Re: [PATCH v2 1/4] memblock tests: add simulation of physical memory with multiple NUMA nodes

[Rebecca Mckeever]

On Wed, Aug 31, 2022 at 06:12:10PM +0300, Mike Rapoport wrote:
> On Tue, Aug 30, 2022 at 10:49:09PM -0500, Rebecca Mckeever wrote:
> > On Tue, Aug 30, 2022 at 01:17:56PM +0200, David Hildenbrand wrote:
> > > On 19.08.22 11:05, Rebecca Mckeever wrote:
> > > > Add functions setup_numa_memblock_generic() and setup_numa_memblock()
> > > > for setting up a memory layout with multiple NUMA nodes in a previously
> > > > allocated dummy physical memory. These functions can be used in place of
> > > > setup_memblock() in tests that need to simulate a NUMA system.
> > > > 
> > > > setup_numa_memblock_generic():
> > > > - allows for setting up a custom memory layout by specifying the amount
> > > >   of memory in each node, the number of nodes, and a factor that will be
> > > >   used to scale the memory in each node
> > > > 
> > > > setup_numa_memblock():
> > > > - allows for setting up a default memory layout
> > > > 
> > > > Introduce constant MEM_FACTOR, which is used to scale the default memory
> > > > layout based on MEM_SIZE.
> > > > 
> > > > Set CONFIG_NODES_SHIFT to 4 when building with NUMA=1 to allow for up to
> > > > 16 NUMA nodes.
> > > > 
> > > > Signed-off-by: Rebecca Mckeever <remckee0@xxxxxxxxx>
> > > > ---
> > > >  .../testing/memblock/scripts/Makefile.include |  2 +-
> > > >  tools/testing/memblock/tests/common.c         | 38 +++++++++++++++++++
> > > >  tools/testing/memblock/tests/common.h         |  9 ++++-
> > > >  3 files changed, 47 insertions(+), 2 deletions(-)
> > > > 
> > > > diff --git a/tools/testing/memblock/scripts/Makefile.include b/tools/testing/memblock/scripts/Makefile.include
> > > > index aa6d82d56a23..998281723590 100644
> > > > --- a/tools/testing/memblock/scripts/Makefile.include
> > > > +++ b/tools/testing/memblock/scripts/Makefile.include
> > > > @@ -3,7 +3,7 @@
> > > >  
> > > >  # Simulate CONFIG_NUMA=y
> > > >  ifeq ($(NUMA), 1)
> > > > -	CFLAGS += -D CONFIG_NUMA
> > > > +	CFLAGS += -D CONFIG_NUMA -D CONFIG_NODES_SHIFT=4
> > > >  endif
> > > >  
> > > >  # Use 32 bit physical addresses.
> > > > diff --git a/tools/testing/memblock/tests/common.c b/tools/testing/memblock/tests/common.c
> > > > index eec6901081af..15d8767dc70c 100644
> > > > --- a/tools/testing/memblock/tests/common.c
> > > > +++ b/tools/testing/memblock/tests/common.c
> > > > @@ -34,6 +34,10 @@ static const char * const help_opts[] = {
> > > >  
> > > >  static int verbose;
> > > >  
> > > > +static const phys_addr_t node_sizes[] = {
> > > > +	SZ_4K, SZ_1K, SZ_2K, SZ_2K, SZ_1K, SZ_1K, SZ_4K, SZ_1K
> > > > +};
> > > > +
> > > >  /* sets global variable returned by movable_node_is_enabled() stub */
> > > >  bool movable_node_enabled;
> > > >  
> > > > @@ -72,6 +76,40 @@ void setup_memblock(void)
> > > >  	fill_memblock();
> > > >  }
> > > >  
> > > > +/**
> > > > + * setup_numa_memblock_generic:
> > > > + * Set up a memory layout with multiple NUMA nodes in a previously allocated
> > > > + * dummy physical memory.
> > > > + * @nodes: an array containing the amount of memory in each node
> > > > + * @node_cnt: the size of @nodes
> > > > + * @factor: a factor that will be used to scale the memory in each node
> > > > + *
> > > > + * The nids will be set to 0 through node_cnt - 1.
> > > > + */
> > > > +void setup_numa_memblock_generic(const phys_addr_t nodes[],
> > > > +				 int node_cnt, int factor)
> > > > +{
> > > > +	phys_addr_t base;
> > > > +	int flags;
> > > > +
> > > > +	reset_memblock_regions();
> > > > +	base = (phys_addr_t)memory_block.base;
> > > > +	flags = (movable_node_is_enabled()) ? MEMBLOCK_NONE : MEMBLOCK_HOTPLUG;
> > > > +
> > > > +	for (int i = 0; i < node_cnt; i++) {
> > > > +		phys_addr_t size = factor * nodes[i];
> > > 
> > > I'm a bit lost why we need the factor if we already provide sizes in the
> > > array.
> > > 
> > > Can you enlighten me? :)
> > > 
> > > Why can't we just stick to the sizes in the array?
> > > 
> > Without the factor, some of the tests will break if we increase MEM_SIZE
> > in the future (which we may need to do). I could rewrite them so that the
> > factor is not needed, but I thought the code would be over-complicated if
> > I did.
> 
> What if we make nodes[] to represent the fraction of the memory rather than
> a node size? Then the factor won't be required.
> 
I think that will work. I'll try it.

> > Thanks,
> > Rebecca
> 
> -- 
> Sincerely yours,
> Mike.

Thanks,
Rebecca