mm/memory_hotplug: full support for

We want to specify flags when hotplugging memory. Let's prepare to pass flags to memblock_add_node() by adjusting all existing users.

Note that when hotplugging memory the system is already up and running and we don't want to add the memory first and apply flags later: it should happen within one memblock call.

Signed-off-by David Hildenbrand <[email protected]>

 arch/m68k/mm/mcfmmu.c            | 3 ++-
 arch/m68k/mm/motorola.c          | 6 ++++--

We want to specify flags when hotplugging memory. Let's prepare to pass flags to memblock_add_node() by adjusting all existing users.

Note that when hotplugging memory the system is already up and running and we don't want to add the memory first and apply flags later: it should happen within one memblock call.

Signed-off-by David Hildenbrand <[email protected]>

---

 arch/s390/kernel/setup.c         | 3 ++-

We want to specify flags when hotplugging memory. Let's prepare to pass flags to memblock_add_node() by adjusting all existing users.

Note that when hotplugging memory the system is already up and running and we don't want to add the memory first and apply flags later: it should happen within one memblock call.

Signed-off-by David Hildenbrand <[email protected]>

---
 arch/arc/mm/init.c               | 4 ++--
 arch/ia64/mm/contig.c            | 2 +-
 arch/ia64/mm/init.c              | 2 +-
 arch/m68k/mm/mcfmmu.c            | 3 ++-
 arch/m68k/mm/motorola.c          | 6 ++++--
 arch/mips/loongson64/init.c      | 4 +++-
 arch/mips/sgi-ip27/ip27-memory.c | 3 ++-
 arch/s390/kernel/setup.c         | 3 ++-
 include/linux/memblock.h         | 3 ++-
 include/linux/mm.h               | 2 +-
 mm/memblock.c                    | 9 +++++----
 mm/memory_hotplug.c              | 2 +-
 12 files changed, 26 insertions(+), 17 deletions(-)

arch/arc/mm/init.c +2 -2

index 699ecf119641..110eb69e9bee 100644
@@ -59,13 +59,13 @@ void __init early_init_dt_add_memory_arch(u64 base, u64 size)
 
 		low_mem_sz = size;
 		in_use = 1;
-		memblock_add_node(base, size, 0);
+		memblock_add_node(base, size, 0, MEMBLOCK_NONE);
 	} else {
 #ifdef CONFIG_HIGHMEM
 		high_mem_start = base;
 		high_mem_sz = size;
 		in_use = 1;
-		memblock_add_node(base, size, 1);
+		memblock_add_node(base, size, 1, MEMBLOCK_NONE);
 		memblock_reserve(base, size);
 #endif
 	}

arch/ia64/mm/contig.c +1 -1

index 42e025cfbd08..24901d809301 100644
@@ -153,7 +153,7 @@ find_memory (void)
 	efi_memmap_walk(find_max_min_low_pfn, NULL);
 	max_pfn = max_low_pfn;
 
-	memblock_add_node(0, PFN_PHYS(max_low_pfn), 0);
+	memblock_add_node(0, PFN_PHYS(max_low_pfn), 0, MEMBLOCK_NONE);
 
 	find_initrd();
 

arch/ia64/mm/init.c +1 -1

index 5c6da8d83c1a..5d165607bf35 100644
@@ -378,7 +378,7 @@ int __init register_active_ranges(u64 start, u64 len, int nid)
 #endif
 
 	if (start < end)
-		memblock_add_node(__pa(start), end - start, nid);
+		memblock_add_node(__pa(start), end - start, nid, MEMBLOCK_NONE);
 	return 0;
 }
 

arch/m68k/mm/mcfmmu.c +2 -1

index eac9dde65193..6f1f25125294 100644
@@ -174,7 +174,8 @@ void __init cf_bootmem_alloc(void)
 	m68k_memory[0].addr = _rambase;
 	m68k_memory[0].size = _ramend - _rambase;
 
-	memblock_add_node(m68k_memory[0].addr, m68k_memory[0].size, 0);
+	memblock_add_node(m68k_memory[0].addr, m68k_memory[0].size, 0,
+			  MEMBLOCK_NONE);
 
