Sanity check that makes sure the nodes cover all memory loops over
numa_meminfo to count the pages that have node id assigned by the firmware,
then loops again over memblock.memory to find the total amount of memory
and in the end checks that the difference between the total memory and
memory that covered by nodes is less than some threshold. Worse, the loop
over numa_meminfo calls __absent_pages_in_range() that also partially
traverses memblock.memory.
It's much simpler and more efficient to have a single traversal of
memblock.memory that verifies that amount of memory not covered by nodes is
less than a threshold.
Introduce memblock_validate_numa_coverage() that does exactly that and use
it instead of numa_meminfo_cover_memory().
Signed-off-by: Liam Ni <[email protected]>
---
arch/loongarch/kernel/numa.c | 28 +---------------------------
arch/x86/mm/numa.c | 34 ++--------------------------------
include/linux/memblock.h | 1 +
mm/memblock.c | 34 ++++++++++++++++++++++++++++++++++
4 files changed, 38 insertions(+), 59 deletions(-)
diff --git a/arch/loongarch/kernel/numa.c b/arch/loongarch/kernel/numa.c
index cb00804826f7..fca94d16be34 100644
--- a/arch/loongarch/kernel/numa.c
+++ b/arch/loongarch/kernel/numa.c
@@ -226,32 +226,6 @@ static void __init node_mem_init(unsigned int node)
#ifdef CONFIG_ACPI_NUMA
-/*
- * Sanity check to catch more bad NUMA configurations (they are amazingly
- * common). Make sure the nodes cover all memory.
- */
-static bool __init numa_meminfo_cover_memory(const struct numa_meminfo *mi)
-{
- int i;
- u64 numaram, biosram;
-
- numaram = 0;
- for (i = 0; i < mi->nr_blks; i++) {
- u64 s = mi->blk[i].start >> PAGE_SHIFT;
- u64 e = mi->blk[i].end >> PAGE_SHIFT;
-
- numaram += e - s;
- numaram -= __absent_pages_in_range(mi->blk[i].nid, s, e);
- if ((s64)numaram < 0)
- numaram = 0;
- }
- max_pfn = max_low_pfn;
- biosram = max_pfn - absent_pages_in_range(0, max_pfn);
-
- BUG_ON((s64)(biosram - numaram) >= (1 << (20 - PAGE_SHIFT)));
- return true;
-}
-
static void __init add_node_intersection(u32 node, u64 start, u64 size, u32 type)
{
static unsigned long num_physpages;
@@ -396,7 +370,7 @@ int __init init_numa_memory(void)
return -EINVAL;
init_node_memblock();
- if (numa_meminfo_cover_memory(&numa_meminfo) == false)
+ if (memblock_validate_numa_coverage(SZ_1M >> 12) == false)
return -EINVAL;
for_each_node_mask(node, node_possible_map) {
diff --git a/arch/x86/mm/numa.c b/arch/x86/mm/numa.c
index 2aadb2019b4f..95376e7c263e 100644
--- a/arch/x86/mm/numa.c
+++ b/arch/x86/mm/numa.c
@@ -447,37 +447,6 @@ int __node_distance(int from, int to)
}
EXPORT_SYMBOL(__node_distance);
-/*
- * Sanity check to catch more bad NUMA configurations (they are amazingly
- * common). Make sure the nodes cover all memory.
- */
-static bool __init numa_meminfo_cover_memory(const struct numa_meminfo *mi)
-{
- u64 numaram, e820ram;
- int i;
-
- numaram = 0;
- for (i = 0; i < mi->nr_blks; i++) {
- u64 s = mi->blk[i].start >> PAGE_SHIFT;
- u64 e = mi->blk[i].end >> PAGE_SHIFT;
- numaram += e - s;
- numaram -= __absent_pages_in_range(mi->blk[i].nid, s, e);
- if ((s64)numaram < 0)
- numaram = 0;
- }
-
- e820ram = max_pfn - absent_pages_in_range(0, max_pfn);
-
- /* We seem to lose 3 pages somewhere. Allow 1M of slack. */
- if ((s64)(e820ram - numaram) >= (1 << (20 - PAGE_SHIFT))) {
- printk(KERN_ERR "NUMA: nodes only cover %LuMB of your %LuMB e820 RAM. Not used.\n",
- (numaram << PAGE_SHIFT) >> 20,
- (e820ram << PAGE_SHIFT) >> 20);
- return false;
- }
- return true;
-}
-
/*
* Mark all currently memblock-reserved physical memory (which covers the
* kernel's own memory ranges) as hot-unswappable.
