[v6,00/12] mm: Sub-section memory hotplug support
mbox series

Message ID 155552633539.2015392.2477781120122237934.stgit@dwillia2-desk3.amr.corp.intel.com
Headers show
Series
  • mm: Sub-section memory hotplug support
Related show

Message

Dan Williams April 17, 2019, 6:38 p.m. UTC
Changes since v5 [1]:

- Rebase on next-20190416 and the new 'struct mhp_restrictions'
  infrastructure.

- Extend mhp_restrictions to the 'remove' case so the sub-section policy
  can be clarified with respect to the memblock-api in a symmetric
  manner with the 'add' case.

- Kill is_dev_zone() since cleanups have now made it moot

[1]: https://lwn.net/Articles/783808/

---

The memory hotplug section is an arbitrary / convenient unit for memory
hotplug. 'Section-size' units have bled into the user interface
('memblock' sysfs) and can not be changed without breaking existing
userspace. The section-size constraint, while mostly benign for typical
memory hotplug, has and continues to wreak havoc with 'device-memory'
use cases, persistent memory (pmem) in particular. Recall that pmem uses
devm_memremap_pages(), and subsequently arch_add_memory(), to allocate a
'struct page' memmap for pmem. However, it does not use the 'bottom
half' of memory hotplug, i.e. never marks pmem pages online and never
exposes the userspace memblock interface for pmem. This leaves an
opening to redress the section-size constraint.

To date, the libnvdimm subsystem has attempted to inject padding to
satisfy the internal constraints of arch_add_memory(). Beyond
complicating the code, leading to bugs [2], wasting memory, and limiting
configuration flexibility, the padding hack is broken when the platform
changes this physical memory alignment of pmem from one boot to the
next. Device failure (intermittent or permanent) and physical
reconfiguration are events that can cause the platform firmware to
change the physical placement of pmem on a subsequent boot, and device
failure is an everyday event in a data-center.

It turns out that sections are only a hard requirement of the
user-facing interface for memory hotplug and with a bit more
infrastructure sub-section arch_add_memory() support can be added for
kernel internal usages like devm_memremap_pages(). Here is an analysis
of the current design assumptions in the current code and how they are
addressed in the new implementation:

Current design assumptions:

- Sections that describe boot memory (early sections) are never
  unplugged / removed.

- pfn_valid(), in the CONFIG_SPARSEMEM_VMEMMAP=y, case devolves to a
  valid_section() check

- __add_pages() and helper routines assume all operations occur in
  PAGES_PER_SECTION units.

- The memblock sysfs interface only comprehends full sections

New design assumptions:

- Sections are instrumented with a sub-section bitmask to track (on x86)
  individual 2MB sub-divisions of a 128MB section.

- Partially populated early sections can be extended with additional
  sub-sections, and those sub-sections can be removed with
  arch_remove_memory(). With this in place we no longer lose usable memory
  capacity to padding.

- pfn_valid() is updated to look deeper than valid_section() to also check the
  active-sub-section mask. This indication is in the same cacheline as
  the valid_section() so the performance impact is expected to be
  negligible. So far the lkp robot has not reported any regressions.

- Outside of the core vmemmap population routines which are replaced,
  other helper routines like shrink_{zone,pgdat}_span() are updated to
  handle the smaller granularity. Core memory hotplug routines that deal
  with online memory are not touched.

- The existing memblock sysfs user api guarantees / assumptions are
  not touched since this capability is limited to !online
  !memblock-sysfs-accessible sections.

Meanwhile the issue reports continue to roll in from users that do not
understand when and how the 128MB constraint will bite them. The current
implementation relied on being able to support at least one misaligned
namespace, but that immediately falls over on any moderately complex
namespace creation attempt. Beyond the initial problem of 'System RAM'
colliding with pmem, and the unsolvable problem of physical alignment
changes, Linux is now being exposed to platforms that collide pmem
ranges with other pmem ranges by default [3]. In short,
devm_memremap_pages() has pushed the venerable section-size constraint
past the breaking point, and the simplicity of section-aligned
arch_add_memory() is no longer tenable.