 	/* compute total pages in system */
 	num_pages = PFN_DOWN(_ramend - _rambase);

arch/m68k/mm/motorola.c +4 -2

index 3a653f0a4188..e80c5d7e6728 100644
@@ -410,7 +410,8 @@ void __init paging_init(void)
 
 	min_addr = m68k_memory[0].addr;
 	max_addr = min_addr + m68k_memory[0].size;
-	memblock_add_node(m68k_memory[0].addr, m68k_memory[0].size, 0);
+	memblock_add_node(m68k_memory[0].addr, m68k_memory[0].size, 0,
+			  MEMBLOCK_NONE);
 	for (i = 1; i < m68k_num_memory;) {
 		if (m68k_memory[i].addr < min_addr) {
 			printk("Ignoring memory chunk at 0x%lx:0x%lx before the first chunk\n",

@@ -421,7 +422,8 @@ void __init paging_init(void)
 				(m68k_num_memory - i) * sizeof(struct m68k_mem_info));
 			continue;
 		}
-		memblock_add_node(m68k_memory[i].addr, m68k_memory[i].size, i);
+		memblock_add_node(m68k_memory[i].addr, m68k_memory[i].size, i,
+				  MEMBLOCK_NONE);
 		addr = m68k_memory[i].addr + m68k_memory[i].size;
 		if (addr > max_addr)
 			max_addr = addr;

arch/mips/loongson64/init.c +3 -1

index 76e0a9636a0e..4ac5ba80bbf6 100644
@@ -77,7 +77,9 @@ void __init szmem(unsigned int node)
 				(u32)node_id, mem_type, mem_start, mem_size);
 			pr_info("       start_pfn:0x%llx, end_pfn:0x%llx, num_physpages:0x%lx\n",
 				start_pfn, end_pfn, num_physpages);
-			memblock_add_node(PFN_PHYS(start_pfn), PFN_PHYS(node_psize), node);
+			memblock_add_node(PFN_PHYS(start_pfn),
+					  PFN_PHYS(node_psize), node,
+					  MEMBLOCK_NONE);
 			break;
 		case SYSTEM_RAM_RESERVED:
 			pr_info("Node%d: mem_type:%d, mem_start:0x%llx, mem_size:0x%llx MB\n",

arch/mips/sgi-ip27/ip27-memory.c +2 -1

index 6173684b5aaa..adc2faeecf7c 100644
@@ -341,7 +341,8 @@ static void __init szmem(void)
 				continue;
 			}
 			memblock_add_node(PFN_PHYS(slot_getbasepfn(node, slot)),
-					  PFN_PHYS(slot_psize), node);
+					  PFN_PHYS(slot_psize), node,
+					  MEMBLOCK_NONE);
 		}
 	}
 }

arch/s390/kernel/setup.c +2 -1

index 67e5fff96ee0..f3943f15af6e 100644
@@ -593,7 +593,8 @@ static void __init setup_resources(void)
 	 * part of the System RAM resource.
 	 */
 	if (crashk_res.end) {
-		memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
+		memblock_add_node(crashk_res.start, resource_size(&crashk_res),
+				  0, MEMBLOCK_NONE);
 		memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 		insert_resource(&iomem_resource, &crashk_res);
 	}

include/linux/memblock.h +2 -1

index 34de69b3b8ba..b49a58f621bc 100644
@@ -100,7 +100,8 @@ static inline void memblock_discard(void) {}
 #endif
 
 void memblock_allow_resize(void);
-int memblock_add_node(phys_addr_t base, phys_addr_t size, int nid);
+int memblock_add_node(phys_addr_t base, phys_addr_t size, int nid,
+		      enum memblock_flags flags);
 int memblock_add(phys_addr_t base, phys_addr_t size);
 int memblock_remove(phys_addr_t base, phys_addr_t size);
 int memblock_free(phys_addr_t base, phys_addr_t size);

include/linux/mm.h +1 -1

index 73a52aba448f..0117cb35b212 100644
@@ -2447,7 +2447,7 @@ static inline unsigned long get_num_physpages(void)
  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
  * 							 max_highmem_pfn};
  * for_each_valid_physical_page_range()
- * 	memblock_add_node(base, size, nid)
+ *	memblock_add_node(base, size, nid, MEMBLOCK_NONE)
  * free_area_init(max_zone_pfns);
  */
 void free_area_init(unsigned long *max_zone_pfn);

