Re: [RFC PATCH] Memory hotplug support for arm64 platform

From: Scott Branden
Date: Wed Feb 08 2017 - 15:08:32 EST

Next message: Christopher Covington: "[PATCH v7] arm64: Work around Falkor erratum 1003"
Previous message: Greg Kroah-Hartman: "Re: [PATCH 4.4 00/29] 4.4.48-stable review"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Hi Andrea,

On 17-02-06 03:17 AM, Andrea Reale wrote:

Hi Scott,
Hi all,

in reply to the issues that Scott reported last month, myself and Maciej
investigated further by running quite a number of experiments on the
physical and virtual environments we have avaialable.

We collected all the results and relevant logs in a Web page at
https://hotplug-tests.eu-gb.mybluemix.net/ so that anyone interested can
go there and check all the details.

The tl;dr version is that, in all configuration, we could not reproduce
what Scott has described as "memory corruption". The only issue we
encountered happens when the system is booted with a small amount of
initial memory (e.g., mem=64M) and one tries to hot-add several sections
of memory in ZONE_MOVABLE; in that case, the process is likely to fail
when vmemmap tries to allocate chunks of 2^9 consecutive pages to make
space for the `struct page`s describing the new memory; in fact, it
seems likely that, in low memory situations, the system cannot find enough
consecutive pages in ZONE_DMA or ZONE_NORMAL. This condition is not
dependand on memory hot-plug; in fact, we counter-tested this by writing
a simple module that just tries to allocate a few chunks of 2^9 pages,
and we experienced that it fails when the system is booted with low
memory (sources and logs in the Web page linked above).

@Scott: were your referring to this issue, by any chance, in your
previous emails? If not, we would really appreciate if you could help us
reproduce the condition you are experiencing and/or give us a more detail
of what are the symptoms of the corruption you are referring to.

The condition I experience is that after hot-adding additional memory (64M pages) the system crashes after a few seconds (differently). I have not had a chance to debug what the root of the problem is as of yet.

I will attempt other memory configurations and try debugging more when I have a chance (Sorry! I've been working on other things and haven't debugged further yet)

We are still running additional tests on other boards and we will update
the Web page while we get them. If anyone happens to try these patches
on their system, we warmly invite to send feedback with either
negative or positive outcomes.

Thanks and best regards,
Andrea

On Wed, Dec 21, 2016 at 05:20:06PM -0800, Scott Branden wrote:

Hi Maciej,

On 16-12-21 01:44 AM, Maciej Bielski wrote:

Hi Scott,

Thanks for testing it and providing us the feedback. For replicating the problem you have reported,
could you provide more info on your system configuration, please?
Among others, few questions that we have in mind are:
* What is the RAM size at boot? Is it multiple of 1GB (or precisely `1<<MIN_MEMORY_BLOCK_SIZE`) ?

RAM size at boot is 64MB, not a multiple of 1GB. I will try next
week on a system with multi-GB at boot to see if it makes a
difference. SECTION_SIZE_BITS in sparsemem.h has been modified to 28
to allow for 256MB hotplug regions to be added.

* Do you run the kernel with `mem=YYY` option?

no, memory is defined in dts file with a few reserved regions:
/memreserve/ 0x75000000 0x00800000;
/memreserve/ 0x77f00000 0x00100000;
&memory {
reg = <0x00000000 0x74000000 0x0 0x04000000>; /* 64MB */
};

* Do you follow steps (1) and (2) for performing the hotplug?

no, only step 1 is done, auto-online is enabled. I'll try turning
that off and see if it makes a difference.

* Do you have any other configuration flags enabled except for CONFIG_MEMORY_HOTPLUG?
* Any other non-defaults?

Yes, a lot of config options have been disabled to reduce kernel
size. The memory auto-online is also enabled. I'll try with the
default defconfig to see if it makes a difference.

