[PATCH v2 0/3] kho: fix bootmem sizing

[PATCH v2 0/3] kho: fix bootmem sizing

[Mike Rapoport (Microsoft)]

There were several reports about issues with KHO bootmem sizing.

The bootmem areas are sized from overlapping views of the memory reserved
in memblock, so on NUMA systems memory allocated on a node is accounted
twice and with CONFIG_NUMA=n the lowmem and the global areas are accounted
once again in the area of node 0. Besides, the lowmem and the global areas
are reserved even when they are empty.

The first patch makes kexec walk the bootmem areas in address order.
Otherwise kexec on arm64 may fail if a small area at the top of the memory
is the first in the list, and with smaller bootmem areas this becomes more
likely.

The second patch skips empty bootmem areas and the third one makes the
lowmem, per-node and global areas cover separate parts of the reserved
memory.

---
Changes in v2:
- add a patch that walks bootmem areas in address order
- update kho_bootmem_cnt directly in kho_reserve_bootmem_area() (Pratyush)
- don't reserve the lowmem area if ARCH_LOW_ADDRESS_LIMIT covers all
  memory (Pratyush)
- extend the comment above bootmem_size_update() (Pratyush)
- Link to v1: https://patch.msgid.link/20261103-kho-bootmem-sizing-v1-0-45aa74098f34@kernel.org

---
Mike Rapoport (Microsoft) (3):
      kho: walk bootmem areas in address order
      kho: skip zero-sized lowmem and global bootmem areas
      kho: fix bootmem sizing

 kernel/liveupdate/kexec_handover.c | 212 ++++++++++++++++++++++---------------
 1 file changed, 128 insertions(+), 84 deletions(-)
---
base-commit: 6628e7d4c6b25c8d18f1345e3a20d21829aae7e4
change-id: 20261103-kho-bootmem-sizing-5d6aed47a243

--
Sincerely yours,
Mike.

[PATCH v2 1/3] kho: walk bootmem areas in address order

[Mike Rapoport (Microsoft)]

kexec_add_buffer() users may depend on the order in which memory holes are
searched. For instance, arm64 places the kernel bottom-up and then
requires the DTB to be above the kernel.

kho_walk_bootmem() walks the bootmem areas in the order they were
reserved, so if the kernel lands in a small area at the top of the memory
there is no room left for the DTB and kexec fails. For example, on an
arm64 VM with 2G of RAM and kho_bootmem=8M,64M,64M, the lowmem area is
reserved first at the top of the memory, the kernel image takes most of
it and kexec fails with

  kexec_file(Image): Could not find any suitable kernel location!

Sort the bootmem areas by address and walk them in ascending order, or in
descending order for top-down searches, like kexec_walk_resources() does.

Fixes: 3bdecc3c93f9 ("kexec: add KHO support to kexec file loads")
Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
---
 kernel/liveupdate/kexec_handover.c | 57 ++++++++++++++++++++++++++++++--------
 1 file changed, 46 insertions(+), 11 deletions(-)

diff --git a/kernel/liveupdate/kexec_handover.c b/kernel/liveupdate/kexec_handover.c
index 6f1f55fe31819..2ac5d94f33fd4 100644
--- a/kernel/liveupdate/kexec_handover.c
+++ b/kernel/liveupdate/kexec_handover.c
@@ -26,6 +26,7 @@
 #include <linux/list.h>
 #include <linux/memblock.h>
 #include <linux/page-isolation.h>
+#include <linux/sort.h>
 #include <linux/unaligned.h>
 #include <linux/vmalloc.h>
 
@@ -820,6 +821,14 @@ bool kho_bootmem_overlap(phys_addr_t phys, size_t size)
 	return false;
 }
 
+static int __init kho_bootmem_cmp(const void *_a, const void *_b)
+{
+	const struct kho_bootmem *a = _a;
+	const struct kho_bootmem *b = _b;
+
+	return cmp_int(a->addr, b->addr);
+}
+
 /**
  * kho_reserve_bootmem - Reserve a contiguous chunk of memory for kexec
  *
@@ -904,6 +913,9 @@ static void __init kho_reserve_bootmem(void)
 	 */
 	kho_bootmem_cnt = i;
 
+	sort(kho_bootmem, kho_bootmem_cnt, sizeof(*kho_bootmem),
+	     kho_bootmem_cmp, NULL);
+
 	return;
 
 err_free_bootmem_areas:
@@ -2081,20 +2093,40 @@ int kho_fill_kimage(struct kimage *image)
 	return 0;
 }
 
