On Tue, Dec 11, 2018 at 10:44:41AM +0100, Michal Hocko wrote: >On Tue 11-12-18 11:51:28, Wei Yang wrote: >> A valid present section number is in [0, __highest_present_section_nr]. >> And the return value of next_present_section_nr() meets this >> requirement. This means it is not necessary to check it with >> __highest_present_section_nr again in for_each_present_section_nr(). >> >> Since we pass an unsigned long *section_nr* to >> for_each_present_section_nr(), we need to cast it to int before >> comparing. > >Why do we want this patch? Is it an improvement? If yes, it is >performance visible change or does it make the code easier to maintain? > Michal I know you concern, maintainance is a very critical part of review. >To me at least the later seems dubious to be honest because it adds a >non-obvious dependency of the terminal condition to the >next_present_section_nr implementation and that might turn out error >prone. > While I think the original code is not that clear about the syntax. When we look at the next_present_section_nr(section_nr), the return value falls into two categories: -1 : no more present section after section_nr other: the next present section number after section_nr Based on this syntax, the iteration could be simpler to terminate when the return value is less than 0. This is what the patch tries to do. Maybe I could do more to help the maintainance: * add some comment about the return value of next_present_section_nr * terminate the loop when section_nr == -1 Hope this would help a little. -- Wei Yang Help you, Help me