@@ -583,7 +552,8 @@ static int __init numa_register_memblks(struct numa_meminfo *mi)
return -EINVAL;
}
}
- if (!numa_meminfo_cover_memory(mi))
+
+ if (!memblock_validate_numa_coverage(SZ_1M >> 12))
return -EINVAL;
/* Finally register nodes. */
diff --git a/include/linux/memblock.h b/include/linux/memblock.h
index 1c1072e3ca06..727242f4b54a 100644
--- a/include/linux/memblock.h
+++ b/include/linux/memblock.h
@@ -120,6 +120,7 @@ int memblock_physmem_add(phys_addr_t base, phys_addr_t size);
void memblock_trim_memory(phys_addr_t align);
bool memblock_overlaps_region(struct memblock_type *type,
phys_addr_t base, phys_addr_t size);
+bool memblock_validate_numa_coverage(const u64 threshold_pages);
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);
diff --git a/mm/memblock.c b/mm/memblock.c
index 0863222af4a4..4f1f2d8a8119 100644
--- a/mm/memblock.c
+++ b/mm/memblock.c
@@ -734,6 +734,40 @@ int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size)
return memblock_add_range(&memblock.memory, base, size, MAX_NUMNODES, 0);
}
+/**
+ * memblock_validate_numa_coverage - calculating memory with no node id assigned by firmware
+ * @threshold_pages: threshold memory of no node id assigned
+ *
+ * calculating memory with no node id assigned by firmware,
+ * If the number is less than the @threshold_pages, it returns true,
+ * otherwise it returns false.
+ *
+ * Return:
+ * true on success, false on failure.
+ */
+bool __init_memblock memblock_validate_numa_coverage(const u64 threshold_pages)
+{
+ unsigned long nr_pages = 0;
+ unsigned long start_pfn, end_pfn, mem_size_mb;
+ int nid, i;
+
+ /* calculate lose page */
+ for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn, &nid) {
+ if (nid == NUMA_NO_NODE)
+ nr_pages += end_pfn - start_pfn;
+ }
+
+ if (nr_pages >= threshold_pages) {
+ mem_size_mb = memblock_phys_mem_size() >> 20;
+ pr_err("NUMA: no nodes coverage for %luMB of %luMB RAM\n",
+ (nr_pages << PAGE_SHIFT) >> 20, mem_size_mb);
+ return false;
+ }
+
+ return true;
+}
+
+
/**
* memblock_isolate_range - isolate given range into disjoint memblocks
* @type: memblock type to isolate range for
--
2.25.1
On Tue, Oct 17, 2023 at 04:30:33PM +0800, Liam Ni wrote:
> Sanity check that makes sure the nodes cover all memory loops over
> numa_meminfo to count the pages that have node id assigned by the firmware,
> then loops again over memblock.memory to find the total amount of memory
> and in the end checks that the difference between the total memory and
> memory that covered by nodes is less than some threshold. Worse, the loop
> over numa_meminfo calls __absent_pages_in_range() that also partially
> traverses memblock.memory.
>
> It's much simpler and more efficient to have a single traversal of
> memblock.memory that verifies that amount of memory not covered by nodes is
> less than a threshold.
>
> Introduce memblock_validate_numa_coverage() that does exactly that and use
> it instead of numa_meminfo_cover_memory().