-static int kho_walk_bootmem(struct kexec_buf *kbuf,
-			    int (*func)(struct resource *, void *))
+static int kho_bootmem_walk_one(struct kexec_buf *kbuf,
+				const struct kho_bootmem *area,
+				int (*func)(struct resource *, void *))
+{
+	struct resource res = {
+		.start = area->addr,
+		.end = area->addr + area->size - 1,
+	};
+
+	/* Try to fit the kbuf into our KHO bootmem region */
+	return func(&res, kbuf);
+}
+
+static int kho_bootmem_walk_top_down(struct kexec_buf *kbuf,
+				     int (*func)(struct resource *, void *))
 {
 	int ret = 0;
-	int i;
 
-	for (i = 0; i < kho_bootmem_cnt; i++) {
-		struct resource res = {
-			.start = kho_bootmem[i].addr,
-			.end = kho_bootmem[i].addr + kho_bootmem[i].size - 1,
-		};
+	for (int i = kho_bootmem_cnt - 1; i >= 0; i--) {
+		ret = kho_bootmem_walk_one(kbuf, &kho_bootmem[i], func);
+		if (ret)
+			break;
+	}
+
+	return ret;
+}
 
-		/* Try to fit the kimage into our KHO bootmem region */
-		ret = func(&res, kbuf);
+static int kho_bootmem_walk_bottom_up(struct kexec_buf *kbuf,
+				      int (*func)(struct resource *, void *))
+{
+	int ret = 0;
+
+	for (int i = 0; i < kho_bootmem_cnt; i++) {
+		ret = kho_bootmem_walk_one(kbuf, &kho_bootmem[i], func);
 		if (ret)
 			break;
 	}
@@ -2110,7 +2142,10 @@ int kho_locate_mem_hole(struct kexec_buf *kbuf,
 	if (!kho_enable || kbuf->image->type == KEXEC_TYPE_CRASH)
 		return 1;
 
-	ret = kho_walk_bootmem(kbuf, func);
+	if (kbuf->top_down)
+		ret = kho_bootmem_walk_top_down(kbuf, func);
+	else
+		ret = kho_bootmem_walk_bottom_up(kbuf, func);
 
 	return ret == 1 ? 0 : -EADDRNOTAVAIL;
 }

-- 
2.53.0

[PATCH v2 2/3] kho: skip zero-sized lowmem and global bootmem areas

[Mike Rapoport (Microsoft)]

The lowmem and the global bootmem areas can have zero size, for example
when none or all of the memory is below ARCH_LOW_ADDRESS_LIMIT, or when
either of those areas is explicitly set to zero with kho_bootmem=.

For a zero-sized allocation memblock returns an address without reserving
anything, and that empty area is recorded in kho_bootmem. If there is no
memory below ARCH_LOW_ADDRESS_LIMIT, allocation of the empty lowmem area
fails and KHO gets disabled.

Extract a helper that checks for zero size and allocates memory from
memblock, and use it to skip empty areas the same way as empty per-node
areas are skipped.

Fixes: 3dc92c311498 ("kexec: add Kexec HandOver (KHO) generation helpers")
Reported-by: Chen Pei <cp0613@linux.alibaba.com>
Closes: https://lore.kernel.org/all/20261026131851.9942-1-cp0613@linux.alibaba.com
Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
---
 kernel/liveupdate/kexec_handover.c | 66 +++++++++++++++++---------------------
 1 file changed, 30 insertions(+), 36 deletions(-)

diff --git a/kernel/liveupdate/kexec_handover.c b/kernel/liveupdate/kexec_handover.c
index 2ac5d94f33fd4..5b1797d1289c8 100644
--- a/kernel/liveupdate/kexec_handover.c
+++ b/kernel/liveupdate/kexec_handover.c
@@ -829,6 +829,26 @@ static int __init kho_bootmem_cmp(const void *_a, const void *_b)
 	return cmp_int(a->addr, b->addr);
 }
 
+static int __init kho_reserve_bootmem_area(phys_addr_t size, phys_addr_t end,
+					   int nid)
+{
+	phys_addr_t addr;
+
+	if (!size)
+		return 0;
+
+	addr = memblock_alloc_range_nid(size, BOOTMEM_ALIGNMENT_BYTES, 0, end,
+					nid, true);
+	if (!addr)
+		return -ENOMEM;
+
+	kho_bootmem[kho_bootmem_cnt].addr = addr;
+	kho_bootmem[kho_bootmem_cnt].size = size;
+	kho_bootmem_cnt++;
+
+	return 0;
+}
+
 /**
  * kho_reserve_bootmem - Reserve a contiguous chunk of memory for kexec
  *
@@ -840,8 +860,8 @@ static int __init kho_bootmem_cmp(const void *_a, const void *_b)
  */
 static void __init kho_reserve_bootmem(void)
 {
-	int nid, i = 0, max_nr_bootmem;
-	phys_addr_t addr, size;
+	int nid, max_nr_bootmem;
+	phys_addr_t size;
 