mm/memblock.c +5 -4

index 184dcd2e5d99..47a56b223141 100644
@@ -655,6 +655,7 @@ static int __init_memblock memblock_add_range(struct memblock_type *type,
  * @base: base address of the new region
  * @size: size of the new region
  * @nid: nid of the new region
+ * @flags: flags of the new region
  *
  * Add new memblock region [@base, @base + @size) to the "memory"
  * type. See memblock_add_range() description for mode details

@@ -663,14 +664,14 @@ static int __init_memblock memblock_add_range(struct memblock_type *type,
  * 0 on success, -errno on failure.
  */
 int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size,
-				       int nid)
+				      int nid, enum memblock_flags flags)
 {
 	phys_addr_t end = base + size - 1;
 
-	memblock_dbg("%s: [%pa-%pa] nid=%d %pS\n", __func__,
-		     &base, &end, nid, (void *)_RET_IP_);
+	memblock_dbg("%s: [%pa-%pa] nid=%d flags=%x %pS\n", __func__,
+		     &base, &end, nid, flags, (void *)_RET_IP_);
 
-	return memblock_add_range(&memblock.memory, base, size, nid, 0);
+	return memblock_add_range(&memblock.memory, base, size, nid, flags);
 }
 
 /**

mm/memory_hotplug.c +1 -1

index 917b3528636d..5f873e7f5b29 100644
@@ -1385,7 +1385,7 @@ int __ref add_memory_resource(int nid, struct resource *res, mhp_t mhp_flags)
 	mem_hotplug_begin();
 
 	if (IS_ENABLED(CONFIG_ARCH_KEEP_MEMBLOCK)) {
-		ret = memblock_add_node(start, size, nid);
+		ret = memblock_add_node(start, size, nid, MEMBLOCK_NONE);
 		if (ret)
 			goto error_mem_hotplug_end;
 	}

Signature · 1 line

2.31.1

Quoted 3 lines · David Hildenbrand

We want to specify flags when hotplugging memory. Let's prepare to pass flags to memblock_add_node() by adjusting all existing users.

Quoted 3 lines

Note that when hotplugging memory the system is already up and running and we don't want to add the memory first and apply flags later: it should happen within one memblock call.

Why is it important that the system is up and why it should happen in a single call?
I don't mind adding flags parameter to memblock_add_node() but this changelog does not really explain the reasons to do it.

Let's add a flag that corresponds to IORESOURCE_SYSRAM_DRIVER_MANAGED. Similar to MEMBLOCK_HOTPLUG, most infrastructure has to treat such memory like ordinary MEMBLOCK_NONE memory -- for example, when selecting memory regions to add to the vmcore for dumping in the crashkernel via for_each_mem_range().

However, especially kexec_file is not supposed to select such memblocks via for_each_free_mem_range() / for_each_free_mem_range_reverse() to place kexec images, similar to how we handle IORESOURCE_SYSRAM_DRIVER_MANAGED without CONFIG_ARCH_KEEP_MEMBLOCK.

Let's document why kexec_walk_memblock() won't try placing images on areas marked MEMBLOCK_DRIVER_MANAGED -- similar to IORESOURCE_SYSRAM_DRIVER_MANAGED handling in locate_mem_hole_callback() via kexec_walk_resources().

We'll make sure that memory hotplug code sets the flag where applicable (IORESOURCE_SYSRAM_DRIVER_MANAGED) next. This prepares architectures that need CONFIG_ARCH_KEEP_MEMBLOCK, such as arm64, for virtio-mem support.

