On 9 April 2015 at 18:23, Christopher Covington
<christopher.coving...@linaro.org> wrote:
> On Fri, Mar 27, 2015 at 12:40 PM, Peter Maydell
> <peter.mayd...@linaro.org> wrote:
>> You need to have the semihosting_enabled check here rather
>> than in the do_interrupt code, because otherwise we won't
>> behave correctly in the disabled case.
>
> Do you have suggestions for getting semihosting_enabled defined in
> translate-a64.c? I'm likely doing something dumb, but while #include
> "sysemu/sysemu.h" at first seemed like the obvious approach, and
> appears to work for -softmmu, I'm getting errors with that when
> building -linux-user.
sysemu.h is an ugly grab-bag of things that are specific to the
system emulator config. Semihosting is always enabled for linux-user,
so it doesn't have an equivalent switch.
Let's clean this up a bit, by creating a new include/exec/semihost.h,
with contents something like:
#ifdef CONFIG_SOFTMMU
bool semihosting_enabled(void);
#else
/* In -user configurations semihosting is always enabled */
bool semihosting_enabled(void)
{
return true;
}
#endif
[plus license header, ifdef guards, etc]
and then in vl.c have a
static bool semihosting_allowed;
and
bool semihosting_enabled(void)
{
return semihosting_allowed;
}
and change the places that set semihosting_enabled to set
this private flag instead. [This is following the pattern
of the existing defaults_enabled/usb_enabled functions.]
And change all the tests of semihosting_enabled to calls to
semihosting_enabled(), and delete the semihosting_enabled
variable completely.
Then we have a place to put future semihosting related
config things, and a small self-contained header you can
include in translate-a64.c without dragging in a lot of
rubbish.
This should all go in a preliminary patch at the start of your
series.
thanks
-- PMM
