[PATCH v4 0/2] mm: using CMA for 1 GB hugepages allocation

[PATCH v4 0/2] mm: using CMA for 1 GB hugepages allocation

[Roman Gushchin]

The patchset adds a hugetlb_cma boot option, which allows
to reserve a cma area which can be later used for 1 GB
hugepages allocations.

This is v4 of the patch(set). It contains a patch from Aslan,
which adds a useful function of the cma side, and the previous
version of the hugetlb_cma patch (v3) with all following cleanups
and fixes squashed, plus the following changes:
1) removed the hard-coded archs list from docs
2) added a warning printing on non-supported archs
3) hugetlb_lock is temporarily dropped in update_and_free_page()

I've retained Michal's and Mike's acks, because changes are
not significant. Please, let me know if there is something
wrong.

Thanks!


Aslan Bakirov (1):
  mm: cma: NUMA node interface

Roman Gushchin (1):
  mm: hugetlb: optionally allocate gigantic hugepages using cma

 .../admin-guide/kernel-parameters.txt         |   8 ++
 arch/arm64/mm/init.c                          |   6 +
 arch/x86/kernel/setup.c                       |   4 +
 include/linux/cma.h                           |  13 ++-
 include/linux/hugetlb.h                       |  12 ++
 include/linux/memblock.h                      |   3 +
 mm/cma.c                                      |  16 +--
 mm/hugetlb.c                                  | 109 ++++++++++++++++++
 mm/memblock.c                                 |   2 +-
 9 files changed, 163 insertions(+), 10 deletions(-)

-- 
2.25.1

[PATCH v4 1/2] mm: cma: NUMA node interface

[Roman Gushchin]

From: Aslan Bakirov <aslan@fb.com>

I've noticed that there is no interfaces exposed by CMA which would let me
to declare contigous memory on particular NUMA node.

This patchset adds the ability to try to allocate contiguous memory on
specific node. It will fallback to other nodes if the specified one
doesn't work.

Implement a new method for declaring contigous memory on particular node
and keep cma_declare_contiguous() as a wrapper.

Signed-off-by: Aslan Bakirov <aslan@fb.com>
---
 include/linux/cma.h      | 13 +++++++++++--
 include/linux/memblock.h |  3 +++
 mm/cma.c                 | 16 +++++++++-------
 mm/memblock.c            |  2 +-
 4 files changed, 24 insertions(+), 10 deletions(-)

diff --git a/include/linux/cma.h b/include/linux/cma.h
index 190184b5ff32..eae834c2162f 100644
--- a/include/linux/cma.h
+++ b/include/linux/cma.h
@@ -24,10 +24,19 @@ extern phys_addr_t cma_get_base(const struct cma *cma);
 extern unsigned long cma_get_size(const struct cma *cma);
 extern const char *cma_get_name(const struct cma *cma);
 
-extern int __init cma_declare_contiguous(phys_addr_t base,
+extern int __init cma_declare_contiguous_nid(phys_addr_t base,
 			phys_addr_t size, phys_addr_t limit,
 			phys_addr_t alignment, unsigned int order_per_bit,
-			bool fixed, const char *name, struct cma **res_cma);
+			bool fixed, const char *name, struct cma **res_cma,
+			int nid);
+static inline int __init cma_declare_contiguous(phys_addr_t base,
+			phys_addr_t size, phys_addr_t limit,
+			phys_addr_t alignment, unsigned int order_per_bit,
+			bool fixed, const char *name, struct cma **res_cma)
+{
+	return cma_declare_contiguous_nid(base, size, limit, alignment,
+			order_per_bit, fixed, name, res_cma, NUMA_NO_NODE);
+}
 extern int cma_init_reserved_mem(phys_addr_t base, phys_addr_t size,
 					unsigned int order_per_bit,
 					const char *name,
diff --git a/include/linux/memblock.h b/include/linux/memblock.h
index 079d17d96410..6bc37a731d27 100644
--- a/include/linux/memblock.h
+++ b/include/linux/memblock.h
@@ -348,6 +348,9 @@ static inline int memblock_get_region_node(const struct memblock_region *r)
 
 phys_addr_t memblock_phys_alloc_range(phys_addr_t size, phys_addr_t align,
 				      phys_addr_t start, phys_addr_t end);
+phys_addr_t memblock_alloc_range_nid(phys_addr_t size,
+				      phys_addr_t align, phys_addr_t start,
+				      phys_addr_t end, int nid, bool exact_nid);
 phys_addr_t memblock_phys_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid);
 
 static inline phys_addr_t memblock_phys_alloc(phys_addr_t size,
diff --git a/mm/cma.c b/mm/cma.c
index be55d1988c67..0463ad2ce06b 100644
--- a/mm/cma.c
+++ b/mm/cma.c
@@ -220,7 +220,7 @@ int __init cma_init_reserved_mem(phys_addr_t base, phys_addr_t size,
 }
 