 	if (!kho_enable)
 		return;
@@ -861,66 +881,40 @@ static void __init kho_reserve_bootmem(void)
 	 * reserve bootmem area in low memory for lowmem allocations in the
 	 * next kernel
 	 */
-	size = bootmem_size_lowmem;
-	addr = memblock_phys_alloc_range(size, BOOTMEM_ALIGNMENT_BYTES, 0,
-					 ARCH_LOW_ADDRESS_LIMIT);
-	if (!addr) {
+	if (kho_reserve_bootmem_area(bootmem_size_lowmem, ARCH_LOW_ADDRESS_LIMIT,
+				     NUMA_NO_NODE)) {
 		pr_err("Failed to reserve lowmem bootmem\n");
 		goto err_free_bootmem_desc;
 	}
 
-	kho_bootmem[i].addr = addr;
-	kho_bootmem[i].size = size;
-	i++;
-
 	/* reserve large contiguous area for allocations without nid */
-	size = bootmem_size_global;
-	addr = memblock_phys_alloc(size, BOOTMEM_ALIGNMENT_BYTES);
-	if (!addr) {
+	if (kho_reserve_bootmem_area(bootmem_size_global,
+				     MEMBLOCK_ALLOC_ACCESSIBLE, NUMA_NO_NODE)) {
 		pr_err("Failed to reserve global bootmem\n");
 		goto err_free_bootmem_areas;
 	}
 
-	kho_bootmem[i].addr = addr;
-	kho_bootmem[i].size = size;
-	i++;
-
 	/*
 	 * Loop over nodes that have both memory and are online. Skip
 	 * memoryless nodes, as we can not allocate bootmem areas there.
 	 */
 	for_each_node_state(nid, N_MEMORY) {
-		size = bootmem_size_node(nid);
-		if (!size)
-			continue;
-
-		addr = memblock_alloc_range_nid(size, BOOTMEM_ALIGNMENT_BYTES,
-						0, MEMBLOCK_ALLOC_ACCESSIBLE,
-						nid, true);
-		if (!addr) {
+		if (kho_reserve_bootmem_area(bootmem_size_node(nid),
+					     MEMBLOCK_ALLOC_ACCESSIBLE, nid)) {
 			pr_err("Failed to reserve nid %d bootmem\n", nid);
 			goto err_free_bootmem_areas;
 		}
-
-		kho_bootmem[i].addr = addr;
-		kho_bootmem[i].size = size;
-		i++;
 	}
 
-	/*
-	 * Nodes that needed no bootmem were skipped, so fewer areas than
-	 * initially estimated may have been reserved.
-	 */
-	kho_bootmem_cnt = i;
-
 	sort(kho_bootmem, kho_bootmem_cnt, sizeof(*kho_bootmem),
 	     kho_bootmem_cmp, NULL);
 
 	return;
 
 err_free_bootmem_areas:
-	for (i--; i >= 0; i--)
+	for (int i = kho_bootmem_cnt - 1; i >= 0; i--)
 		memblock_phys_free(kho_bootmem[i].addr, kho_bootmem[i].size);
+	kho_bootmem_cnt = 0;
 err_free_bootmem_desc:
 	memblock_free(kho_bootmem, max_nr_bootmem * sizeof(*kho_bootmem));
 err_disable_kho:

-- 
2.53.0

[PATCH v2 3/3] kho: fix bootmem sizing

[Mike Rapoport (Microsoft)]

Bootmem areas are sized from overlapping views of the memory reserved in
memblock:

* the global area is sized from all reserved memory, including memory
  allocated on specific nodes, so on NUMA systems everything allocated
  on a node is accounted in both the global area and the area of that
  node.
* per-node sizes are calculated after the lowmem and the global areas
  are reserved. With CONFIG_NUMA=n all memblock regions belong to node 0,
  so the lowmem and the global areas are accounted in the size of node 0
  area and scaled once again.

As a result, bootmem is about twice bigger than it should be on NUMA
systems and about four times bigger with CONFIG_NUMA=n. On small systems
reservation of the bootmem may fail and then KHO gets disabled.

Make the areas cover separate parts of the reserved memory: the lowmem
area covers allocations below ARCH_LOW_ADDRESS_LIMIT, per-node areas
cover allocations above it on each node and the global area covers
allocations that are not bound to any node. Calculate all sizes before
reserving any of the areas.

With CONFIG_NUMA=n the global area becomes empty and it is not reserved.

If ARCH_LOW_ADDRESS_LIMIT covers all memory, as may happen on arm64, any
area can serve lowmem allocations, so don't reserve the lowmem area.