>
> Signed-off-by: Liam Ni <[email protected]>
> ---
> arch/loongarch/kernel/numa.c | 28 +---------------------------
> arch/x86/mm/numa.c | 34 ++--------------------------------
> include/linux/memblock.h | 1 +
> mm/memblock.c | 34 ++++++++++++++++++++++++++++++++++
> 4 files changed, 38 insertions(+), 59 deletions(-)
>
> diff --git a/arch/loongarch/kernel/numa.c b/arch/loongarch/kernel/numa.c
> index cb00804826f7..fca94d16be34 100644
> --- a/arch/loongarch/kernel/numa.c
> +++ b/arch/loongarch/kernel/numa.c
> @@ -226,32 +226,6 @@ static void __init node_mem_init(unsigned int node)
>
> #ifdef CONFIG_ACPI_NUMA
>
> -/*
> - * Sanity check to catch more bad NUMA configurations (they are amazingly
> - * common). Make sure the nodes cover all memory.
> - */
> -static bool __init numa_meminfo_cover_memory(const struct numa_meminfo *mi)
> -{
> - int i;
> - u64 numaram, biosram;
> -
> - numaram = 0;
> - for (i = 0; i < mi->nr_blks; i++) {
> - u64 s = mi->blk[i].start >> PAGE_SHIFT;
> - u64 e = mi->blk[i].end >> PAGE_SHIFT;
> -
> - numaram += e - s;
> - numaram -= __absent_pages_in_range(mi->blk[i].nid, s, e);
> - if ((s64)numaram < 0)
> - numaram = 0;
> - }
> - max_pfn = max_low_pfn;
> - biosram = max_pfn - absent_pages_in_range(0, max_pfn);
> -
> - BUG_ON((s64)(biosram - numaram) >= (1 << (20 - PAGE_SHIFT)));
> - return true;
> -}
> -
> static void __init add_node_intersection(u32 node, u64 start, u64 size, u32 type)
> {
> static unsigned long num_physpages;
> @@ -396,7 +370,7 @@ int __init init_numa_memory(void)
> return -EINVAL;
>
> init_node_memblock();
> - if (numa_meminfo_cover_memory(&numa_meminfo) == false)
> + if (memblock_validate_numa_coverage(SZ_1M >> 12) == false)
No magic constants please.
Either use
SZ_1M >> PAGE_SIZE
here, or make threshold in bytes and convert it to number of pages in
memblock_validate_numa_coverage().
Besides, no need to compare to false,
if (!memblock_validate_numa_coverage())
will do
> return -EINVAL;
>
> for_each_node_mask(node, node_possible_map) {
> diff --git a/arch/x86/mm/numa.c b/arch/x86/mm/numa.c
> index 2aadb2019b4f..95376e7c263e 100644
> --- a/arch/x86/mm/numa.c
> +++ b/arch/x86/mm/numa.c
> @@ -447,37 +447,6 @@ int __node_distance(int from, int to)
> }
> EXPORT_SYMBOL(__node_distance);
>
> -/*
> - * Sanity check to catch more bad NUMA configurations (they are amazingly
> - * common). Make sure the nodes cover all memory.
> - */
> -static bool __init numa_meminfo_cover_memory(const struct numa_meminfo *mi)
> -{
> - u64 numaram, e820ram;
> - int i;
> -
> - numaram = 0;
> - for (i = 0; i < mi->nr_blks; i++) {
> - u64 s = mi->blk[i].start >> PAGE_SHIFT;
> - u64 e = mi->blk[i].end >> PAGE_SHIFT;
> - numaram += e - s;
> - numaram -= __absent_pages_in_range(mi->blk[i].nid, s, e);
> - if ((s64)numaram < 0)
> - numaram = 0;
> - }
> -
> - e820ram = max_pfn - absent_pages_in_range(0, max_pfn);
> -
> - /* We seem to lose 3 pages somewhere. Allow 1M of slack. */
> - if ((s64)(e820ram - numaram) >= (1 << (20 - PAGE_SHIFT))) {
> - printk(KERN_ERR "NUMA: nodes only cover %LuMB of your %LuMB e820 RAM. Not used.\n",
> - (numaram << PAGE_SHIFT) >> 20,
> - (e820ram << PAGE_SHIFT) >> 20);
> - return false;
> - }
> - return true;
> -}
> -
> /*
> * Mark all currently memblock-reserved physical memory (which covers the
> * kernel's own memory ranges) as hot-unswappable.
