On 09/20/2017 08:41 AM, Jan Kara wrote:
> On Tue 19-09-17 13:53:06, Jens Axboe wrote:
>> Now that we have no external callers of wb_start_writeback(),
>> we can move the nr_pages == 0 logic into that function.
>>
>> Signed-off-by: Jens Axboe <axboe@xxxxxxxxx>
>
> ...
>
>> +static unsigned long get_nr_dirty_pages(void)
>> +{
>> + return global_node_page_state(NR_FILE_DIRTY) +
>> + global_node_page_state(NR_UNSTABLE_NFS) +
>> + get_nr_dirty_inodes();
>> +}
>> +
>> static void wb_start_writeback(struct bdi_writeback *wb, long nr_pages,
>> bool range_cyclic, enum wb_reason reason)
>> {
>> @@ -942,6 +953,12 @@ static void wb_start_writeback(struct bdi_writeback *wb, long nr_pages,
>> return;
>>
>> /*
>> + * If someone asked for zero pages, we write out the WORLD
>> + */
>> + if (!nr_pages)
>> + nr_pages = get_nr_dirty_pages();
>> +
>
> So for 'wb' we have a better estimate of the amount we should write - use
> wb_stat_sum(wb, WB_RECLAIMABLE) statistics - that is essentially dirty +
> unstable_nfs broken down to bdi_writeback.
I don't mind making that change, but I think that should be a separate
patch. We're using get_nr_dirty_pages() in existing locations where
we have the 'wb', like in wb_check_old_data_flush().
--
Jens Axboe
