| Message ID | 20160204165744.GG14425@dhcp22.suse.cz (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
2016-02-05 1:57 GMT+09:00 Michal Hocko <mhocko@kernel.org>: > [CCing Minchan] > > I am getting many compilation errors if !CONFIG_COMPACTION and > CONFIG_MEMORY_ISOLATION=y. I didn't get to check what has hanged in that > regards because this code is quite old but the previous mmotm seemed ok, > maybe some subtle change in my config? It would be caused by Vlastimil's "mm, hugetlb: don't require CMA for runtime gigantic pages". https://lkml.org/lkml/2016/2/3/841 Fix is already there. Thanks. -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri 05-02-16 02:02:08, Joonsoo Kim wrote: > 2016-02-05 1:57 GMT+09:00 Michal Hocko <mhocko@kernel.org>: > > [CCing Minchan] > > > > I am getting many compilation errors if !CONFIG_COMPACTION and > > CONFIG_MEMORY_ISOLATION=y. I didn't get to check what has hanged in that > > regards because this code is quite old but the previous mmotm seemed ok, > > maybe some subtle change in my config? > > It would be caused by Vlastimil's "mm, hugetlb: don't require CMA for > runtime gigantic pages". > > https://lkml.org/lkml/2016/2/3/841 > > Fix is already there. What would be the fix? Maybe I have missed it in the current mmotm...
On Thu 04-02-16 18:10:54, Michal Hocko wrote: > On Fri 05-02-16 02:02:08, Joonsoo Kim wrote: > > 2016-02-05 1:57 GMT+09:00 Michal Hocko <mhocko@kernel.org>: > > > [CCing Minchan] > > > > > > I am getting many compilation errors if !CONFIG_COMPACTION and > > > CONFIG_MEMORY_ISOLATION=y. I didn't get to check what has hanged in that > > > regards because this code is quite old but the previous mmotm seemed ok, > > > maybe some subtle change in my config? > > > > It would be caused by Vlastimil's "mm, hugetlb: don't require CMA for > > runtime gigantic pages". > > > > https://lkml.org/lkml/2016/2/3/841 > > > > Fix is already there. > > What would be the fix? Maybe I have missed it in the current mmotm... Ohh, v2, I should have read the rest of the thread.
diff --git a/mm/internal.h b/mm/internal.h index 271ad9567e32..f7c0024a013a 100644 --- a/mm/internal.h +++ b/mm/internal.h @@ -150,7 +150,7 @@ extern bool is_free_buddy_page(struct page *page); #endif extern int user_min_free_kbytes; -#if defined CONFIG_COMPACTION || defined CONFIG_CMA +#if defined CONFIG_COMPACTION || defined CONFIG_CMA || defined MEMORY_ISOLATION /* * in mm/compaction.c
[CCing Minchan] I am getting many compilation errors if !CONFIG_COMPACTION and CONFIG_MEMORY_ISOLATION=y. I didn't get to check what has hanged in that regards because this code is quite old but the previous mmotm seemed ok, maybe some subtle change in my config? --- mm/page_alloc.c:6893:6: warning: ‘struct compact_control’ declared inside parameter list [enabled by default] unsigned long start, unsigned long end) ^ mm/page_alloc.c:6893:6: warning: its scope is only this definition or declaration, which is probably not what you want [enabled by default] mm/page_alloc.c: In function ‘__alloc_contig_migrate_range’: mm/page_alloc.c:6903:37: error: dereferencing pointer to incomplete type while (pfn < end || !list_empty(&cc->migratepages)) { ^ mm/page_alloc.c:6909:21: error: dereferencing pointer to incomplete type if (list_empty(&cc->migratepages)) { ^ mm/page_alloc.c:6910:6: error: dereferencing pointer to incomplete type cc->nr_migratepages = 0; ^ mm/page_alloc.c:6911:4: error: implicit declaration of function ‘isolate_migratepages_range’ [-Werror=implicit-function-declaration] pfn = isolate_migratepages_range(cc, pfn, end); ^ mm/page_alloc.c:6922:50: error: dereferencing pointer to incomplete type nr_reclaimed = reclaim_clean_pages_from_list(cc->zone, ^ mm/page_alloc.c:6923:11: error: dereferencing pointer to incomplete type &cc->migratepages); ^ mm/page_alloc.c:6924:5: error: dereferencing pointer to incomplete type cc->nr_migratepages -= nr_reclaimed; ^ mm/page_alloc.c:6926:26: error: dereferencing pointer to incomplete type ret = migrate_pages(&cc->migratepages, alloc_migrate_target, ^ mm/page_alloc.c:6927:20: error: dereferencing pointer to incomplete type NULL, 0, cc->mode, MR_CMA); ^ mm/page_alloc.c:6930:28: error: dereferencing pointer to incomplete type putback_movable_pages(&cc->migratepages); ^ mm/page_alloc.c: In function ‘alloc_contig_range’: mm/page_alloc.c:6963:9: error: variable ‘cc’ has initializer but incomplete type struct compact_control cc = { ^ mm/page_alloc.c:6964:3: error: unknown field ‘nr_migratepages’ specified in initializer .nr_migratepages = 0, --- I have tried to do but that failed during linkage: mm/built-in.o: In function `alloc_contig_range': /dev/shm/mhocko/mmotm/mm/page_alloc.c:7062: undefined reference to `isolate_freepages_range' mm/built-in.o: In function `__alloc_contig_migrate_range': /dev/shm/mhocko/mmotm/mm/page_alloc.c:6911: undefined reference to `isolate_migratepages_range' So it really seems that CONFIG_MEMORY_ISOLATION depends on unctionality provided by CONFIG_COMPACTION. The whole MEMORY_ISOLATION came in ee6f509c3274 ("mm: factor out memory isolate functions") in 3.6 so quite some time ago. I do not have a good idea how to untangle this. We can move isolate_freepages_range and isolate_migratepages_range but then it needs to do ifdef for both MEMORY_ISOLATION and COMPACTION. Another option would be to make MEMORY_ISOLATION depend on COMPACTION. Do we ever need this functionality without compaction?