On 3/15/20 9:39 PM, Richard Henderson wrote:
> On 3/13/20 11:36 AM, Philippe Mathieu-Daudé wrote:
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
>> ---
>> exec.c | 4 ----
>> 1 file changed, 4 deletions(-)
>>
>> diff --git a/exec.c b/exec.c
>> index 7bc9828c5b..f258502966 100644
>> --- a/exec.c
>> +++ b/exec.c
>> @@ -185,10 +185,6 @@ struct DirtyBitmapSnapshot {
>> unsigned long dirty[];
>> };
>>
>> -#endif
>> -
>> -#if !defined(CONFIG_USER_ONLY)
>> -
>> static void phys_map_node_reserve(PhysPageMap *map, unsigned nodes)
>> {
>> static unsigned alloc_hint = 16;
>>
>
> There's even more than that. Looking further down,
>
>> 745 #endif
>> 746
>> 747 #if !defined(CONFIG_USER_ONLY)
>
> This is the #endif that paired with the one at 190.
>
> Later,
>
>> 988 #if defined(CONFIG_USER_ONLY)
> ...
>> 1000 #else
> ...
>> 1031 #endif
>> 1032
>> 1033 #ifndef CONFIG_USER_ONLY
>
> So those three lines are redundant.
>
> Later,
>
>> 1252 #if !defined(CONFIG_USER_ONLY)
> ...
>> 1438 #endif /* defined(CONFIG_USER_ONLY) */
>> 1439
>> 1440 #if !defined(CONFIG_USER_ONLY)
>
> Clearly these ifdefs are very hard to follow. I would thus welcome a split of
> this file.
>
> Possibly into exec-common.c (with functions present in both softmmu and
> user-only, with ifdefs *inside* functions only), and exec-system.c (with no
> /#if.*CONFIG_USER_ONLY/).
>
> But exec.c is over 4000 lines, so if there's another logical split into even
> more files, that would be even better.
OK, I am taking notes for the 5.0 cycle.
We can drop this patch for this series objective, as it is mostly cosmetic.
>
>
> r~
>