These patches are exposed to the kbuild robot on my libnvdimm-pending
branch [4], and a preview of the unit test for this functionality is
available on the 'subsection-pending' branch of ndctl [5].

[2]: https://lore.kernel.org/r/155000671719.348031.2347363160141119237.stgit@dwillia2-desk3.amr.corp.intel.com
[3]: https://github.com/pmem/ndctl/issues/76
[4]: https://git.kernel.org/pub/scm/linux/kernel/git/djbw/nvdimm.git/log/?h=libnvdimm-pending
[5]: https://github.com/pmem/ndctl/commit/7c59b4867e1c

---

Dan Williams (12):
      mm/sparsemem: Introduce struct mem_section_usage
      mm/sparsemem: Introduce common definitions for the size and mask of a section
      mm/sparsemem: Add helpers track active portions of a section at boot
      mm/hotplug: Prepare shrink_{zone,pgdat}_span for sub-section removal
      mm/sparsemem: Convert kmalloc_section_memmap() to populate_section_memmap()
      mm/hotplug: Add mem-hotplug restrictions for remove_memory()
      mm: Kill is_dev_zone() helper
      mm/sparsemem: Prepare for sub-section ranges
      mm/sparsemem: Support sub-section hotplug
      mm/devm_memremap_pages: Enable sub-section remap
      libnvdimm/pfn: Fix fsdax-mode namespace info-block zero-fields
      libnvdimm/pfn: Stop padding pmem namespaces to section alignment


 arch/ia64/mm/init.c            |    4 
 arch/powerpc/mm/mem.c          |    5 -
 arch/s390/mm/init.c            |    2 
 arch/sh/mm/init.c              |    4 
 arch/x86/mm/init_32.c          |    4 
 arch/x86/mm/init_64.c          |    9 +
 drivers/nvdimm/dax_devs.c      |    2 
 drivers/nvdimm/pfn.h           |   12 -
 drivers/nvdimm/pfn_devs.c      |   93 +++-------
 include/linux/memory_hotplug.h |   12 +
 include/linux/mm.h             |    4 
 include/linux/mmzone.h         |   72 ++++++--
 kernel/memremap.c              |   70 +++-----
 mm/hmm.c                       |    2 
 mm/memory_hotplug.c            |  148 +++++++++-------
 mm/page_alloc.c                |    8 +
 mm/sparse-vmemmap.c            |   21 ++
 mm/sparse.c                    |  371 +++++++++++++++++++++++++++-------------
 18 files changed, 503 insertions(+), 340 deletions(-)

Comments

Andrew Morton April 17, 2019, 10:03 p.m. UTC | #1
On Wed, 17 Apr 2019 11:38:55 -0700 Dan Williams <dan.j.williams@intel.com> wrote:

> The memory hotplug section is an arbitrary / convenient unit for memory
> hotplug. 'Section-size' units have bled into the user interface
> ('memblock' sysfs) and can not be changed without breaking existing
> userspace. The section-size constraint, while mostly benign for typical
> memory hotplug, has and continues to wreak havoc with 'device-memory'
> use cases, persistent memory (pmem) in particular. Recall that pmem uses
> devm_memremap_pages(), and subsequently arch_add_memory(), to allocate a
> 'struct page' memmap for pmem. However, it does not use the 'bottom
> half' of memory hotplug, i.e. never marks pmem pages online and never
> exposes the userspace memblock interface for pmem. This leaves an
> opening to redress the section-size constraint.

v6 and we're not showing any review activity.  Who would be suitable
people to help out here?
Dan Williams April 17, 2019, 10:59 p.m. UTC | #2
On Wed, Apr 17, 2019 at 3:04 PM Andrew Morton <akpm@linux-foundation.org> wrote:
>
> On Wed, 17 Apr 2019 11:38:55 -0700 Dan Williams <dan.j.williams@intel.com> wrote:
>
> > The memory hotplug section is an arbitrary / convenient unit for memory
> > hotplug. 'Section-size' units have bled into the user interface
> > ('memblock' sysfs) and can not be changed without breaking existing
> > userspace. The section-size constraint, while mostly benign for typical
> > memory hotplug, has and continues to wreak havoc with 'device-memory'
> > use cases, persistent memory (pmem) in particular. Recall that pmem uses
> > devm_memremap_pages(), and subsequently arch_add_memory(), to allocate a
> > 'struct page' memmap for pmem. However, it does not use the 'bottom
> > half' of memory hotplug, i.e. never marks pmem pages online and never
> > exposes the userspace memblock interface for pmem. This leaves an
> > opening to redress the section-size constraint.
>
> v6 and we're not showing any review activity.  Who would be suitable
> people to help out here?