Bootmem sizes in MiB in x86 QEMU VMs, before -> after:

                     CONFIG_NUMA=y     CONFIG_NUMA=n
  1G,  1 node         176 ->   92       320 ->   80
  4G,  1 node         508 ->  332      1264 ->  316
  16G, 2 nodes       1268 ->  704      2944 ->  736
  32G, 4 nodes       2296 -> 1224      5196 -> 1300
  64G, 4 nodes       4344 -> 2248      9664 -> 2420

Fixes: 3dc92c311498 ("kexec: add Kexec HandOver (KHO) generation helpers")
Reported-by: Kexin Liu <liukexin@kylinos.cn>
Reported-by: George Guo <guodongtai@kylinos.cn>
Closes: https://lore.kernel.org/all/20260904025101.9959-1-dongtai.guo@linux.dev
Reported-by: Sourabh Jain <sourabhjain@linux.ibm.com>
Closes: https://lore.kernel.org/all/20260922131217.698809-1-sourabhjain@linux.ibm.com
Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
---
 kernel/liveupdate/kexec_handover.c | 93 ++++++++++++++++++++++----------------
 1 file changed, 54 insertions(+), 39 deletions(-)

diff --git a/kernel/liveupdate/kexec_handover.c b/kernel/liveupdate/kexec_handover.c
index 5b1797d1289c8..c4b697a612f67 100644
--- a/kernel/liveupdate/kexec_handover.c
+++ b/kernel/liveupdate/kexec_handover.c
@@ -685,6 +685,7 @@ static unsigned int bootmem_scale __initdata = 200;
 static phys_addr_t bootmem_size_global __initdata;
 static phys_addr_t bootmem_size_pernode __initdata;
 static phys_addr_t bootmem_size_lowmem __initdata;
+static phys_addr_t bootmem_size_node[MAX_NUMNODES] __initdata;
 
 static int __init kho_parse_bootmem_size(char *p)
 {
@@ -754,29 +755,58 @@ static int __init kho_parse_bootmem_size(char *p)
 }
 early_param("kho_bootmem", kho_parse_bootmem_size);
 
+static phys_addr_t __init reserved_mem_size(phys_addr_t limit, int nid)
+{
+	return memblock_reserved_kern_size(limit, nid) -
+	       memblock_reserved_hugetlb_size(limit, nid);
+}
+
+static phys_addr_t __init bootmem_scale_size(phys_addr_t size)
+{
+	return size * bootmem_scale / 100;
+}
+
+/*
+ * The lowmem area covers allocations below ARCH_LOW_ADDRESS_LIMIT, per-node
+ * areas cover allocations above it on each node and the global area covers
+ * allocations that are not bound to any node.
+ *
+ * The reserved memory on a node and the total reserved memory include
+ * lowmem, so it is subtracted from both.
+ *
+ * If ARCH_LOW_ADDRESS_LIMIT covers all memory, any area can serve lowmem
+ * allocations and there is no lowmem area.
+ *
+ * All sizes must be calculated before any area is reserved, otherwise the
+ * areas reserved first will be accounted in the sizes of the others.
+ */
 static void __init bootmem_size_update(void)
 {
-	/*
-	 * If fixed sizes are not provided via command line, calculate them now.
-	 * Remove HugeTLB allocations from it because they never get allocated
-	 * from bootmem.
-	 */
+	int nid;
+
 	if (bootmem_scale) {
-		phys_addr_t size;
-
-		size = memblock_reserved_kern_size(ARCH_LOW_ADDRESS_LIMIT,
-						   NUMA_NO_NODE);
-		size -= memblock_reserved_hugetlb_size(ARCH_LOW_ADDRESS_LIMIT,
-						       NUMA_NO_NODE);
-		size = size * bootmem_scale / 100;
-		bootmem_size_lowmem = size;
-
-		size = memblock_reserved_kern_size(MEMBLOCK_ALLOC_ANYWHERE,
-						   NUMA_NO_NODE);
-		size -= memblock_reserved_hugetlb_size(MEMBLOCK_ALLOC_ANYWHERE,
-						       NUMA_NO_NODE);
-		size = size * bootmem_scale / 100 - bootmem_size_lowmem;
-		bootmem_size_global = size;
+		phys_addr_t low_limit = ARCH_LOW_ADDRESS_LIMIT;
+		phys_addr_t lowmem, global, size;
+
+		if (low_limit >= memblock_end_of_DRAM() - 1)
+			low_limit = 0;
+
+		lowmem = reserved_mem_size(low_limit, NUMA_NO_NODE);
+		global = reserved_mem_size(MEMBLOCK_ALLOC_ANYWHERE, NUMA_NO_NODE);
+		global -= lowmem;
+
+		for_each_node_state(nid, N_MEMORY) {
+			size = reserved_mem_size(MEMBLOCK_ALLOC_ANYWHERE, nid) -
+			       reserved_mem_size(low_limit, nid);
+			bootmem_size_node[nid] = bootmem_scale_size(size);
+			global -= size;
+		}
+
+		bootmem_size_lowmem = bootmem_scale_size(lowmem);
+		bootmem_size_global = bootmem_scale_size(global);
+	} else {
+		for_each_node_state(nid, N_MEMORY)
+			bootmem_size_node[nid] = bootmem_size_pernode;
 	}
 
