On 21.11.18 07:04, Baoquan He wrote: > On 11/19/18 at 11:16am, David Hildenbrand wrote: >> diff --git a/kernel/crash_core.c b/kernel/crash_core.c >> index 933cb3e45b98..093c9f917ed0 100644 >> --- a/kernel/crash_core.c >> +++ b/kernel/crash_core.c >> @@ -464,6 +464,8 @@ static int __init crash_save_vmcoreinfo_init(void) >> VMCOREINFO_NUMBER(PAGE_BUDDY_MAPCOUNT_VALUE); >> #ifdef CONFIG_HUGETLB_PAGE >> VMCOREINFO_NUMBER(HUGETLB_PAGE_DTOR); >> +#define PAGE_OFFLINE_MAPCOUNT_VALUE (~PG_offline) >> + VMCOREINFO_NUMBER(PAGE_OFFLINE_MAPCOUNT_VALUE); >> #endif > > This solution looks good to me. One small concern is why we don't > export PG_offline to vmcoreinfo directly, then define > PAGE_OFFLINE_MAPCOUNT_VALUE in makedumpfile. We have been exporting > kernel data/MACRO directly, why this one is exceptional. > 1. We are much more similar to PG_buddy (in contrast to actual page flags), and for PG_buddy it is historically handled like this (and I think it makes sense to expose these as actual MAPCOUNT_VALUEs). 2. Right now only one page type per page is supported. Therefore only exactly one value in mapcount indicates e.g. PageBuddy()/PageOffline(). Now, if we ever decide to change this (e.g. treat them like real flags), it is much easier to switch to PG_offline/PG_buddy then. We can directly see in makedumpfile that .*_MAPCOUNT_VALUE is no longer available but instead e.g. PG_offline and PG_buddy. Instead we would no see a change in makedumpfile and would have to rely on other properties. If there are no strong opinions I will leave it like this. Thanks! > Thanks > Baoquan > -- Thanks, David / dhildenb
