| Message ID | 20240522072959.23741-2-richard.weiyang@gmail.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | [1/2] mm/mm_init.c: use memblock_region_memory_base_pfn() to get startpfn | expand |
On Wed, May 22, 2024 at 07:29:59AM +0000, Wei Yang wrote: > Since we always populate low zones, it is not necessary to count > nr_initialised for them. Let's move the last zone check ahead to skip > the nr_initialised reset. We also don't need to count initialized pfns after first_deferred_pfn is detected, so the reset of nr_initialised can be moved even lower. > Assume low zones counts for 4G, this change may reduce 1M times > comparison of prev_end_pfn and end_pfn. > > Signed-off-by: Wei Yang <richard.weiyang@gmail.com> > --- > mm/mm_init.c | 9 +++++---- > 1 file changed, 5 insertions(+), 4 deletions(-) > > diff --git a/mm/mm_init.c b/mm/mm_init.c > index 2dfb87841fdb..4cc627b4ebfe 100644 > --- a/mm/mm_init.c > +++ b/mm/mm_init.c > @@ -676,6 +676,11 @@ defer_init(int nid, unsigned long pfn, unsigned long end_pfn) > > if (early_page_ext_enabled()) > return false; > + > + /* Always populate low zones for address-constrained allocations */ > + if (end_pfn < pgdat_end_pfn(NODE_DATA(nid))) > + return false; > + > /* > * prev_end_pfn static that contains the end of previous zone > * No need to protect because called very early in boot before smp_init. > @@ -685,10 +690,6 @@ defer_init(int nid, unsigned long pfn, unsigned long end_pfn) > nr_initialised = 0; > } > > - /* Always populate low zones for address-constrained allocations */ > - if (end_pfn < pgdat_end_pfn(NODE_DATA(nid))) > - return false; > - > if (NODE_DATA(nid)->first_deferred_pfn != ULONG_MAX) > return true; > /* > -- > 2.34.1 >
On Thu, May 23, 2024 at 05:53:09PM +0300, Mike Rapoport wrote: >On Wed, May 22, 2024 at 07:29:59AM +0000, Wei Yang wrote: >> Since we always populate low zones, it is not necessary to count >> nr_initialised for them. Let's move the last zone check ahead to skip >> the nr_initialised reset. > >We also don't need to count initialized pfns after first_deferred_pfn is >detected, so the reset of nr_initialised can be moved even lower. > Agree, will update in v2.
diff --git a/mm/mm_init.c b/mm/mm_init.c index 2dfb87841fdb..4cc627b4ebfe 100644 --- a/mm/mm_init.c +++ b/mm/mm_init.c @@ -676,6 +676,11 @@ defer_init(int nid, unsigned long pfn, unsigned long end_pfn) if (early_page_ext_enabled()) return false; + + /* Always populate low zones for address-constrained allocations */ + if (end_pfn < pgdat_end_pfn(NODE_DATA(nid))) + return false; + /* * prev_end_pfn static that contains the end of previous zone * No need to protect because called very early in boot before smp_init. @@ -685,10 +690,6 @@ defer_init(int nid, unsigned long pfn, unsigned long end_pfn) nr_initialised = 0; } - /* Always populate low zones for address-constrained allocations */ - if (end_pfn < pgdat_end_pfn(NODE_DATA(nid))) - return false; - if (NODE_DATA(nid)->first_deferred_pfn != ULONG_MAX) return true; /*
Since we always populate low zones, it is not necessary to count nr_initialised for them. Let's move the last zone check ahead to skip the nr_initialised reset. Assume low zones counts for 4G, this change may reduce 1M times comparison of prev_end_pfn and end_pfn. Signed-off-by: Wei Yang <richard.weiyang@gmail.com> --- mm/mm_init.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-)