, char **envp)
> > }
> > break;
> > case QEMU_OPTION_no_kvm:
> > +#ifdef CONFIG_TCG
> > olist = qemu_find_opts("machine");
> > qemu_opts_parse_noisily(olist, "accel=tcg", false);
> > +#else
> > + error_report("TCG is disabled");
>
> This should not be necessary, it would work anyway.
>
Hello Paolo,
You are right, i justly did the test to verify your comment, i will remove
this CONFIG_TCG. thanks!
Yang
> > + exit(1);
> > +#endif
> > break;
> > case QEMU_OPTION_no_kvm_pit: {
> > error_report("warning: ignoring deprecated option");
> > @@ -3915,10 +3920,15 @@ int main(int argc, char **argv, char **envp)
> > configure_rtc(opts);
> > break;
> > case QEMU_OPTION_tb_size:
> > +#ifdef CONFIG_TCG
> > tcg_tb_size = strtol(optarg, NULL, 0);
> > if (tcg_tb_size < 0) {
> > tcg_tb_size = 0;
> > }
> > +#else
> > + error_report("TCG is disabled");
> > + exit(1);
> > +#endif
>
> This would regress the case where people are using -tb-size with KVM.
> Just leave it aside.
>
Hello Paolo,
The tb_size is only useful in TCG mode, only tcg_init() will call this
variable.
If TCG is disabled in compile, the KVM will not need this variable. Of
course, if
we keep this variable in pure KVM mode, there is not any side effect. What's
your
opinion? thanks!
Yang
> > break;
> > case QEMU_OPTION_icount:
> > icount_opts =
> > qemu_opts_parse_noisily(qemu_find_opts("icount"),
> > @@ -4457,7 +4467,9 @@ int main(int argc, char **argv, char **envp)
> > qemu_opts_del(icount_opts);
> > }
> >
> > +#ifdef CONFIG_TCG
> > qemu_tcg_configure(accel_opts, &error_fatal);
>
> Why is this appearing in this patch?
>
Hello Paolo,
If the TCG is disabled, the qemu_tcg_configure() is not useful, this function
only set
the single or mutiple thread for TCG. thanks!
Yang
> > +#endif