Signed-off-by David Hildenbrand <[email protected]>

---
 include/linux/memblock.h | 16 ++++++++++++++--
 kernel/kexec_file.c      |  5 +++++
 mm/memblock.c            |  4 ++++
 3 files changed, 23 insertions(+), 2 deletions(-)

include/linux/memblock.h +14 -2

index b49a58f621bc..7d8d656d5082 100644
@@ -33,12 +33,17 @@ extern unsigned long long max_possible_pfn;
  * @MEMBLOCK_NOMAP: don't add to kernel direct mapping and treat as
  * reserved in the memory map; refer to memblock_mark_nomap() description
  * for further details
+ * @MEMBLOCK_DRIVER_MANAGED: memory region that is always detected via a driver,
+ * corresponding to IORESOURCE_SYSRAM_DRIVER_MANAGED in the kernel resource
+ * tree. Especially kexec should never use this memory for placing images and
+ * shouldn't expose this memory to the second kernel.
  */
 enum memblock_flags {
 	MEMBLOCK_NONE		= 0x0,	/* No special request */
 	MEMBLOCK_HOTPLUG	= 0x1,	/* hotpluggable region */
 	MEMBLOCK_MIRROR		= 0x2,	/* mirrored region */
 	MEMBLOCK_NOMAP		= 0x4,	/* don't add to kernel direct mapping */
+	MEMBLOCK_DRIVER_MANAGED = 0x8,	/* always detected via a driver */
 };
 
 /**

@@ -209,7 +214,8 @@ static inline void __next_physmem_range(u64 *idx, struct memblock_type *type,
  */
 #define for_each_mem_range(i, p_start, p_end) \
 	__for_each_mem_range(i, &memblock.memory, NULL, NUMA_NO_NODE,	\
-			     MEMBLOCK_HOTPLUG, p_start, p_end, NULL)
+			     MEMBLOCK_HOTPLUG | MEMBLOCK_DRIVER_MANAGED, \
+			     p_start, p_end, NULL)
 
 /**
  * for_each_mem_range_rev - reverse iterate through memblock areas from

@@ -220,7 +226,8 @@ static inline void __next_physmem_range(u64 *idx, struct memblock_type *type,
  */
 #define for_each_mem_range_rev(i, p_start, p_end)			\
 	__for_each_mem_range_rev(i, &memblock.memory, NULL, NUMA_NO_NODE, \
-				 MEMBLOCK_HOTPLUG, p_start, p_end, NULL)
+				 MEMBLOCK_HOTPLUG | MEMBLOCK_DRIVER_MANAGED,\
+				 p_start, p_end, NULL)
 
 /**
  * for_each_reserved_mem_range - iterate over all reserved memblock areas

@@ -250,6 +257,11 @@ static inline bool memblock_is_nomap(struct memblock_region *m)
 	return m->flags & MEMBLOCK_NOMAP;
 }
 
+static inline bool memblock_is_driver_managed(struct memblock_region *m)
+{
+	return m->flags & MEMBLOCK_DRIVER_MANAGED;
+}
+
 int memblock_search_pfn_nid(unsigned long pfn, unsigned long *start_pfn,
 			    unsigned long  *end_pfn);
 void __next_mem_pfn_range(int *idx, int nid, unsigned long *out_start_pfn,

kernel/kexec_file.c +5 -0

index 33400ff051a8..8347fc158d2b 100644
@@ -556,6 +556,11 @@ static int kexec_walk_memblock(struct kexec_buf *kbuf,
 	if (kbuf->image->type == KEXEC_TYPE_CRASH)
 		return func(&crashk_res, kbuf);
 
+	/*
+	 * Using MEMBLOCK_NONE will properly skip MEMBLOCK_DRIVER_MANAGED. See
+	 * IORESOURCE_SYSRAM_DRIVER_MANAGED handling in
+	 * locate_mem_hole_callback().
+	 */
 	if (kbuf->top_down) {
 		for_each_free_mem_range_reverse(i, NUMA_NO_NODE, MEMBLOCK_NONE,
 						&mstart, &mend, NULL) {

mm/memblock.c +4 -0

index 47a56b223141..540a35317fb0 100644
@@ -979,6 +979,10 @@ static bool should_skip_region(struct memblock_type *type,
 	if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m))
 		return true;
 
+	/* skip driver-managed memory unless we were asked for it explicitly */
+	if (!(flags & MEMBLOCK_DRIVER_MANAGED) && memblock_is_driver_managed(m))
+		return true;
+
 	return false;
 }
 

Signature · 1 line

2.31.1

Hi,

Quoted 5 lines · David Hildenbrand

Let's add a flag that corresponds to IORESOURCE_SYSRAM_DRIVER_MANAGED. Similar to MEMBLOCK_HOTPLUG, most infrastructure has to treat such memory like ordinary MEMBLOCK_NONE memory -- for example, when selecting memory regions to add to the vmcore for dumping in the crashkernel via for_each_mem_range().