Up to now the patch was passing our simple tests but if we miss something we will be very thankful
if you could help us to sort it out. If you have found something after your deeper analysis, please
keep us informed.

I think the path forward is to get a normal multi-GB system going
and then reduce the memory on boot and then hotplug it in after to
see if that works. Or, if you can reduce your memory down and give
it a try that might uncover a problem as well.

On 20/12/2016 20:12, Scott Branden wrote:

Hi Maciej,

I have applied that patch ontop of the patches I previously sent out
and tested.

It does recognized the memory in /proc/iomem but I get memory corruption of the original system
RAM soon after. It appears the page allocation gets corrupted. I will try to dig into it further
but if somebody else could try it out in their system to see what results they get it would help.

Regards,
Scott

On 16-12-14 04:16 AM, Maciej Bielski wrote:

This patch relates to the work previously announced in [1]. This builds on the
work by Scott Branden [2] and, henceforth, it needs to be applied on top of
Scott's patches [2]. Comments are very welcome.

Changes from the original patchset and known issues:

- Compared to Scott's original patchset, this work adds the mapping of
the new hotplugged pages into the kernel page tables. This is done by
copying the old swapper_pg_dir over a new page, adding the new mappings,
and then switching to the newly built pg_dir (see `hotplug_paging` in
arch/arm64/mmu.c). There might be better ways to to this: suggestions
are more than welcome.

- The stub function for `arch_remove_memory` has been removed for now; we
are working in parallel on memory hot remove, and we plan to contribute
it as a separate patch.

- Corresponding Kconfig flags have been added;

- Note that this patch does not work when NUMA is enabled; in fact,
the function `memory_add_physaddr_to_nid` does not have an
implementation when the NUMA flag is on: this function is supposed to
return the nid the hotplugged memory should be associated with. However
it is not really clear to us yet what the semantics of this function
in the context of a NUMA system should be. A quick and dirty fix would
be to always attach to the first available NUMA node.

- In arch/arm64/mm/init.c `arch_add_memory`, we are doing a hack with the
nomap memory block flags to satisfy preconditions and postconditions of
`__add_pages` and postconditions of `arch_add_memory`. Compared to
memory hotplug implementation for other architectures, the "issue"
seems to be in the implemenation of `pfn_valid`. Suggestions on how
to cleanly avoid this hack are welcome.

This patchset can be tested by starting the kernel with the `mem=X` flag, where
X is less than the total available physical memory and has to be multiple of
MIN_MEMORY_BLOCK_SIZE. We also tested it on a customised version of QEMU
capable to emulate physical hotplug on arm64 platform.

To enable the feature the CONFIG_MEMORY_HOTPLUG compilation flag
needs to be set to true. Then, after memory is physically hotplugged,
the standard two steps to make it available (as also documented in
Documentation/memory-hotplug.txt) are:

(1) Notify memory hot-add
echo '0xYY000000' > /sys/devices/system/memory/probe

where 0xYY000000 is the first physical address of the new memory section.

(2) Online new memory block(s)
echo online > /sys/devices/system/memory/memoryXXX/state

where XXX corresponds to the ids of newly added blocks.

Onlining can optionally be automatic at hot-add notification by enabling
the global flag:
echo online > /sys/devices/system/memory/auto_online_blocks
or by setting the corresponding config flag in the kernel build.

Again, any comment is highly appreciated.