 	/*
@@ -785,24 +815,9 @@ static void __init bootmem_size_update(void)
 	 */
 	bootmem_size_lowmem = round_up(bootmem_size_lowmem, BOOTMEM_ALIGNMENT_BYTES);
 	bootmem_size_global = round_up(bootmem_size_global, BOOTMEM_ALIGNMENT_BYTES);
-}
-
-static phys_addr_t __init bootmem_size_node(int nid)
-{
-	phys_addr_t size;
-
-	if (bootmem_scale) {
-		size = memblock_reserved_kern_size(MEMBLOCK_ALLOC_ANYWHERE,
-						   nid);
-		/* Do not count HugeTLB pages. */
-		size -= memblock_reserved_hugetlb_size(MEMBLOCK_ALLOC_ANYWHERE,
-						       nid);
-		size = size * bootmem_scale / 100;
-	} else {
-		size = bootmem_size_pernode;
-	}
-
-	return round_up(size, BOOTMEM_ALIGNMENT_BYTES);
+	for_each_node_state(nid, N_MEMORY)
+		bootmem_size_node[nid] = round_up(bootmem_size_node[nid],
+						  BOOTMEM_ALIGNMENT_BYTES);
 }
 
 bool kho_bootmem_overlap(phys_addr_t phys, size_t size)
@@ -899,7 +914,7 @@ static void __init kho_reserve_bootmem(void)
 	 * memoryless nodes, as we can not allocate bootmem areas there.
 	 */
 	for_each_node_state(nid, N_MEMORY) {
-		if (kho_reserve_bootmem_area(bootmem_size_node(nid),
+		if (kho_reserve_bootmem_area(bootmem_size_node[nid],
 					     MEMBLOCK_ALLOC_ACCESSIBLE, nid)) {
 			pr_err("Failed to reserve nid %d bootmem\n", nid);
 			goto err_free_bootmem_areas;

-- 
2.53.0

Re: [PATCH v2 1/3] kho: walk bootmem areas in address order

[Pasha Tatashin]

On Wed, 04 Nov 2026 18:50:09 +0200, Mike Rapoport (Microsoft) <rppt@kernel.org> wrote:
> kexec_add_buffer() users may depend on the order in which memory holes are
> searched. For instance, arm64 places the kernel bottom-up and then
> requires the DTB to be above the kernel.
> 
> kho_walk_bootmem() walks the bootmem areas in the order they were
> reserved, so if the kernel lands in a small area at the top of the memory
> there is no room left for the DTB and kexec fails. For example, on an
> arm64 VM with 2G of RAM and kho_bootmem=8M,64M,64M, the lowmem area is
> reserved first at the top of the memory, the kernel image takes most of
> it and kexec fails with
> 
> [...]

Reviewed-by: Pasha Tatashin <pasha.tatashin@soleen.com>

-- 
Pasha

Re: [PATCH v2 2/3] kho: skip zero-sized lowmem and global bootmem areas

[Pasha Tatashin]

On Wed, 04 Nov 2026 18:50:10 +0200, Mike Rapoport (Microsoft) <rppt@kernel.org> wrote:
> The lowmem and the global bootmem areas can have zero size, for example
> when none or all of the memory is below ARCH_LOW_ADDRESS_LIMIT, or when
> either of those areas is explicitly set to zero with kho_bootmem=.
> 
> For a zero-sized allocation memblock returns an address without reserving
> anything, and that empty area is recorded in kho_bootmem. If there is no
> memory below ARCH_LOW_ADDRESS_LIMIT, allocation of the empty lowmem area
> fails and KHO gets disabled.
> 
> [...]