Can you please elaborate on the difference in semantics of MEMBLOCK_HOTPLUG and MEMBLOCK_DRIVER_MANAGED?
Unless I'm missing something they both mark memory that can be unplugged anytime and so it should not be used in certain cases. Why is there a need for a new flag?

Quoted 8 lines · Mike Rapoport

Hi,

Quoted 5 lines · David Hildenbrand

Let's add a flag that corresponds to IORESOURCE_SYSRAM_DRIVER_MANAGED. Similar to MEMBLOCK_HOTPLUG, most infrastructure has to treat such memory like ordinary MEMBLOCK_NONE memory -- for example, when selecting memory regions to add to the vmcore for dumping in the crashkernel via for_each_mem_range().

Quoted 5 lines

Can you please elaborate on the difference in semantics of MEMBLOCK_HOTPLUG and MEMBLOCK_DRIVER_MANAGED?
Unless I'm missing something they both mark memory that can be unplugged anytime and so it should not be used in certain cases. Why is there a need for a new flag?

In the cover letter I have "Alternative B: Reuse MEMBLOCK_HOTPLUG. MEMBLOCK_HOTPLUG serves a different purpose, though.", but looking into the details it won't work as is.

MEMBLOCK_HOTPLUG is used to mark memory early during boot that can later get hotunplugged again and should be placed into ZONE_MOVABLE if the "movable_node" kernel parameter is set.

The confusing part is that we talk about "hotpluggable" but really mean "hotunpluggable": the reason is that HW flags DIMM slots that can later be hotplugged as "hotpluggable" even though there is already something hotplugged.

For example, ranges in the ACPI SRAT that are marked as ACPI_SRAT_MEM_HOT_PLUGGABLE will be marked MEMBLOCK_HOTPLUG early during boot (drivers/acpi/numa/srat.c:acpi_numa_memory_affinity_init()). Later, we use that information to size ZONE_MOVABLE (mm/page_alloc.c:find_zone_movable_pfns_for_nodes()). This will make sure that these "hotpluggable" DIMMs can later get hotunplugged.

Also, see should_skip_region() how this relates to the "movable_node" kernel parameter:

	/* skip hotpluggable memory regions if needed */
	if (movable_node_is_enabled() && memblock_is_hotpluggable(m) &&
	    (flags & MEMBLOCK_HOTPLUG))
		return true;

Quoted 19 lines · David Hildenbrand

Quoted 8 lines · Mike Rapoport

Hi,

Quoted 5 lines · David Hildenbrand

Let's add a flag that corresponds to IORESOURCE_SYSRAM_DRIVER_MANAGED. Similar to MEMBLOCK_HOTPLUG, most infrastructure has to treat such memory like ordinary MEMBLOCK_NONE memory -- for example, when selecting memory regions to add to the vmcore for dumping in the crashkernel via for_each_mem_range().

Quoted 5 lines

Can you please elaborate on the difference in semantics of MEMBLOCK_HOTPLUG and MEMBLOCK_DRIVER_MANAGED?
Unless I'm missing something they both mark memory that can be unplugged anytime and so it should not be used in certain cases. Why is there a need for a new flag?

In the cover letter I have "Alternative B: Reuse MEMBLOCK_HOTPLUG. MEMBLOCK_HOTPLUG serves a different purpose, though.", but looking into the details it won't work as is.

Quoted 8 lines

MEMBLOCK_HOTPLUG is used to mark memory early during boot that can later get hotunplugged again and should be placed into ZONE_MOVABLE if the "movable_node" kernel parameter is set.

The confusing part is that we talk about "hotpluggable" but really mean "hotunpluggable": the reason is that HW flags DIMM slots that can later be hotplugged as "hotpluggable" even though there is already something hotplugged.