 /**
- * cma_declare_contiguous() - reserve custom contiguous area
+ * cma_declare_contiguous_nid() - reserve custom contiguous area
  * @base: Base address of the reserved area optional, use 0 for any
  * @size: Size of the reserved area (in bytes),
  * @limit: End address of the reserved memory (optional, 0 for any).
@@ -229,6 +229,7 @@ int __init cma_init_reserved_mem(phys_addr_t base, phys_addr_t size,
  * @fixed: hint about where to place the reserved area
  * @name: The name of the area. See function cma_init_reserved_mem()
  * @res_cma: Pointer to store the created cma region.
+ * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
  *
  * This function reserves memory from early allocator. It should be
  * called by arch specific code once the early allocator (memblock or bootmem)
@@ -238,10 +239,11 @@ int __init cma_init_reserved_mem(phys_addr_t base, phys_addr_t size,
  * If @fixed is true, reserve contiguous area at exactly @base.  If false,
  * reserve in range from @base to @limit.
  */
-int __init cma_declare_contiguous(phys_addr_t base,
+int __init cma_declare_contiguous_nid(phys_addr_t base,
 			phys_addr_t size, phys_addr_t limit,
 			phys_addr_t alignment, unsigned int order_per_bit,
-			bool fixed, const char *name, struct cma **res_cma)
+			bool fixed, const char *name, struct cma **res_cma,
+			int nid)
 {
 	phys_addr_t memblock_end = memblock_end_of_DRAM();
 	phys_addr_t highmem_start;
@@ -336,14 +338,14 @@ int __init cma_declare_contiguous(phys_addr_t base,
 		 * memory in case of failure.
 		 */
 		if (base < highmem_start && limit > highmem_start) {
-			addr = memblock_phys_alloc_range(size, alignment,
-							 highmem_start, limit);
+			addr = memblock_alloc_range_nid(size, alignment,
+					highmem_start, limit, nid, false);
 			limit = highmem_start;
 		}
 
 		if (!addr) {
-			addr = memblock_phys_alloc_range(size, alignment, base,
-							 limit);
+			addr = memblock_alloc_range_nid(size, alignment, base,
+					limit, nid, false);
 			if (!addr) {
 				ret = -ENOMEM;
 				goto err;
diff --git a/mm/memblock.c b/mm/memblock.c
index 4d06bbaded0f..c79ba6f9920c 100644
--- a/mm/memblock.c
+++ b/mm/memblock.c
@@ -1349,7 +1349,7 @@ __next_mem_pfn_range_in_zone(u64 *idx, struct zone *zone,
  * Return:
  * Physical address of allocated memory block on success, %0 on failure.
  */
-static phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size,
+phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size,
 					phys_addr_t align, phys_addr_t start,
 					phys_addr_t end, int nid,
 					bool exact_nid)
-- 
2.25.1

[PATCH v4 2/2] mm: hugetlb: optionally allocate gigantic hugepages using cma

[Roman Gushchin]

Commit 944d9fec8d7a ("hugetlb: add support for gigantic page allocation at
runtime") has added the run-time allocation of gigantic pages.  However it
actually works only at early stages of the system loading, when the
majority of memory is free.  After some time the memory gets fragmented by
non-movable pages, so the chances to find a contiguous 1 GB block are
getting close to zero.  Even dropping caches manually doesn't help a lot.

At large scale rebooting servers in order to allocate gigantic hugepages
is quite expensive and complex.  At the same time keeping some constant
percentage of memory in reserved hugepages even if the workload isn't
using it is a big waste: not all workloads can benefit from using 1 GB
pages.

The following solution can solve the problem:
1) On boot time a dedicated cma area* is reserved. The size is passed
   as a kernel argument.
2) Run-time allocations of gigantic hugepages are performed using the
   cma allocator and the dedicated cma area

In this case gigantic hugepages can be allocated successfully with a high
probability, however the memory isn't completely wasted if nobody is using
1GB hugepages: it can be used for pagecache, anon memory, THPs, etc.

* On a multi-node machine a per-node cma area is allocated on each node.
  Following gigantic hugetlb allocation are using the first available
  numa node if the mask isn't specified by a user.

Usage:
1) configure the kernel to allocate a cma area for hugetlb allocations:
   pass hugetlb_cma=10G as a kernel argument

2) allocate hugetlb pages as usual, e.g.
   echo 10 > /sys/kernel/mm/hugepages/hugepages-1048576kB/nr_hugepages

If the option isn't enabled or the allocation of the cma area failed,
the current behavior of the system is preserved.

x86 and arm-64 are covered by this patch, other architectures can be
trivially added later.

The patch contains clean-ups and fixes proposed and implemented by
Aslan Bakirov and Randy Dunlap. It also contains ideas and suggestions
proposed by Rik van Riel, Michal Hocko and Mike Kravetz. Thanks!