Reviewed-by: Pasha Tatashin <pasha.tatashin@soleen.com>

-- 
Pasha

Re: [PATCH v2 3/3] kho: fix bootmem sizing

[Pasha Tatashin]

> Bootmem areas are sized from overlapping views of the memory reserved in
> memblock:
> 
> * the global area is sized from all reserved memory, including memory
>   allocated on specific nodes, so on NUMA systems everything allocated
>   on a node is accounted in both the global area and the area of that
>   node.
> * per-node sizes are calculated after the lowmem and the global areas
>   are reserved. With CONFIG_NUMA=n all memblock regions belong to node 0,
>   so the lowmem and the global areas are accounted in the size of node 0
>   area and scaled once again.
> 
> As a result, bootmem is about twice bigger than it should be on NUMA
> systems and about four times bigger with CONFIG_NUMA=n. On small systems
> reservation of the bootmem may fail and then KHO gets disabled.
> 
> Make the areas cover separate parts of the reserved memory: the lowmem
> area covers allocations below ARCH_LOW_ADDRESS_LIMIT, per-node areas
> cover allocations above it on each node and the global area covers
> allocations that are not bound to any node. Calculate all sizes before
> reserving any of the areas.
> 
> With CONFIG_NUMA=n the global area becomes empty and it is not reserved.
> 
> If ARCH_LOW_ADDRESS_LIMIT covers all memory, as may happen on arm64, any
> area can serve lowmem allocations, so don't reserve the lowmem area.
> 
> Bootmem sizes in MiB in x86 QEMU VMs, before -> after:
> 
>                      CONFIG_NUMA=y     CONFIG_NUMA=n
>   1G,  1 node         176 ->   92       320 ->   80
>   4G,  1 node         508 ->  332      1264 ->  316
>   16G, 2 nodes       1268 ->  704      2944 ->  736
>   32G, 4 nodes       2296 -> 1224      5196 -> 1300
>   64G, 4 nodes       4344 -> 2248      9664 -> 2420
> 
> Fixes: 3dc92c311498 ("kexec: add Kexec HandOver (KHO) generation helpers")
> Reported-by: Kexin Liu <liukexin@kylinos.cn>
> Reported-by: George Guo <guodongtai@kylinos.cn>
> Closes: https://lore.kernel.org/all/20260904025101.9959-1-dongtai.guo@linux.dev
> Reported-by: Sourabh Jain <sourabhjain@linux.ibm.com>
> Closes: https://lore.kernel.org/all/20260922131217.698809-1-sourabhjain@linux.ibm.com
> Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
>
> diff --git a/kernel/liveupdate/kexec_handover.c b/kernel/liveupdate/kexec_handover.c
> index 5b1797d1289c..c4b697a612f6 100644
> --- a/kernel/liveupdate/kexec_handover.c
> +++ b/kernel/liveupdate/kexec_handover.c
> @@ -685,6 +685,7 @@ static unsigned int bootmem_scale __initdata = 200;
>  static phys_addr_t bootmem_size_global __initdata;
>  static phys_addr_t bootmem_size_pernode __initdata;
>  static phys_addr_t bootmem_size_lowmem __initdata;
> +static phys_addr_t bootmem_size_node[MAX_NUMNODES] __initdata;
>  
>  static int __init kho_parse_bootmem_size(char *p)
>  {
> @@ -754,29 +755,58 @@ static int __init kho_parse_bootmem_size(char *p)
>  }
>  early_param("kho_bootmem", kho_parse_bootmem_size);
>  
> +static phys_addr_t __init reserved_mem_size(phys_addr_t limit, int nid)
> +{
> +	return memblock_reserved_kern_size(limit, nid) -
> +	       memblock_reserved_hugetlb_size(limit, nid);
> +}
> +
> +static phys_addr_t __init bootmem_scale_size(phys_addr_t size)
> +{
> +	return size * bootmem_scale / 100;
> +}
> +
> +/*
> + * The lowmem area covers allocations below ARCH_LOW_ADDRESS_LIMIT, per-node
> + * areas cover allocations above it on each node and the global area covers
> + * allocations that are not bound to any node.
> + *
> + * The reserved memory on a node and the total reserved memory include
> + * lowmem, so it is subtracted from both.
> + *
> + * If ARCH_LOW_ADDRESS_LIMIT covers all memory, any area can serve lowmem
> + * allocations and there is no lowmem area.
> + *
> + * All sizes must be calculated before any area is reserved, otherwise the
> + * areas reserved first will be accounted in the sizes of the others.
> + */
>  static void __init bootmem_size_update(void)
>  {
> -	/*
> -	 * If fixed sizes are not provided via command line, calculate them now.
> -	 * Remove HugeTLB allocations from it because they never get allocated
> -	 * from bootmem.
> -	 */
> +	int nid;
> +
>  	if (bootmem_scale) {
> -		phys_addr_t size;
> -
> -		size = memblock_reserved_kern_size(ARCH_LOW_ADDRESS_LIMIT,
> -						   NUMA_NO_NODE);
> -		size -= memblock_reserved_hugetlb_size(ARCH_LOW_ADDRESS_LIMIT,
> -						       NUMA_NO_NODE);
> -		size = size * bootmem_scale / 100;
> -		bootmem_size_lowmem = size;
> -
> -		size = memblock_reserved_kern_size(MEMBLOCK_ALLOC_ANYWHERE,
> -						   NUMA_NO_NODE);
> -		size -= memblock_reserved_hugetlb_size(MEMBLOCK_ALLOC_ANYWHERE,
> -						       NUMA_NO_NODE);
> -		size = size * bootmem_scale / 100 - bootmem_size_lowmem;
> -		bootmem_size_global = size;
> +		phys_addr_t low_limit = ARCH_LOW_ADDRESS_LIMIT;
> +		phys_addr_t lowmem, global, size;
> +
> +		if (low_limit >= memblock_end_of_DRAM() - 1)
> +			low_limit = 0;
> +
> +		lowmem = reserved_mem_size(low_limit, NUMA_NO_NODE);
> +		global = reserved_mem_size(MEMBLOCK_ALLOC_ANYWHERE, NUMA_NO_NODE);
> +		global -= lowmem;
> +
> +		for_each_node_state(nid, N_MEMORY) {
> +			size = reserved_mem_size(MEMBLOCK_ALLOC_ANYWHERE, nid) -
> +			       reserved_mem_size(low_limit, nid);
> +			bootmem_size_node[nid] = bootmem_scale_size(size);
> +			global -= size;
> +		}
> +
> +		bootmem_size_lowmem = bootmem_scale_size(lowmem);
> +		bootmem_size_global = bootmem_scale_size(global);
> +	} else {
> +		for_each_node_state(nid, N_MEMORY)
> +			bootmem_size_node[nid] = bootmem_size_pernode;
>  	}
>  
>  	/*
> @@ -785,24 +815,9 @@ static void __init bootmem_size_update(void)
>  	 */
>  	bootmem_size_lowmem = round_up(bootmem_size_lowmem, BOOTMEM_ALIGNMENT_BYTES);
>  	bootmem_size_global = round_up(bootmem_size_global, BOOTMEM_ALIGNMENT_BYTES);
> -}
> -
> -static phys_addr_t __init bootmem_size_node(int nid)
> -{
> -	phys_addr_t size;
> -
> -	if (bootmem_scale) {
> -		size = memblock_reserved_kern_size(MEMBLOCK_ALLOC_ANYWHERE,
> -						   nid);
> -		/* Do not count HugeTLB pages. */
> -		size -= memblock_reserved_hugetlb_size(MEMBLOCK_ALLOC_ANYWHERE,
> -						       nid);
> -		size = size * bootmem_scale / 100;
> -	} else {
> -		size = bootmem_size_pernode;
> -	}
> -
> -	return round_up(size, BOOTMEM_ALIGNMENT_BYTES);
> +	for_each_node_state(nid, N_MEMORY)
> +		bootmem_size_node[nid] = round_up(bootmem_size_node[nid],
> +						  BOOTMEM_ALIGNMENT_BYTES);