MEMBLOCK_HOTPLUG name is indeed somewhat confusing, but still it's core meaning "this memory may be removed" which does not differ from what IORESOURCE_SYSRAM_DRIVER_MANAGED means.

MEMBLOCK_HOTPLUG regions are indeed placed into ZONE_MOVABLE, but more importantly, they are avoided when we allocate memory from memblock.

So, in my view, both flags mean that the memory may be removed and it should not be used for certain types of allocations.

Quoted 11 lines

For example, ranges in the ACPI SRAT that are marked as ACPI_SRAT_MEM_HOT_PLUGGABLE will be marked MEMBLOCK_HOTPLUG early during boot (drivers/acpi/numa/srat.c:acpi_numa_memory_affinity_init()). Later, we use that information to size ZONE_MOVABLE (mm/page_alloc.c:find_zone_movable_pfns_for_nodes()). This will make sure that these "hotpluggable" DIMMs can later get hotunplugged.

Also, see should_skip_region() how this relates to the "movable_node" kernel parameter:

Quoted 4 lines

	/* skip hotpluggable memory regions if needed */
	if (movable_node_is_enabled() && memblock_is_hotpluggable(m) &&
	    (flags & MEMBLOCK_HOTPLUG))
		return true;

Hmm, I think that the movable_node_is_enabled() check here is excessive, but I suspect we cannot simply remove it without breaking anything.

I'll take a deeper look on the potential consequences.

BTW, is there anything that prevents putting kexec to hot-unplugable memory that was cold-plugged on boot?

Quoted 33 lines · Mike Rapoport

Quoted 19 lines · David Hildenbrand

Quoted 8 lines · Mike Rapoport

Hi,

Quoted 5 lines · David Hildenbrand

Let's add a flag that corresponds to IORESOURCE_SYSRAM_DRIVER_MANAGED. Similar to MEMBLOCK_HOTPLUG, most infrastructure has to treat such memory like ordinary MEMBLOCK_NONE memory -- for example, when selecting memory regions to add to the vmcore for dumping in the crashkernel via for_each_mem_range().

Quoted 5 lines

Can you please elaborate on the difference in semantics of MEMBLOCK_HOTPLUG and MEMBLOCK_DRIVER_MANAGED?
Unless I'm missing something they both mark memory that can be unplugged anytime and so it should not be used in certain cases. Why is there a need for a new flag?

In the cover letter I have "Alternative B: Reuse MEMBLOCK_HOTPLUG. MEMBLOCK_HOTPLUG serves a different purpose, though.", but looking into the details it won't work as is.

Quoted 8 lines

MEMBLOCK_HOTPLUG is used to mark memory early during boot that can later get hotunplugged again and should be placed into ZONE_MOVABLE if the "movable_node" kernel parameter is set.

The confusing part is that we talk about "hotpluggable" but really mean "hotunpluggable": the reason is that HW flags DIMM slots that can later be hotplugged as "hotpluggable" even though there is already something hotplugged.

MEMBLOCK_HOTPLUG name is indeed somewhat confusing, but still it's core meaning "this memory may be removed" which does not differ from what IORESOURCE_SYSRAM_DRIVER_MANAGED means.

Quoted 5 lines

MEMBLOCK_HOTPLUG regions are indeed placed into ZONE_MOVABLE, but more importantly, they are avoided when we allocate memory from memblock.

So, in my view, both flags mean that the memory may be removed and it should not be used for certain types of allocations.

The semantics are different:

MEMBLOCK_HOTPLUG: memory is indicated as "System RAM" in the firmware-provided memory map and added to the system early during boot; we want this memory to be managed by ZONE_MOVABLE with "movable_node" set on the kernel command line, because only then we want it to be hotpluggable again. kexec *has to* indicate this memory to the second kernel and can place kexec-images on this memory. After memory hotunplug, kexec has to be re-armed.

MEMBLOCK_DRIVER_MANAGED: memory is not indicated as System RAM" in the firmware-provided memory map; this memory is always detected and added to the system by a driver; memory might not actually be physically hotunpluggable and the ZONE selection does not depend on "movable_core". kexec *must not* indicate this memory to the second kernel and *must not* place kexec-images on this memory.