> @@ -583,7 +552,8 @@ static int __init numa_register_memblks(struct numa_meminfo *mi)
> return -EINVAL;
> }
> }
> - if (!numa_meminfo_cover_memory(mi))
> +
> + if (!memblock_validate_numa_coverage(SZ_1M >> 12))
> return -EINVAL;
>
> /* Finally register nodes. */
> diff --git a/include/linux/memblock.h b/include/linux/memblock.h
> index 1c1072e3ca06..727242f4b54a 100644
> --- a/include/linux/memblock.h
> +++ b/include/linux/memblock.h
> @@ -120,6 +120,7 @@ int memblock_physmem_add(phys_addr_t base, phys_addr_t size);
> void memblock_trim_memory(phys_addr_t align);
> bool memblock_overlaps_region(struct memblock_type *type,
> phys_addr_t base, phys_addr_t size);
> +bool memblock_validate_numa_coverage(const u64 threshold_pages);
> 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);
> diff --git a/mm/memblock.c b/mm/memblock.c
> index 0863222af4a4..4f1f2d8a8119 100644
> --- a/mm/memblock.c
> +++ b/mm/memblock.c
> @@ -734,6 +734,40 @@ int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size)
> return memblock_add_range(&memblock.memory, base, size, MAX_NUMNODES, 0);
> }
>
> +/**
> + * memblock_validate_numa_coverage - calculating memory with no node id assigned by firmware
> + * @threshold_pages: threshold memory of no node id assigned
> + *
> + * calculating memory with no node id assigned by firmware,
> + * If the number is less than the @threshold_pages, it returns true,
> + * otherwise it returns false.
> + *
> + * Return:
> + * true on success, false on failure.
> + */
I'd suggest the below version:
/**
* memblock_validate_numa_coverage - check if amount of memory with
* no node ID assigned is less than a threshold
* @threshold_pages: maximal number of pages that can have unassigned node
* ID (in pages).
*
* A buggy firmware may report memory that does not belong to any node.
* Check if amount of such memory is below @threshold_pages.
*
* Return: true on success, false on failure.
*/
> +bool __init_memblock memblock_validate_numa_coverage(const u64 threshold_pages)
> +{
> + unsigned long nr_pages = 0;
> + unsigned long start_pfn, end_pfn, mem_size_mb;
> + int nid, i;
> +
> + /* calculate lose page */
> + for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn, &nid) {
> + if (nid == NUMA_NO_NODE)
> + nr_pages += end_pfn - start_pfn;
> + }
> +
> + if (nr_pages >= threshold_pages) {
> + mem_size_mb = memblock_phys_mem_size() >> 20;
> + pr_err("NUMA: no nodes coverage for %luMB of %luMB RAM\n",
> + (nr_pages << PAGE_SHIFT) >> 20, mem_size_mb);
> + return false;
> + }
> +
> + return true;
> +}
> +
> +
> /**
> * memblock_isolate_range - isolate given range into disjoint memblocks
> * @type: memblock type to isolate range for
> --
> 2.25.1
>
--
Sincerely yours,
Mike.
Thanks, I will prepare V6 based on your suggestion.
On Thu, 19 Oct 2023 at 15:36, Mike Rapoport <[email protected]> wrote:
>
> On Tue, Oct 17, 2023 at 04:30:33PM +0800, Liam Ni wrote:
> > Sanity check that makes sure the nodes cover all memory loops over
> > numa_meminfo to count the pages that have node id assigned by the firmware,
> > then loops again over memblock.memory to find the total amount of memory
> > and in the end checks that the difference between the total memory and
> > memory that covered by nodes is less than some threshold. Worse, the loop
> > over numa_meminfo calls __absent_pages_in_range() that also partially
> > traverses memblock.memory.
> >
> > It's much simpler and more efficient to have a single traversal of
> > memblock.memory that verifies that amount of memory not covered by nodes is
> > less than a threshold.
> >
> > Introduce memblock_validate_numa_coverage() that does exactly that and use
> > it instead of numa_meminfo_cover_memory().