[1] https://lkml.org/lkml/2016/11/17/49
[2] https://lkml.org/lkml/2016/12/1/811

Signed-off-by: Maciej Bielski <m.bielski@xxxxxxxxxxxxxxxxxxxxxx>
Signed-off-by: Andrea Reale <ar@xxxxxxxxxxxxxxxxxx>
---
arch/arm64/Kconfig | 4 +--
arch/arm64/include/asm/mmu.h | 3 +++
arch/arm64/mm/init.c | 58 +++++++++++++++++++++++++++++++-------------
arch/arm64/mm/mmu.c | 24 ++++++++++++++++++
include/linux/memblock.h | 1 +
mm/memblock.c | 10 ++++++++
6 files changed, 80 insertions(+), 20 deletions(-)

diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
index 2482fdd..bd8ddf2 100644
--- a/arch/arm64/Kconfig
+++ b/arch/arm64/Kconfig
@@ -577,9 +577,7 @@ config HOTPLUG_CPU
can be controlled through /sys/devices/system/cpu.

config ARCH_ENABLE_MEMORY_HOTPLUG
- def_bool y
-
-config ARCH_ENABLE_MEMORY_HOTREMOVE
+ depends on !NUMA
def_bool y

# Common NUMA Features
diff --git a/arch/arm64/include/asm/mmu.h b/arch/arm64/include/asm/mmu.h
index 8d9fce0..2499745 100644
--- a/arch/arm64/include/asm/mmu.h
+++ b/arch/arm64/include/asm/mmu.h
@@ -36,5 +36,8 @@ extern void create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys,
unsigned long virt, phys_addr_t size,
pgprot_t prot, bool allow_block_mappings);
extern void *fixmap_remap_fdt(phys_addr_t dt_phys);
+#ifdef CONFIG_MEMORY_HOTPLUG
+extern void hotplug_paging(phys_addr_t start, phys_addr_t size);
+#endif

#endif
diff --git a/arch/arm64/mm/init.c b/arch/arm64/mm/init.c
index 687d087..a7c740e 100644
--- a/arch/arm64/mm/init.c
+++ b/arch/arm64/mm/init.c
@@ -544,37 +544,61 @@ int arch_add_memory(int nid, u64 start, u64 size, bool for_device)
struct zone *zone;
unsigned long start_pfn = start >> PAGE_SHIFT;
unsigned long nr_pages = size >> PAGE_SHIFT;
+ unsigned long end_pfn = start_pfn + nr_pages;
+ unsigned long max_sparsemem_pfn = 1UL << (MAX_PHYSMEM_BITS-PAGE_SHIFT);
+ unsigned long pfn;
int ret;

+ if (end_pfn > max_sparsemem_pfn) {
+ pr_err("end_pfn too big");
+ return -1;
+ }
+ hotplug_paging(start, size);
+
+ /*
+ * Mark all the page range as unsuable.
+ * This is needed because __add_section (within __add_pages)
+ * wants pfn_valid to be false, and in arm64 pfn falid is implemented
+ * by just checking at the nomap flag for existing blocks
+ */
+ memblock_mark_nomap(start, size);
+
pgdat = NODE_DATA(nid);

zone = pgdat->node_zones +
zone_for_memory(nid, start, size, ZONE_NORMAL, for_device);
ret = __add_pages(nid, zone, start_pfn, nr_pages);

- if (ret)
- pr_warn("%s: Problem encountered in __add_pages() ret=%d\n",
- __func__, ret);
+ /*
+ * Make the pages usable after they have been added.
+ * This will make pfn_valid return true
+ */
+ memblock_clear_nomap(start, size);

- return ret;
-}
+ /*
+ * This is a hack to avoid having to mix arch specific code into arch
+ * independent code. SetPageReserved is supposed to be called by __add_zone
+ * (within __add_section, within __add_pages). However, when it is called
+ * there, it assumes that pfn_valid returns true. For the way pfn_valid is
+ * implemented in arm64 (a check on the nomap flag), the only way to make
+ * this evaluate true inside __add_zone is to clear the nomap flags of
+ * blocks in architecture independent code.
+ *
+ * To avoid this, we set the Reserved flag here after we cleared the nomap
+ * flag in the line above.
+ */
+ for (pfn = start_pfn; pfn < start_pfn + nr_pages; pfn++) {
+ if (!pfn_valid(pfn))
+ continue;

