On Mon, Dec 10, 2018 at 3:46 PM Vlastimil Babka <vbabka@xxxxxxx> wrote:
>
> On 12/10/18 4:15 AM, Pingfan Liu wrote:
> > On Fri, Dec 7, 2018 at 3:36 PM Vlastimil Babka <vbabka@xxxxxxx> wrote:
> >>
> >> On 12/7/18 5:53 AM, Pingfan Liu wrote:
> >>> Currently, NR_PAGEBLOCK_BITS and MIGRATE_TYPES are not associated by code.
> >>> If someone adds extra migrate type, then he may forget to enlarge the
> >>> NR_PAGEBLOCK_BITS. Hence it requires some way to fix.
> >>> NR_PAGEBLOCK_BITS depends on MIGRATE_TYPES, while these macro
> >>> spread on two different .h file with reverse dependency, it is a little
> >>> hard to refer to MIGRATE_TYPES in pageblock-flag.h. This patch tries to
> >>> remind such relation in compiling-time.
> >>>
> >>> Signed-off-by: Pingfan Liu <kernelfans@xxxxxxxxx>
> >>> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> >>> Cc: Michal Hocko <mhocko@xxxxxxxx>
> >>> Cc: Pavel Tatashin <pavel.tatashin@xxxxxxxxxxxxx>
> >>> Cc: Vlastimil Babka <vbabka@xxxxxxx>
> >>> Cc: Oscar Salvador <osalvador@xxxxxxx>
> >>> Cc: Mike Rapoport <rppt@xxxxxxxxxxxxxxxxxx>
> >>> Cc: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx>
> >>> Cc: Alexander Duyck <alexander.h.duyck@xxxxxxxxxxxxxxx>
> >>> ---
> >>> include/linux/pageblock-flags.h | 5 +++--
> >>> mm/page_alloc.c | 3 ++-
> >>> 2 files changed, 5 insertions(+), 3 deletions(-)
> >>>
> >>> diff --git a/include/linux/pageblock-flags.h b/include/linux/pageblock-flags.h
> >>> index 9132c5c..fe0aec4 100644
> >>> --- a/include/linux/pageblock-flags.h
> >>> +++ b/include/linux/pageblock-flags.h
> >>> @@ -25,11 +25,12 @@
> >>>
> >>> #include <linux/types.h>
> >>>
> >>> +#define PB_migratetype_bits 3
> >>> /* Bit indices that affect a whole block of pages */
> >>> enum pageblock_bits {
> >>> PB_migrate,
> >>> - PB_migrate_end = PB_migrate + 3 - 1,
> >>> - /* 3 bits required for migrate types */
> >>> + PB_migrate_end = PB_migrate + PB_migratetype_bits - 1,
> >>> + /* n bits required for migrate types */
> >>> PB_migrate_skip,/* If set the block is skipped by compaction */
> >>>
> >>> /*
> >>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> >>> index 2ec9cc4..1a22d8d 100644
> >>> --- a/mm/page_alloc.c
> >>> +++ b/mm/page_alloc.c
> >>> @@ -425,7 +425,8 @@ void set_pfnblock_flags_mask(struct page *page, unsigned long flags,
> >>> unsigned long bitidx, word_bitidx;
> >>> unsigned long old_word, word;
> >>>
> >>> - BUILD_BUG_ON(NR_PAGEBLOCK_BITS != 4);
> >>
> >> Why delete this one? It's for something a bit different and also still
> >> valid.
> >>
> >
> > I think it is dependent on PB_migratetype_bits (plus 1 dedicated bit
> > for skip), hence the later one implies it.
>
> No, NR_PAGEBLOCK_BITS simply to be 4 (or other number that can divide
> the number of bits in a word with no remainder), or
> set_pfnblock_flags_mask() will not function correctly.
>
Yes, you are right. It will break the atomic.
> >>> + BUILD_BUG_ON(order_base_2(MIGRATE_TYPES)
> >>> + != (PB_migratetype_bits - 1));
> >>
> >> I think this should use the '>' operator. It's fine if there are less
> >
> > If ">" is chosen, then it allows wasted bits. I had thought it is not
> > the purpose of the design, hence disard the ">".
>
> We do allow wasted bits, because we need NR_PAGEBLOCK_BITS to be 4 in
> any case, as explained above.
>
Right!
> > Otherwise, what about
> > using warning on ">"?
>
> I think it would be needed.
>
OK.
> >> types than what can fit into 3 bits. AFAICS for !CONFIG_DMA and
> >> !CONFIG_MEMORY_ISOLATION there are just 4 types that fit into 2 bits...
> >>
> > Oh, yes, you are right. How about this:
> > #if defined(CONFIG_DMA) || defined(CONFIG_MEMORY_ISOLATION)
> > #define PB_migratetype_bits 3
> > #else
> > #define PB_migratetype_bits 2
> > #endif
>
> I think it's not necessary. Really, we need to have 4 NR_PAGEBLOCK_BITS
> with the current code, and this leaves us with 3 bits for migratetype.
> The only thing to check is whether migratetypes fit into these 3 bits, IMHO.
>
OK. I will update v3.
Thanks and regards,
Pingfan