> >
> > Signed-off-by: Liam Ni <[email protected]>
> > ---
> > arch/loongarch/kernel/numa.c | 28 +---------------------------
> > arch/x86/mm/numa.c | 34 ++--------------------------------
> > include/linux/memblock.h | 1 +
> > mm/memblock.c | 34 ++++++++++++++++++++++++++++++++++
> > 4 files changed, 38 insertions(+), 59 deletions(-)
> >
> > diff --git a/arch/loongarch/kernel/numa.c b/arch/loongarch/kernel/numa.c
> > index cb00804826f7..fca94d16be34 100644
> > --- a/arch/loongarch/kernel/numa.c
> > +++ b/arch/loongarch/kernel/numa.c
> > @@ -226,32 +226,6 @@ static void __init node_mem_init(unsigned int node)
> >
> > #ifdef CONFIG_ACPI_NUMA
> >
> > -/*
> > - * Sanity check to catch more bad NUMA configurations (they are amazingly
> > - * common). Make sure the nodes cover all memory.
> > - */
> > -static bool __init numa_meminfo_cover_memory(const struct numa_meminfo *mi)
> > -{
> > - int i;
> > - u64 numaram, biosram;
> > -
> > - numaram = 0;
> > - for (i = 0; i < mi->nr_blks; i++) {
> > - u64 s = mi->blk[i].start >> PAGE_SHIFT;
> > - u64 e = mi->blk[i].end >> PAGE_SHIFT;
> > -
> > - numaram += e - s;
> > - numaram -= __absent_pages_in_range(mi->blk[i].nid, s, e);
> > - if ((s64)numaram < 0)
> > - numaram = 0;
> > - }
> > - max_pfn = max_low_pfn;
> > - biosram = max_pfn - absent_pages_in_range(0, max_pfn);
> > -
> > - BUG_ON((s64)(biosram - numaram) >= (1 << (20 - PAGE_SHIFT)));
> > - return true;
> > -}
> > -
> > static void __init add_node_intersection(u32 node, u64 start, u64 size, u32 type)
> > {
> > static unsigned long num_physpages;
> > @@ -396,7 +370,7 @@ int __init init_numa_memory(void)
> > return -EINVAL;
> >
> > init_node_memblock();
> > - if (numa_meminfo_cover_memory(&numa_meminfo) == false)
> > + if (memblock_validate_numa_coverage(SZ_1M >> 12) == false)
>
> No magic constants please.
> Either use
>
> SZ_1M >> PAGE_SIZE
>
> here, or make threshold in bytes and convert it to number of pages in
> memblock_validate_numa_coverage().
>
> Besides, no need to compare to false,
>
> if (!memblock_validate_numa_coverage())
>
> will do
>
> > return -EINVAL;
> >
> > for_each_node_mask(node, node_possible_map) {
> > diff --git a/arch/x86/mm/numa.c b/arch/x86/mm/numa.c
> > index 2aadb2019b4f..95376e7c263e 100644
> > --- a/arch/x86/mm/numa.c
> > +++ b/arch/x86/mm/numa.c
> > @@ -447,37 +447,6 @@ int __node_distance(int from, int to)
> > }
> > EXPORT_SYMBOL(__node_distance);
> >
> > -/*
> > - * Sanity check to catch more bad NUMA configurations (they are amazingly
> > - * common). Make sure the nodes cover all memory.
> > - */
> > -static bool __init numa_meminfo_cover_memory(const struct numa_meminfo *mi)
> > -{
> > - u64 numaram, e820ram;
> > - int i;
> > -
> > - numaram = 0;
> > - for (i = 0; i < mi->nr_blks; i++) {
> > - u64 s = mi->blk[i].start >> PAGE_SHIFT;
> > - u64 e = mi->blk[i].end >> PAGE_SHIFT;
> > - numaram += e - s;
> > - numaram -= __absent_pages_in_range(mi->blk[i].nid, s, e);
> > - if ((s64)numaram < 0)
> > - numaram = 0;
> > - }
> > -
> > - e820ram = max_pfn - absent_pages_in_range(0, max_pfn);
> > -
> > - /* We seem to lose 3 pages somewhere. Allow 1M of slack. */
> > - if ((s64)(e820ram - numaram) >= (1 << (20 - PAGE_SHIFT))) {
> > - printk(KERN_ERR "NUMA: nodes only cover %LuMB of your %LuMB e820 RAM. Not used.\n",
> > - (numaram << PAGE_SHIFT) >> 20,
> > - (e820ram << PAGE_SHIFT) >> 20);
> > - return false;
> > - }
> > - return true;
> > -}
> > -
> > /*
> > * Mark all currently memblock-reserved physical memory (which covers the
> > * kernel's own memory ranges) as hot-unswappable.