-#ifdef CONFIG_MEMORY_HOTREMOVE
-int arch_remove_memory(u64 start, u64 size)
-{
- unsigned long start_pfn = start >> PAGE_SHIFT;
- unsigned long nr_pages = size >> PAGE_SHIFT;
- struct zone *zone;
- int ret;
+ SetPageReserved(pfn_to_page(pfn));
+ }

- zone = page_zone(pfn_to_page(start_pfn));
- ret = __remove_pages(zone, start_pfn, nr_pages);
if (ret)
- pr_warn("%s: Problem encountered in __remove_pages() ret=%d\n",
+ pr_warn("%s: Problem encountered in __add_pages() ret=%d\n",
__func__, ret);

return ret;
}
#endif
-#endif

diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
index 05615a3..9efa7d1 100644
--- a/arch/arm64/mm/mmu.c
+++ b/arch/arm64/mm/mmu.c
@@ -493,6 +493,30 @@ void __init paging_init(void)
SWAPPER_DIR_SIZE - PAGE_SIZE);
}

+#ifdef CONFIG_MEMORY_HOTPLUG
+/*
+ * hotplug_paging() is used by memory hotplug to build new page tables
+ * for hot added memory.
+ */
+void hotplug_paging(phys_addr_t start, phys_addr_t size)
+{
+ phys_addr_t pgd_phys = pgd_pgtable_alloc();
+ pgd_t *pgd = pgd_set_fixmap(pgd_phys);
+
+ memcpy(pgd, swapper_pg_dir, PAGE_SIZE);
+
+ __create_pgd_mapping(pgd, start, __phys_to_virt(start), size,
+ PAGE_KERNEL, pgd_pgtable_alloc, false);
+
+ cpu_replace_ttbr1(__va(pgd_phys));
+ memcpy(swapper_pg_dir, pgd, PAGE_SIZE);
+ cpu_replace_ttbr1(swapper_pg_dir);
+
+ pgd_clear_fixmap();
+ memblock_free(pgd_phys, PAGE_SIZE);
+}
+#endif
+
/*
* Check whether a kernel address is valid (derived from arch/x86/).
*/
diff --git a/include/linux/memblock.h b/include/linux/memblock.h
index 5b759c9..5f78257 100644
--- a/include/linux/memblock.h
+++ b/include/linux/memblock.h
@@ -92,6 +92,7 @@ int memblock_mark_hotplug(phys_addr_t base, phys_addr_t size);
int memblock_clear_hotplug(phys_addr_t base, phys_addr_t size);
int memblock_mark_mirror(phys_addr_t base, phys_addr_t size);
int memblock_mark_nomap(phys_addr_t base, phys_addr_t size);
+int memblock_clear_nomap(phys_addr_t base, phys_addr_t size);
ulong choose_memblock_flags(void);

/* Low level functions */
diff --git a/mm/memblock.c b/mm/memblock.c
index 7608bc3..05e7676 100644
--- a/mm/memblock.c
+++ b/mm/memblock.c
@@ -814,6 +814,16 @@ int __init_memblock memblock_mark_nomap(phys_addr_t base, phys_addr_t size)
}

/**
+ * memblock_clear_nomap - Clear a flag of MEMBLOCK_NOMAP memory region
+ * @base: the base phys addr of the region
+ * @size: the size of the region
+ */
+int __init_memblock memblock_clear_nomap(phys_addr_t base, phys_addr_t size)
+{
+ return memblock_setclr_flag(base, size, 0, MEMBLOCK_NOMAP);
+}
+
+/**
* __next_reserved_mem_region - next function for for_each_reserved_region()
* @idx: pointer to u64 loop variable
* @out_start: ptr to phys_addr_t for start address of the region, can be %NULL

BR,

Next message: Christopher Covington: "[PATCH v7] arm64: Work around Falkor erratum 1003"
Previous message: Greg Kroah-Hartman: "Re: [PATCH 4.4 00/29] 4.4.48-stable review"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]