[PATCH v5 3/9] writeback: convert variables to unsigned

[Greg Thelen]

On Sun, Feb 27, 2011 at 8:07 AM, Minchan Kim <minchan.kim at gmail.com> wrote:
> On Fri, Feb 25, 2011 at 01:35:54PM -0800, Greg Thelen wrote:
>> Convert two balance_dirty_pages() page counter variables (nr_reclaimable
>> and nr_writeback) from 'long' to 'unsigned long'.
>>
>> These two variables are used to store results from global_page_state().
>> global_page_state() returns unsigned long and carefully sums per-cpu
>> counters explicitly avoiding returning a negative value.
>>
>> Signed-off-by: Greg Thelen <gthelen at google.com>
> Reviewed-by: Minchan Kim <minchan.kim at gmail.com>
>
>> ---
>> Changelog since v4:
>> - Created this patch for clarity.  Previously this patch was integrated within
>>   the "writeback: create dirty_info structure" patch.
>>
>>  mm/page-writeback.c |    6 ++++--
>>  1 files changed, 4 insertions(+), 2 deletions(-)
>>
>> diff --git a/mm/page-writeback.c b/mm/page-writeback.c
>> index 2cb01f6..4408e54 100644
>> --- a/mm/page-writeback.c
>> +++ b/mm/page-writeback.c
>> @@ -478,8 +478,10 @@ unsigned long bdi_dirty_limit(struct backing_dev_info *bdi, unsigned long dirty)
>>  static void balance_dirty_pages(struct address_space *mapping,
>>                               unsigned long write_chunk)
>>  {
>> -     long nr_reclaimable, bdi_nr_reclaimable;
>> -     long nr_writeback, bdi_nr_writeback;
>> +     unsigned long nr_reclaimable;
>> +     long bdi_nr_reclaimable;
>> +     unsigned long nr_writeback;
>> +     long bdi_nr_writeback;
>>       unsigned long background_thresh;
>>       unsigned long dirty_thresh;
>>       unsigned long bdi_thresh;
>> --
>> 1.7.3.1
>>
> bdi_nr_[reclaimable|writeback] can return negative value?
> When I just look through bdi_stat_sum, it uses *percpu_counter_sum_positive*.
> So I guess it always returns positive value.
> If it is right, could you change it, too?

Yes, I think we can also change bdi_nr_[reclaimable|writeback] to unsigned long.

bdi_stat_sum() and bdi_stat() both call percpu_counter_sum_positive(),
which return a positive number.  bdi_stat[_sum]() return s64.  Should
we also change bdi_stat[_sum]() to return unsigned long rather than
s64?  I would like the return value type to match the type of the
corresponding local variables in balance_dirty_pages().  All current
callers appear to expect bdi_stat[_sum]() to return unsigned long.

> --
> Kind regards,
> Minchan Kim
>