There was quite a bit of review of the cover letter from Michal and
David, but you're right the details not so much as of yet. I'd like to
call out other people where I can reciprocate with some review of my
own. Oscar's altmap work looks like a good candidate for that.
Dan Williams April 18, 2019, 2:09 a.m. UTC | #3
On Wed, Apr 17, 2019 at 3:59 PM Dan Williams <dan.j.williams@intel.com> wrote:
>
> On Wed, Apr 17, 2019 at 3:04 PM Andrew Morton <akpm@linux-foundation.org> wrote:
> >
> > On Wed, 17 Apr 2019 11:38:55 -0700 Dan Williams <dan.j.williams@intel.com> wrote:
> >
> > > The memory hotplug section is an arbitrary / convenient unit for memory
> > > hotplug. 'Section-size' units have bled into the user interface
> > > ('memblock' sysfs) and can not be changed without breaking existing
> > > userspace. The section-size constraint, while mostly benign for typical
> > > memory hotplug, has and continues to wreak havoc with 'device-memory'
> > > use cases, persistent memory (pmem) in particular. Recall that pmem uses
> > > devm_memremap_pages(), and subsequently arch_add_memory(), to allocate a
> > > 'struct page' memmap for pmem. However, it does not use the 'bottom
> > > half' of memory hotplug, i.e. never marks pmem pages online and never
> > > exposes the userspace memblock interface for pmem. This leaves an
> > > opening to redress the section-size constraint.
> >
> > v6 and we're not showing any review activity.  Who would be suitable
> > people to help out here?
>
> There was quite a bit of review of the cover letter from Michal and
> David, but you're right the details not so much as of yet. I'd like to
> call out other people where I can reciprocate with some review of my
> own. Oscar's altmap work looks like a good candidate for that.

I'm also hoping Jeff can give a tested-by for the customer scenarios
that fall over with the current implementation.
Jeff Moyer April 18, 2019, 12:45 p.m. UTC | #4
Dan Williams <dan.j.williams@intel.com> writes:

>> On Wed, Apr 17, 2019 at 3:59 PM Dan Williams <dan.j.williams@intel.com> wrote:
>>
>> On Wed, Apr 17, 2019 at 3:04 PM Andrew Morton <akpm@linux-foundation.org> wrote:
>> >
>> > On Wed, 17 Apr 2019 11:38:55 -0700 Dan Williams <dan.j.williams@intel.com> wrote:
>> >
>> > > The memory hotplug section is an arbitrary / convenient unit for memory
>> > > hotplug. 'Section-size' units have bled into the user interface
>> > > ('memblock' sysfs) and can not be changed without breaking existing
>> > > userspace. The section-size constraint, while mostly benign for typical
>> > > memory hotplug, has and continues to wreak havoc with 'device-memory'
>> > > use cases, persistent memory (pmem) in particular. Recall that pmem uses
>> > > devm_memremap_pages(), and subsequently arch_add_memory(), to allocate a
>> > > 'struct page' memmap for pmem. However, it does not use the 'bottom
>> > > half' of memory hotplug, i.e. never marks pmem pages online and never
>> > > exposes the userspace memblock interface for pmem. This leaves an
>> > > opening to redress the section-size constraint.
>> >
>> > v6 and we're not showing any review activity.  Who would be suitable
>> > people to help out here?
>>
>> There was quite a bit of review of the cover letter from Michal and
>> David, but you're right the details not so much as of yet. I'd like to
>> call out other people where I can reciprocate with some review of my
>> own. Oscar's altmap work looks like a good candidate for that.
>
> I'm also hoping Jeff can give a tested-by for the customer scenarios
> that fall over with the current implementation.

Sure.  I'll also have a look over the patches.