Nit, but I prefer to avoid multi-line indented statements without 
braces. Just put a braces around the above.

Reviewed-by: Pasha Tatashin <pasha.tatashin@soleen.com>

-- 
Pasha

Re: [PATCH v2 1/3] kho: walk bootmem areas in address order

[Jose A. Perez de Azpillaga]

On 11/4/26 5:50 PM, Mike Rapoport (Microsoft) wrote:
> kexec_add_buffer() users may depend on the order in which memory holes are
> searched. For instance, arm64 places the kernel bottom-up and then
> requires the DTB to be above the kernel.
> 
> kho_walk_bootmem() walks the bootmem areas in the order they were
> reserved, so if the kernel lands in a small area at the top of the memory
> there is no room left for the DTB and kexec fails. For example, on an
> arm64 VM with 2G of RAM and kho_bootmem=8M,64M,64M, the lowmem area is
> reserved first at the top of the memory, the kernel image takes most of
> it and kexec fails with
> 
>   kexec_file(Image): Could not find any suitable kernel location!
> 
> Sort the bootmem areas by address and walk them in ascending order, or in
> descending order for top-down searches, like kexec_walk_resources() does.
> 
> Fixes: 3bdecc3c93f9 ("kexec: add KHO support to kexec file loads")
> Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
Reviewed-by: Jose A. Perez de Azpillaga <azpijr@gmail.com>

-- 
cheers,
jose a. p-a

Re: [PATCH v2 2/3] kho: skip zero-sized lowmem and global bootmem areas

[Jose A. Perez de Azpillaga]

