Re: [PATCH] mm/memory-failure: Disable soft offline for HugeTLB pages by default

[Kyle Meyer <kyle.meyer@hpe.com>]

On Fri, Sep 12, 2025 at 09:53:02AM +0200, David Hildenbrand wrote:
> On 11.09.25 19:56, Luck, Tony wrote:
> > On Thu, Sep 11, 2025 at 10:46:10AM +0200, David Hildenbrand wrote:
> > > On 10.09.25 18:15, Kyle Meyer wrote:
> > > > Soft offlining a HugeTLB page reduces the available HugeTLB page pool.
> > > > Since HugeTLB pages are preallocated, reducing the available HugeTLB
> > > > page pool can cause allocation failures.
> > > > 
> > > > /proc/sys/vm/enable_soft_offline provides a sysctl interface to
> > > > disable/enable soft offline:
> > > > 
> > > > 0 - Soft offline is disabled.
> > > > 1 - Soft offline is enabled.
> > > > 
> > > > The current sysctl interface does not distinguish between HugeTLB pages
> > > > and other page types.
> > > > 
> > > > Disable soft offline for HugeTLB pages by default (1) and extend the
> > > > sysctl interface to preserve existing behavior (2):
> > > > 
> > > > 0 - Soft offline is disabled.
> > > > 1 - Soft offline is enabled (excluding HugeTLB pages).
> > > > 2 - Soft offline is enabled (including HugeTLB pages).
> > > > 
> > > > Update documentation for the sysctl interface, reference the sysctl
> > > > interface in the sysfs ABI documentation, and update HugeTLB soft
> > > > offline selftests.
> > > 
> > > I'm sure you spotted that the documentation for
> > > "/sys/devices/system/memory/soft_offline_pag" resides under "testing".
> > 
> > But that is only one of several places in the kernel that
> > feed into the page offline code.
> 
> Right, I can see one more call to soft_offline_page() from
> arch/parisc/kernel/pdt.c.
> 
> And there is memory_failure_work_func() that I missed.
> 
> So agreed that this goes beyond testing.
> 
> It caught my attention because you ended up modifying documentation residing
> in Documentation/ABI/testing/sysfs-memory-page-offline.
> 
> Reading 56374430c5dfc that Kyle pointed out is gets clearer.
> 
> So the patch motivation/idea makes sense to me.
> 
> 
> I'll note two things:
> 
> (1) The interface design is not really extensible. Imagine if we want to
> exclude yet another page type.
> 
> Can we maybe add a second interface that defines a filter for types?
> 
> Alternatively, you could use all the remaining flags as such a filter.
> 
> 0 - Soft offline is completely disabled.
> 1 - Soft offline is enabled except for manually disabled types.
> 
> Filter
> 
> 2 - disable hugetlb.
> 
> So value 3 would give you "enable all except hugetlb" etc.
> 
> We could add in the future
> 
> 4 - disable guest_memfd (just some random example)
> 
> 
> Then you
> 
> 2) Changing the semantics of the value "1"
> 
> IIUC, you are changing the semantics of value "1". It used to mean
> "SOFT_OFFLINE_ENABLED" now it is "SOFT_OFFLINE_ENABLED_SKIP_HUGETLB", which
> is a change in behavior.
> 
> If that is the case, I don't think that's okay.
> 
> 
> 2) I am not sure about changing the default. That should be an admin/
>    distro decision.

Thank you, that sounds good to me. I'll put something together.

Thanks,
Kyle Meyer