-Jeff
Dan Williams April 19, 2019, 3:25 a.m. UTC | #5
On Thu, Apr 18, 2019 at 5:45 AM Jeff Moyer <jmoyer@redhat.com> wrote:
[..]
> >> > v6 and we're not showing any review activity.  Who would be suitable
> >> > people to help out here?
> >>
> >> There was quite a bit of review of the cover letter from Michal and
> >> David, but you're right the details not so much as of yet. I'd like to
> >> call out other people where I can reciprocate with some review of my
> >> own. Oscar's altmap work looks like a good candidate for that.
> >
> > I'm also hoping Jeff can give a tested-by for the customer scenarios
> > that fall over with the current implementation.
>
> Sure.  I'll also have a look over the patches.

Andrew, heads up it looks like there is a memory corruption bug in
these patches as I've gotten a few reported of "bad page state" at
boot. Please drop until I can track down the failure.
Oscar Salvador April 23, 2019, 1:16 p.m. UTC | #6
On Wed, 2019-04-17 at 15:59 -0700, Dan Williams wrote:
> On Wed, Apr 17, 2019 at 3:04 PM Andrew Morton <akpm@linux-foundation.
> org> wrote:
> > 
> > On Wed, 17 Apr 2019 11:38:55 -0700 Dan Williams <dan.j.williams@int
> > el.com> wrote:
> > 
> > > The memory hotplug section is an arbitrary / convenient unit for
> > > memory
> > > hotplug. 'Section-size' units have bled into the user interface
> > > ('memblock' sysfs) and can not be changed without breaking
> > > existing
> > > userspace. The section-size constraint, while mostly benign for
> > > typical
> > > memory hotplug, has and continues to wreak havoc with 'device-
> > > memory'
> > > use cases, persistent memory (pmem) in particular. Recall that
> > > pmem uses
> > > devm_memremap_pages(), and subsequently arch_add_memory(), to
> > > allocate a
> > > 'struct page' memmap for pmem. However, it does not use the
> > > 'bottom
> > > half' of memory hotplug, i.e. never marks pmem pages online and
> > > never
> > > exposes the userspace memblock interface for pmem. This leaves an
> > > opening to redress the section-size constraint.
> > 
> > v6 and we're not showing any review activity.  Who would be
> > suitable
> > people to help out here?
> 
> There was quite a bit of review of the cover letter from Michal and
> David, but you're right the details not so much as of yet. I'd like
> to
> call out other people where I can reciprocate with some review of my
> own. Oscar's altmap work looks like a good candidate for that.

Thanks Dan for ccing me.
I will take a look at the patches soon.
Pavel Tatashin April 24, 2019, 8:43 p.m. UTC | #7
I am also taking a look at this work now. I will review and test it in
the next couple of days.

Pasha

On Tue, Apr 23, 2019 at 9:17 AM Oscar Salvador <osalvador@suse.de> wrote:
>
> On Wed, 2019-04-17 at 15:59 -0700, Dan Williams wrote:
> > On Wed, Apr 17, 2019 at 3:04 PM Andrew Morton <akpm@linux-foundation.
> > org> wrote:
> > >
> > > On Wed, 17 Apr 2019 11:38:55 -0700 Dan Williams <dan.j.williams@int
> > > el.com> wrote:
> > >
> > > > The memory hotplug section is an arbitrary / convenient unit for
> > > > memory
> > > > hotplug. 'Section-size' units have bled into the user interface
> > > > ('memblock' sysfs) and can not be changed without breaking
> > > > existing
> > > > userspace. The section-size constraint, while mostly benign for
> > > > typical
> > > > memory hotplug, has and continues to wreak havoc with 'device-
> > > > memory'
> > > > use cases, persistent memory (pmem) in particular. Recall that
> > > > pmem uses
> > > > devm_memremap_pages(), and subsequently arch_add_memory(), to
> > > > allocate a
> > > > 'struct page' memmap for pmem. However, it does not use the
> > > > 'bottom
> > > > half' of memory hotplug, i.e. never marks pmem pages online and
> > > > never
> > > > exposes the userspace memblock interface for pmem. This leaves an
> > > > opening to redress the section-size constraint.
> > >
> > > v6 and we're not showing any review activity.  Who would be
> > > suitable
> > > people to help out here?
> >
> > There was quite a bit of review of the cover letter from Michal and
> > David, but you're right the details not so much as of yet. I'd like
> > to
> > call out other people where I can reciprocate with some review of my
> > own. Oscar's altmap work looks like a good candidate for that.
>
> Thanks Dan for ccing me.
> I will take a look at the patches soon.
>
> --
> Oscar Salvador
> SUSE L3
Pavel Tatashin May 2, 2019, 10:46 p.m. UTC | #8
Hi Dan,

How do you test these patches? Do you have any instructions?