On 11/4/26 5:50 PM, Mike Rapoport (Microsoft) wrote:
> The lowmem and the global bootmem areas can have zero size, for example
> when none or all of the memory is below ARCH_LOW_ADDRESS_LIMIT, or when
> either of those areas is explicitly set to zero with kho_bootmem=.
> 
> For a zero-sized allocation memblock returns an address without reserving
> anything, and that empty area is recorded in kho_bootmem. If there is no
> memory below ARCH_LOW_ADDRESS_LIMIT, allocation of the empty lowmem area
> fails and KHO gets disabled.
> 
> Extract a helper that checks for zero size and allocates memory from
> memblock, and use it to skip empty areas the same way as empty per-node
> areas are skipped.
> 
> Fixes: 3dc92c311498 ("kexec: add Kexec HandOver (KHO) generation helpers")
> Reported-by: Chen Pei <cp0613@linux.alibaba.com>
> Closes: https://lore.kernel.org/all/20261026131851.9942-1-cp0613@linux.alibaba.com
> Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
Reviewed-by: Jose A. Perez de Azpillaga <azpijr@gmail.com>

-- 
cheers,
jose a. p-a

Re: [PATCH v2 3/3] kho: fix bootmem sizing

[Jose A. Perez de Azpillaga]

On 11/4/26 5:50 PM, Mike Rapoport (Microsoft) wrote:
> Bootmem areas are sized from overlapping views of the memory reserved in
> memblock:
> 
> * the global area is sized from all reserved memory, including memory
>   allocated on specific nodes, so on NUMA systems everything allocated
>   on a node is accounted in both the global area and the area of that
>   node.
> * per-node sizes are calculated after the lowmem and the global areas
>   are reserved. With CONFIG_NUMA=n all memblock regions belong to node 0,
>   so the lowmem and the global areas are accounted in the size of node 0
>   area and scaled once again.
> 
> As a result, bootmem is about twice bigger than it should be on NUMA
> systems and about four times bigger with CONFIG_NUMA=n. On small systems
> reservation of the bootmem may fail and then KHO gets disabled.
> 
> Make the areas cover separate parts of the reserved memory: the lowmem
> area covers allocations below ARCH_LOW_ADDRESS_LIMIT, per-node areas
> cover allocations above it on each node and the global area covers
> allocations that are not bound to any node. Calculate all sizes before
> reserving any of the areas.
> 
> With CONFIG_NUMA=n the global area becomes empty and it is not reserved.
> 
> If ARCH_LOW_ADDRESS_LIMIT covers all memory, as may happen on arm64, any
> area can serve lowmem allocations, so don't reserve the lowmem area.
> 
> Bootmem sizes in MiB in x86 QEMU VMs, before -> after:
> 
>                      CONFIG_NUMA=y     CONFIG_NUMA=n
>   1G,  1 node         176 ->   92       320 ->   80
>   4G,  1 node         508 ->  332      1264 ->  316
>   16G, 2 nodes       1268 ->  704      2944 ->  736
>   32G, 4 nodes       2296 -> 1224      5196 -> 1300
>   64G, 4 nodes       4344 -> 2248      9664 -> 2420
> 
> Fixes: 3dc92c311498 ("kexec: add Kexec HandOver (KHO) generation helpers")
> Reported-by: Kexin Liu <liukexin@kylinos.cn>
> Reported-by: George Guo <guodongtai@kylinos.cn>
> Closes: https://lore.kernel.org/all/20260904025101.9959-1-dongtai.guo@linux.dev
> Reported-by: Sourabh Jain <sourabhjain@linux.ibm.com>
> Closes: https://lore.kernel.org/all/20260922131217.698809-1-sourabhjain@linux.ibm.com
> Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
Reviewed-by: Jose A. Perez de Azpillaga <azpijr@gmail.com>

-- 
cheers,
jose a. p-a

Re: [PATCH v2 3/3] kho: fix bootmem sizing

[Mike Rapoport]

On Wed, Nov 04, 2026 at 01:57:03PM -0500, Pasha Tatashin wrote:
> > -
> > -	return round_up(size, BOOTMEM_ALIGNMENT_BYTES);
> > +	for_each_node_state(nid, N_MEMORY)
> > +		bootmem_size_node[nid] = round_up(bootmem_size_node[nid],
> > +						  BOOTMEM_ALIGNMENT_BYTES);
> 
> Nit, but I prefer to avoid multi-line indented statements without 
> braces. Just put a braces around the above.

Yep, I'll fix it up while applying.
 
> Reviewed-by: Pasha Tatashin <pasha.tatashin@soleen.com>
> 
> -- 
> Pasha
> 

-- 
Sincerely yours,
Mike.