Dan Williams dan.j.williams@intel.com writes:

Changes since v9 [1]:

• Fix multiple issues related to the fact that pfn_valid() has traditionally returned true for any pfn in an 'early' (onlined at boot) section regardless of whether that pfn represented 'System RAM'. Teach pfn_valid() to maintain its traditional behavior in the presence of subsections. Specifically, subsection precision for pfn_valid() is only considered for non-early / hot-plugged sections. (Qian)

• Related to the first item introduce a SECTION_IS_EARLY (->section_mem_map flag) to remove the existing hacks for determining an early section by looking at whether the usemap was allocated from the slab.

• Kill off the EEXIST hackery in __add_pages(). It breaks (arch_add_memory() false-positive) the detection of subsection collisions reported by section_activate(). It is also obviated by David's recent reworks to move the 'System RAM' request_region() earlier in the add_memory() sequence().

• Switch to an arch-independent / static subsection-size of 2MB. Otherwise, a per-arch subsection-size is a roadblock on the path to persistent memory namespace compatibility across archs. (Jeff)

• Update the changelog for "libnvdimm/pfn: Fix fsdax-mode namespace info-block zero-fields" to clarify that the "Cc: stable" is only there as safety measure for a distro that decides to backport "libnvdimm/pfn: Stop padding pmem namespaces to section alignment", otherwise there is no known bug exposure in older kernels. (Andrew)

• Drop some redundant subsection checks (Oscar)

• Collect some reviewed-bys

You can add Tested-by: Aneesh Kumar K.V aneesh.kumar@linux.ibm.com for ppc64.

BTW even after this series we have the kernel crash mentioned in the below email on reconfigure.

https://lore.kernel.org/linux-mm/20190514025354.9108-1-aneesh.kumar@linux.ib...

I guess we need to conclude how the reserve space struct page should be initialized ?

---

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 a subsection-v10 branch [4], and a preview of the unit test for this functionality is available on the 'subsection-pending' branch of ndctl [5].

---

Dan Williams (13): mm/sparsemem: Introduce struct mem_section_usage mm/sparsemem: Introduce a SECTION_IS_EARLY flag 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: Kill is_dev_zone() usage in __remove_pages() mm: Kill is_dev_zone() helper mm/sparsemem: Prepare for sub-section ranges mm/sparsemem: Support sub-section hotplug mm: Document ZONE_DEVICE memory-model implications 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

Documentation/vm/memory-model.rst | 39 ++++ arch/x86/mm/init_64.c | 4 drivers/nvdimm/dax_devs.c | 2 drivers/nvdimm/pfn.h | 15 -- drivers/nvdimm/pfn_devs.c | 95 +++------- include/linux/memory_hotplug.h | 7 - include/linux/mm.h | 4 include/linux/mmzone.h | 84 +++++++-- kernel/memremap.c | 61 +++---- mm/memory_hotplug.c | 173 +++++++++---------- mm/page_alloc.c | 16 +- mm/sparse-vmemmap.c | 21 ++ mm/sparse.c | 335 ++++++++++++++++++++++++------------- 13 files changed, 494 insertions(+), 362 deletions(-)

On Tue, Jun 18, 2019 at 10:51:33PM -0700, Dan Williams wrote:

Changes since v9 [1]:

• Fix multiple issues related to the fact that pfn_valid() has traditionally returned true for any pfn in an 'early' (onlined at boot) section regardless of whether that pfn represented 'System RAM'. Teach pfn_valid() to maintain its traditional behavior in the presence of subsections. Specifically, subsection precision for pfn_valid() is only considered for non-early / hot-plugged sections. (Qian)

• Related to the first item introduce a SECTION_IS_EARLY (->section_mem_map flag) to remove the existing hacks for determining an early section by looking at whether the usemap was allocated from the slab.

• Kill off the EEXIST hackery in __add_pages(). It breaks (arch_add_memory() false-positive) the detection of subsection collisions reported by section_activate(). It is also obviated by David's recent reworks to move the 'System RAM' request_region() earlier in the add_memory() sequence().

• Switch to an arch-independent / static subsection-size of 2MB. Otherwise, a per-arch subsection-size is a roadblock on the path to persistent memory namespace compatibility across archs. (Jeff)

• Update the changelog for "libnvdimm/pfn: Fix fsdax-mode namespace info-block zero-fields" to clarify that the "Cc: stable" is only there as safety measure for a distro that decides to backport "libnvdimm/pfn: Stop padding pmem namespaces to section alignment", otherwise there is no known bug exposure in older kernels. (Andrew)

• Drop some redundant subsection checks (Oscar)

• Collect some reviewed-bys

Hi Dan,

I am planning to give it a final review later tomorrow. Now that this work is settled, I took the chance to dust off and push my vmemmap-hotplug, and I am working on that right now. But I would definetely come back to this tomorrow.

Thanks for the work