Re: [RESEND PATCH V3] NUMA:Improve the efficiency of calculating pages loss

[Mike Rapoport <rppt@kernel.org>]

On Tue, Aug 22, 2023 at 07:49:05PM +0800, Liam Ni wrote:
> On Tue, 15 Aug 2023 at 00:00, Mike Rapoport <rppt@kernel.org> wrote:
> >
> > On Fri, Aug 04, 2023 at 11:32:51PM +0800, Liam Ni wrote:
> > > Optimize the way of calculating missing pages.
> > >
> > > In the previous implementation, We calculate missing pages as follows:
> > > 1. calculate numaram by traverse all the numa_meminfo's and for each of
> > > them traverse all the regions in memblock.memory to prepare for
> > > counting missing pages.
> > >
> > > 2. Traverse all the regions in memblock.memory again to get e820ram.
> > >
> > > 3. the missing page is (e820ram - numaram )
> > >
> > > But,it's enough to count memory in ‘memblock.memory’ that doesn't have
> > > the node assigned.
> > >
> > > V2:https://lore.kernel.org/all/20230619075315.49114-1-zhiguangni01@gmail.com/
> > > V1:https://lore.kernel.org/all/20230615142016.419570-1-zhiguangni01@gmail.com/
> > >
> > > Signed-off-by: Liam Ni <zhiguangni01@gmail.com>
> > > ---
> > >  arch/loongarch/kernel/numa.c | 23 ++++++++---------------
> > >  arch/x86/mm/numa.c           | 26 +++++++-------------------
> > >  include/linux/mm.h           |  1 +
> > >  mm/mm_init.c                 | 20 ++++++++++++++++++++
> > >  4 files changed, 36 insertions(+), 34 deletions(-)
> > >
> > > diff --git a/arch/loongarch/kernel/numa.c b/arch/loongarch/kernel/numa.c
> > > index 708665895b47..0239891e4d19 100644
> > > --- a/arch/loongarch/kernel/numa.c
> > > +++ b/arch/loongarch/kernel/numa.c
> > > @@ -262,25 +262,18 @@ static void __init node_mem_init(unsigned int node)
> > >   * Sanity check to catch more bad NUMA configurations (they are amazingly
> > >   * common).  Make sure the nodes cover all memory.
> > >   */
> > > -static bool __init numa_meminfo_cover_memory(const struct numa_meminfo *mi)
> > > +static bool __init memblock_validate_numa_coverage(const u64 limit)
> >
> > There is no need to have arch specific memblock_validate_numa_coverage().
> > You can add this function to memblock and call it from NUMA initialization
> > instead of numa_meminfo_cover_memory().
> 
> Remove implementation of numa_meminfo_cover_memory function?
 
Yes, that's the idea.