I see for example that check_hotplug_memory_range() still enforces
memory_block_size_bytes() alignment.

Also, after removing check_hotplug_memory_range(), I tried to online
16M aligned DAX memory, and got the following panic:

# echo online > /sys/devices/system/memory/memory7/state
[  202.193132] WARNING: CPU: 2 PID: 351 at drivers/base/memory.c:207
memory_block_action+0x110/0x178
[  202.193391] Modules linked in:
[  202.193698] CPU: 2 PID: 351 Comm: sh Not tainted
5.1.0-rc7_pt_devdax-00038-g865af4385544-dirty #9
[  202.193909] Hardware name: linux,dummy-virt (DT)
[  202.194122] pstate: 60000005 (nZCv daif -PAN -UAO)
[  202.194243] pc : memory_block_action+0x110/0x178
[  202.194404] lr : memory_block_action+0x90/0x178
[  202.194506] sp : ffff000016763ca0
[  202.194592] x29: ffff000016763ca0 x28: ffff80016fd29b80
[  202.194724] x27: 0000000000000000 x26: 0000000000000000
[  202.194838] x25: ffff000015546000 x24: 00000000001c0000
[  202.194949] x23: 0000000000000000 x22: 0000000000040000
[  202.195058] x21: 00000000001c0000 x20: 0000000000000008
[  202.195168] x19: 0000000000000007 x18: 0000000000000000
[  202.195281] x17: 0000000000000000 x16: 0000000000000000
[  202.195393] x15: 0000000000000000 x14: 0000000000000000
[  202.195505] x13: 0000000000000000 x12: 0000000000000000
[  202.195614] x11: 0000000000000000 x10: 0000000000000000
[  202.195744] x9 : 0000000000000000 x8 : 0000000180000000
[  202.195858] x7 : 0000000000000018 x6 : ffff000015541930
[  202.195966] x5 : ffff000015541930 x4 : 0000000000000001
[  202.196074] x3 : 0000000000000001 x2 : 0000000000000000
[  202.196185] x1 : 0000000000000070 x0 : 0000000000000000
[  202.196366] Call trace:
[  202.196455]  memory_block_action+0x110/0x178
[  202.196589]  memory_subsys_online+0x3c/0x80
[  202.196681]  device_online+0x6c/0x90
[  202.196761]  state_store+0x84/0x100
[  202.196841]  dev_attr_store+0x18/0x28
[  202.196927]  sysfs_kf_write+0x40/0x58
[  202.197010]  kernfs_fop_write+0xcc/0x1d8
[  202.197099]  __vfs_write+0x18/0x40
[  202.197187]  vfs_write+0xa4/0x1b0
[  202.197295]  ksys_write+0x64/0xd8
[  202.197430]  __arm64_sys_write+0x18/0x20
[  202.197521]  el0_svc_common.constprop.0+0x7c/0xe8
[  202.197621]  el0_svc_handler+0x28/0x78
[  202.197706]  el0_svc+0x8/0xc
[  202.197828] ---[ end trace 57719823dda6d21e ]---

Thank you,
Pasha
Dan Williams May 2, 2019, 11:20 p.m. UTC | #9
On Thu, May 2, 2019 at 3:46 PM Pavel Tatashin <pasha.tatashin@soleen.com> wrote:
>
> Hi Dan,
>
> How do you test these patches? Do you have any instructions?

