On Thu, 21 Jan 2021 at 11:01, Christoph Lameter [off-list ref] wrote:

On Thu, 21 Jan 2021, Bharata B Rao wrote:

quoted

quoted

The problem is that calculate_order() is called a number of times
before secondaries CPUs are booted and it returns 1 instead of 224.
This makes the use of num_online_cpus() irrelevant for those cases

After adding in my command line "slub_min_objects=36" which equals to
4 * (fls(num_online_cpus()) + 1) with a correct num_online_cpus == 224
, the regression diseapears:

9 iterations of hackbench -l 16000 -g 16: 3.201sec (+/- 0.90%)

Should we have switched to num_present_cpus() rather than
num_online_cpus()? If so, the below patch should address the
above problem.

There is certainly an initcall after secondaries are booted where we could
redo the calculate_order?

Or the num_online_cpus needs to be up to date earlier. Why does this issue
not occur on x86? Does x86 have an up to date num_online_cpus earlier?

I have added a printk in calculate_order() :
        pr_info(" SLUB calculate_order cmd  %d min %d online %d
present %d possible %d cpus %d", slub_min_objects, min_objects,
num_online_cpus(), num_present_cpus(), num_possible_cpus(),
nr_cpu_ids);

And checked with
qemu-system-x86_64 -kernel bzImage -nographic -smp 4 -append "console=ttyS0"

[    0.064927]  SLUB calculate_order cmd  0 min 8 online 1 present 4
possible 4 cpus 4
[    0.064970] SLUB: HWalign=64, Order=0-3, MinObjects=0, CPUs=4, Nodes=1

The num_online_cpus has the same behavior as on my platform and
num_online_cpus =  1 when kmem_cache_init() is called

Only the num_present_cpus = 4 from the beginning but that's probably
just because it runs in a VM

Also it's interesting to notice that num_possible_cpus and nr_cpu_ids
are set to the correct value