> > The memblock_validate_numa_coverage() will count all the pages without node
> > ID set and compare to the threshold provided by the architectures.
> >
> > >  {
> > > -       int i;
> > > -       u64 numaram, biosram;
> > > +       u64 lo_pg;
> > >
> > > -       numaram = 0;
> > > -       for (i = 0; i < mi->nr_blks; i++) {
> > > -               u64 s = mi->blk[i].start >> PAGE_SHIFT;
> > > -               u64 e = mi->blk[i].end >> PAGE_SHIFT;
> > > +       lo_pg = max_pfn - calculate_without_node_pages_in_range();
> > >
> > > -               numaram += e - s;
> > > -               numaram -= __absent_pages_in_range(mi->blk[i].nid, s, e);
> > > -               if ((s64)numaram < 0)
> > > -                       numaram = 0;
> > > +       /* We seem to lose 3 pages somewhere. Allow 1M of slack. */
> > > +       if (lo_pg >= limit) {
> > > +               pr_err("NUMA: We lost 1m size page.\n");
> > > +               return false;
> > >         }
> > > -       max_pfn = max_low_pfn;
> > > -       biosram = max_pfn - absent_pages_in_range(0, max_pfn);
> > >
> > > -       BUG_ON((s64)(biosram - numaram) >= (1 << (20 - PAGE_SHIFT)));
> > >         return true;
> > >  }
> > >
> > > @@ -428,7 +421,7 @@ int __init init_numa_memory(void)
> > >                 return -EINVAL;
> > >
> > >         init_node_memblock();
> > > -       if (numa_meminfo_cover_memory(&numa_meminfo) == false)
> > > +       if (memblock_validate_numa_coverage(SZ_1M) == false)
> > >                 return -EINVAL;
> > >
> > >         for_each_node_mask(node, node_possible_map) {
> > > diff --git a/arch/x86/mm/numa.c b/arch/x86/mm/numa.c
> > > index 2aadb2019b4f..14feec144675 100644
> > > --- a/arch/x86/mm/numa.c
> > > +++ b/arch/x86/mm/numa.c
> > > @@ -451,30 +451,18 @@ EXPORT_SYMBOL(__node_distance);
> > >   * Sanity check to catch more bad NUMA configurations (they are amazingly
> > >   * common).  Make sure the nodes cover all memory.
> > >   */
> > > -static bool __init numa_meminfo_cover_memory(const struct numa_meminfo *mi)
> > > +static bool __init memblock_validate_numa_coverage(const u64 limit)
> > >  {
> > > -       u64 numaram, e820ram;
> > > -       int i;
> > > +       u64 lo_pg;
> > >
> > > -       numaram = 0;
> > > -       for (i = 0; i < mi->nr_blks; i++) {
> > > -               u64 s = mi->blk[i].start >> PAGE_SHIFT;
> > > -               u64 e = mi->blk[i].end >> PAGE_SHIFT;
> > > -               numaram += e - s;
> > > -               numaram -= __absent_pages_in_range(mi->blk[i].nid, s, e);
> > > -               if ((s64)numaram < 0)
> > > -                       numaram = 0;
> > > -       }
> > > -
> > > -       e820ram = max_pfn - absent_pages_in_range(0, max_pfn);
> > > +       lo_pg = max_pfn - calculate_without_node_pages_in_range();
> > >
> > >         /* We seem to lose 3 pages somewhere. Allow 1M of slack. */
> > > -       if ((s64)(e820ram - numaram) >= (1 << (20 - PAGE_SHIFT))) {
> > > -               printk(KERN_ERR "NUMA: nodes only cover %LuMB of your
> > > %LuMB e820 RAM. Not used.\n",
> > > -                      (numaram << PAGE_SHIFT) >> 20,
> > > -                      (e820ram << PAGE_SHIFT) >> 20);
> > > +       if (lo_pg >= limit) {
> > > +               pr_err("NUMA: We lost 1m size page.\n");
> > >                 return false;
> > >         }
> > > +
> > >         return true;
> > >  }
> > >
> > > @@ -583,7 +571,7 @@ static int __init numa_register_memblks(struct
> > > numa_meminfo *mi)
> > >                         return -EINVAL;
> > >                 }
> > >         }
> > > -       if (!numa_meminfo_cover_memory(mi))
> > > +       if (!memblock_validate_numa_coverage(SZ_1M))
> > >                 return -EINVAL;
> > >
> > >         /* Finally register nodes. */
> > > diff --git a/include/linux/mm.h b/include/linux/mm.h
> > > index 0daef3f2f029..b32457ad1ae3 100644
> > > --- a/include/linux/mm.h
> > > +++ b/include/linux/mm.h
> > > @@ -3043,6 +3043,7 @@ unsigned long __absent_pages_in_range(int nid,
> > > unsigned long start_pfn,
> > >                                                 unsigned long end_pfn);
> > >  extern unsigned long absent_pages_in_range(unsigned long start_pfn,
> > >                                                 unsigned long end_pfn);
> > > +extern unsigned long calculate_without_node_pages_in_range(void);
> > >  extern void get_pfn_range_for_nid(unsigned int nid,
> > >                         unsigned long *start_pfn, unsigned long *end_pfn);
> > >
> > > diff --git a/mm/mm_init.c b/mm/mm_init.c
> > > index 3ddd18a89b66..13a4883787e3 100644
> > > --- a/mm/mm_init.c
> > > +++ b/mm/mm_init.c
> > > @@ -1132,6 +1132,26 @@ static void __init
> > > adjust_zone_range_for_zone_movable(int nid,
> > >         }
> > >  }
> > >
> > > +/**
> > > + * @start_pfn: The start PFN to start searching for holes
> > > + * @end_pfn: The end PFN to stop searching for holes
> > > + *
> > > + * Return: Return the number of page frames without node assigned
> > > within a range.
> > > + */
> > > +unsigned long __init calculate_without_node_pages_in_range(void)
> > > +{
> > > +       unsigned long num_pages;
> > > +       unsigned long start_pfn, end_pfn;
> > > +       int nid, i;
> > > +
> > > +       for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn, &nid) {
> > > +               if (nid == NUMA_NO_NODE)
> > > +                       num_pages += end_pfn - start_pfn;
> > > +       }
> > > +
> > > +       return num_pages;
> > > +}
> > > +
> > >  /*
> > >   * Return the number of holes in a range on a node. If nid is MAX_NUMNODES,
> > >   * then all holes in the requested range will be accounted for.
> > > --
> > > 2.25.1
> >
> > --
> > Sincerely yours,
> > Mike.

-- 
Sincerely yours,
Mike.