Yes, I briefly mentioned this in the cover letter, but here is the
test I am using:

>
> I see for example that check_hotplug_memory_range() still enforces
> memory_block_size_bytes() alignment.
>
> Also, after removing check_hotplug_memory_range(), I tried to online
> 16M aligned DAX memory, and got the following panic:

Right, this functionality is currently strictly limited to the
devm_memremap_pages() case where there are guarantees that the memory
will never be onlined. This is due to the fact that the section size
is entangled with the memblock api. That said I would have expected
you to trigger the warning in subsection_check() before getting this
far into the hotplug process.
>
> # echo online > /sys/devices/system/memory/memory7/state
> [  202.193132] WARNING: CPU: 2 PID: 351 at drivers/base/memory.c:207
> memory_block_action+0x110/0x178
> [  202.193391] Modules linked in:
> [  202.193698] CPU: 2 PID: 351 Comm: sh Not tainted
> 5.1.0-rc7_pt_devdax-00038-g865af4385544-dirty #9
> [  202.193909] Hardware name: linux,dummy-virt (DT)
> [  202.194122] pstate: 60000005 (nZCv daif -PAN -UAO)
> [  202.194243] pc : memory_block_action+0x110/0x178
> [  202.194404] lr : memory_block_action+0x90/0x178
> [  202.194506] sp : ffff000016763ca0
> [  202.194592] x29: ffff000016763ca0 x28: ffff80016fd29b80
> [  202.194724] x27: 0000000000000000 x26: 0000000000000000
> [  202.194838] x25: ffff000015546000 x24: 00000000001c0000
> [  202.194949] x23: 0000000000000000 x22: 0000000000040000
> [  202.195058] x21: 00000000001c0000 x20: 0000000000000008
> [  202.195168] x19: 0000000000000007 x18: 0000000000000000
> [  202.195281] x17: 0000000000000000 x16: 0000000000000000
> [  202.195393] x15: 0000000000000000 x14: 0000000000000000
> [  202.195505] x13: 0000000000000000 x12: 0000000000000000
> [  202.195614] x11: 0000000000000000 x10: 0000000000000000
> [  202.195744] x9 : 0000000000000000 x8 : 0000000180000000
> [  202.195858] x7 : 0000000000000018 x6 : ffff000015541930
> [  202.195966] x5 : ffff000015541930 x4 : 0000000000000001
> [  202.196074] x3 : 0000000000000001 x2 : 0000000000000000
> [  202.196185] x1 : 0000000000000070 x0 : 0000000000000000
> [  202.196366] Call trace:
> [  202.196455]  memory_block_action+0x110/0x178
> [  202.196589]  memory_subsys_online+0x3c/0x80
> [  202.196681]  device_online+0x6c/0x90
> [  202.196761]  state_store+0x84/0x100
> [  202.196841]  dev_attr_store+0x18/0x28
> [  202.196927]  sysfs_kf_write+0x40/0x58
> [  202.197010]  kernfs_fop_write+0xcc/0x1d8
> [  202.197099]  __vfs_write+0x18/0x40
> [  202.197187]  vfs_write+0xa4/0x1b0
> [  202.197295]  ksys_write+0x64/0xd8
> [  202.197430]  __arm64_sys_write+0x18/0x20
> [  202.197521]  el0_svc_common.constprop.0+0x7c/0xe8
> [  202.197621]  el0_svc_handler+0x28/0x78
> [  202.197706]  el0_svc+0x8/0xc
> [  202.197828] ---[ end trace 57719823dda6d21e ]---
>
> Thank you,
> Pasha
Dan Williams May 2, 2019, 11:21 p.m. UTC | #10
On Thu, May 2, 2019 at 4:20 PM Dan Williams <dan.j.williams@intel.com> wrote:
>
> On Thu, May 2, 2019 at 3:46 PM Pavel Tatashin <pasha.tatashin@soleen.com> wrote:
> >
> > Hi Dan,
> >
> > How do you test these patches? Do you have any instructions?
>
> Yes, I briefly mentioned this in the cover letter, but here is the
> test I am using:

Sorry, fumble fingered the 'send' button, here is that link:

https://github.com/pmem/ndctl/blob/subsection-pending/test/sub-section.sh
Oscar Salvador May 3, 2019, 10:48 a.m. UTC | #11
On Thu, May 02, 2019 at 04:20:03PM -0700, Dan Williams wrote:
> On Thu, May 2, 2019 at 3:46 PM Pavel Tatashin <pasha.tatashin@soleen.com> wrote:
> >
> > Hi Dan,
> >
> > How do you test these patches? Do you have any instructions?
> 
> Yes, I briefly mentioned this in the cover letter, but here is the
> test I am using:
> 
> >
> > I see for example that check_hotplug_memory_range() still enforces
> > memory_block_size_bytes() alignment.
> >
> > Also, after removing check_hotplug_memory_range(), I tried to online
> > 16M aligned DAX memory, and got the following panic:
> 
> Right, this functionality is currently strictly limited to the
> devm_memremap_pages() case where there are guarantees that the memory
> will never be onlined. This is due to the fact that the section size
> is entangled with the memblock api. That said I would have expected
> you to trigger the warning in subsection_check() before getting this
> far into the hotplug process.
> >
> > # echo online > /sys/devices/system/memory/memory7/state
> > [  202.193132] WARNING: CPU: 2 PID: 351 at drivers/base/memory.c:207
> > memory_block_action+0x110/0x178
> > [  202.193391] Modules linked in:
> > [  202.193698] CPU: 2 PID: 351 Comm: sh Not tainted
> > 5.1.0-rc7_pt_devdax-00038-g865af4385544-dirty #9
> > [  202.193909] Hardware name: linux,dummy-virt (DT)
> > [  202.194122] pstate: 60000005 (nZCv daif -PAN -UAO)
> > [  202.194243] pc : memory_block_action+0x110/0x178
> > [  202.194404] lr : memory_block_action+0x90/0x178
> > [  202.194506] sp : ffff000016763ca0
> > [  202.194592] x29: ffff000016763ca0 x28: ffff80016fd29b80
> > [  202.194724] x27: 0000000000000000 x26: 0000000000000000
> > [  202.194838] x25: ffff000015546000 x24: 00000000001c0000
> > [  202.194949] x23: 0000000000000000 x22: 0000000000040000
> > [  202.195058] x21: 00000000001c0000 x20: 0000000000000008
> > [  202.195168] x19: 0000000000000007 x18: 0000000000000000
> > [  202.195281] x17: 0000000000000000 x16: 0000000000000000
> > [  202.195393] x15: 0000000000000000 x14: 0000000000000000
> > [  202.195505] x13: 0000000000000000 x12: 0000000000000000
> > [  202.195614] x11: 0000000000000000 x10: 0000000000000000
> > [  202.195744] x9 : 0000000000000000 x8 : 0000000180000000
> > [  202.195858] x7 : 0000000000000018 x6 : ffff000015541930
> > [  202.195966] x5 : ffff000015541930 x4 : 0000000000000001
> > [  202.196074] x3 : 0000000000000001 x2 : 0000000000000000
> > [  202.196185] x1 : 0000000000000070 x0 : 0000000000000000
> > [  202.196366] Call trace:
> > [  202.196455]  memory_block_action+0x110/0x178
> > [  202.196589]  memory_subsys_online+0x3c/0x80
> > [  202.196681]  device_online+0x6c/0x90
> > [  202.196761]  state_store+0x84/0x100
> > [  202.196841]  dev_attr_store+0x18/0x28
> > [  202.196927]  sysfs_kf_write+0x40/0x58
> > [  202.197010]  kernfs_fop_write+0xcc/0x1d8
> > [  202.197099]  __vfs_write+0x18/0x40
> > [  202.197187]  vfs_write+0xa4/0x1b0
> > [  202.197295]  ksys_write+0x64/0xd8
> > [  202.197430]  __arm64_sys_write+0x18/0x20
> > [  202.197521]  el0_svc_common.constprop.0+0x7c/0xe8
> > [  202.197621]  el0_svc_handler+0x28/0x78
> > [  202.197706]  el0_svc+0x8/0xc
> > [  202.197828] ---[ end trace 57719823dda6d21e ]---

This warning relates to:

        for (; section_nr < section_nr_end; section_nr++) {
                if (WARN_ON_ONCE(!pfn_valid(pfn)))
                        return false;

from pages_correctly_probed().
AFAICS, this is orthogonal to subsection_check().