> > @@ -583,7 +552,8 @@ static int __init numa_register_memblks(struct numa_meminfo *mi)
> > return -EINVAL;
> > }
> > }
> > - if (!numa_meminfo_cover_memory(mi))
> > +
> > + if (!memblock_validate_numa_coverage(SZ_1M >> 12))
> > return -EINVAL;
> >
> > /* Finally register nodes. */
> > diff --git a/include/linux/memblock.h b/include/linux/memblock.h
> > index 1c1072e3ca06..727242f4b54a 100644
> > --- a/include/linux/memblock.h
> > +++ b/include/linux/memblock.h
> > @@ -120,6 +120,7 @@ int memblock_physmem_add(phys_addr_t base, phys_addr_t size);
> > void memblock_trim_memory(phys_addr_t align);
> > bool memblock_overlaps_region(struct memblock_type *type,
> > phys_addr_t base, phys_addr_t size);
> > +bool memblock_validate_numa_coverage(const u64 threshold_pages);
> > 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);
> > diff --git a/mm/memblock.c b/mm/memblock.c
> > index 0863222af4a4..4f1f2d8a8119 100644
> > --- a/mm/memblock.c
> > +++ b/mm/memblock.c
> > @@ -734,6 +734,40 @@ int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size)
> > return memblock_add_range(&memblock.memory, base, size, MAX_NUMNODES, 0);
> > }
> >
> > +/**
> > + * memblock_validate_numa_coverage - calculating memory with no node id assigned by firmware
> > + * @threshold_pages: threshold memory of no node id assigned
> > + *
> > + * calculating memory with no node id assigned by firmware,
> > + * If the number is less than the @threshold_pages, it returns true,
> > + * otherwise it returns false.
> > + *
> > + * Return:
> > + * true on success, false on failure.
> > + */
>
> I'd suggest the below version:
>
> /**
> * memblock_validate_numa_coverage - check if amount of memory with
> * no node ID assigned is less than a threshold
> * @threshold_pages: maximal number of pages that can have unassigned node
> * ID (in pages).
> *
> * A buggy firmware may report memory that does not belong to any node.
> * Check if amount of such memory is below @threshold_pages.
> *
> * Return: true on success, false on failure.
> */
>
> > +bool __init_memblock memblock_validate_numa_coverage(const u64 threshold_pages)
> > +{
> > + unsigned long nr_pages = 0;
> > + unsigned long start_pfn, end_pfn, mem_size_mb;
> > + int nid, i;
> > +
> > + /* calculate lose page */
> > + for_each_mem_pfn_range(i, MAX_NUMNODES, &start_pfn, &end_pfn, &nid) {
> > + if (nid == NUMA_NO_NODE)
> > + nr_pages += end_pfn - start_pfn;
> > + }
> > +
> > + if (nr_pages >= threshold_pages) {
> > + mem_size_mb = memblock_phys_mem_size() >> 20;
> > + pr_err("NUMA: no nodes coverage for %luMB of %luMB RAM\n",
> > + (nr_pages << PAGE_SHIFT) >> 20, mem_size_mb);
> > + return false;
> > + }
> > +
> > + return true;
> > +}
> > +
> > +
> > /**
> > * memblock_isolate_range - isolate given range into disjoint memblocks
> > * @type: memblock type to isolate range for
> > --
> > 2.25.1
> >
>
> --
> Sincerely yours,
> Mike.