Link: http://lkml.kernel.org/r/20200311220920.2487528-1-guro@fb.com
Signed-off-by: Roman Gushchin <guro@fb.com>
Tested-by: Andreas Schaufler <andreas.schaufler@gmx.de>
Acked-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Michal Hocko <mhocko@kernel.org>
Cc: Aslan Bakirov <aslan@fb.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Joonsoo Kim <js1304@gmail.com>
Signed-off-by: Roman Gushchin <guro@fb.com>
---
 .../admin-guide/kernel-parameters.txt         |   8 ++
 arch/arm64/mm/init.c                          |   6 +
 arch/x86/kernel/setup.c                       |   4 +
 include/linux/hugetlb.h                       |  12 ++
 mm/hugetlb.c                                  | 109 ++++++++++++++++++
 5 files changed, 139 insertions(+)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 4d5a4fe22703..59cca49a5249 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -1473,6 +1473,14 @@
 	hpet_mmap=	[X86, HPET_MMAP] Allow userspace to mmap HPET
 			registers.  Default set by CONFIG_HPET_MMAP_DEFAULT.
 
+	hugetlb_cma=	[HW] The size of a cma area used for allocation
+			of gigantic hugepages.
+			Format: nn[KMGTPE]
+
+			Reserve a cma area of given size and allocate gigantic
+			hugepages using the cma allocator. If enabled, the
+			boot-time allocation of gigantic hugepages is skipped.
+
 	hugepages=	[HW,X86-32,IA-64] HugeTLB pages to allocate at boot.
 	hugepagesz=	[HW,IA-64,PPC,X86-64] The size of the HugeTLB pages.
 			On x86-64 and powerpc, this option can be specified
diff --git a/arch/arm64/mm/init.c b/arch/arm64/mm/init.c
index b65dffdfb201..e42727e3568e 100644
--- a/arch/arm64/mm/init.c
+++ b/arch/arm64/mm/init.c
@@ -29,6 +29,7 @@
 #include <linux/mm.h>
 #include <linux/kexec.h>
 #include <linux/crash_dump.h>
+#include <linux/hugetlb.h>
 
 #include <asm/boot.h>
 #include <asm/fixmap.h>
@@ -457,6 +458,11 @@ void __init arm64_memblock_init(void)
 	high_memory = __va(memblock_end_of_DRAM() - 1) + 1;
 
 	dma_contiguous_reserve(arm64_dma32_phys_limit);
+
+#ifdef CONFIG_ARM64_4K_PAGES
+	hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
+#endif
+
 }
 
 void __init bootmem_init(void)
diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
index e6b545047f38..4b3fa6cd3106 100644
--- a/arch/x86/kernel/setup.c
+++ b/arch/x86/kernel/setup.c
@@ -16,6 +16,7 @@
 #include <linux/pci.h>
 #include <linux/root_dev.h>
 #include <linux/sfi.h>
+#include <linux/hugetlb.h>
 #include <linux/tboot.h>
 #include <linux/usb/xhci-dbgp.h>
 
@@ -1157,6 +1158,9 @@ void __init setup_arch(char **cmdline_p)
 	initmem_init();
 	dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
 
+	if (boot_cpu_has(X86_FEATURE_GBPAGES))
+		hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
+
 	/*
 	 * Reserve memory for crash kernel after SRAT is parsed so that it
 	 * won't consume hotpluggable memory.
diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 5ea05879a0a9..43a1cef8f0f1 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -895,4 +895,16 @@ static inline spinlock_t *huge_pte_lock(struct hstate *h,
 	return ptl;
 }
 
+#if defined(CONFIG_HUGETLB_PAGE) && defined(CONFIG_CMA)
+extern void __init hugetlb_cma_reserve(int order);
+extern void __init hugetlb_cma_check(void);
+#else
+static inline __init void hugetlb_cma_reserve(int order)
+{
+}
+static inline __init void hugetlb_cma_check(void)
+{
+}
+#endif
+
 #endif /* _LINUX_HUGETLB_H */
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index f9ea1e5197b4..b2c56c3c3e67 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -28,6 +28,7 @@
 #include <linux/jhash.h>
 #include <linux/numa.h>
 #include <linux/llist.h>
+#include <linux/cma.h>
 
 #include <asm/page.h>
 #include <asm/pgtable.h>
@@ -44,6 +45,9 @@
 int hugetlb_max_hstate __read_mostly;
 unsigned int default_hstate_idx;
 struct hstate hstates[HUGE_MAX_HSTATE];
+
+static struct cma *hugetlb_cma[MAX_NUMNODES];
+
 /*
  * Minimum page order among possible hugepage sizes, set to a proper value
  * at boot time.
@@ -1228,6 +1232,14 @@ static void destroy_compound_gigantic_page(struct page *page,
 
 static void free_gigantic_page(struct page *page, unsigned int order)
 {
+	/*
+	 * If the page isn't allocated using the cma allocator,
+	 * cma_release() returns false.
+	 */
+	if (IS_ENABLED(CONFIG_CMA) &&
+	    cma_release(hugetlb_cma[page_to_nid(page)], page, 1 << order))
+		return;
+
 	free_contig_range(page_to_pfn(page), 1 << order);
 }
 
@@ -1237,6 +1249,21 @@ static struct page *alloc_gigantic_page(struct hstate *h, gfp_t gfp_mask,
 {
 	unsigned long nr_pages = 1UL << huge_page_order(h);
 
+	if (IS_ENABLED(CONFIG_CMA)) {
+		struct page *page;
+		int node;
+
+		for_each_node_mask(node, *nodemask) {
+			if (!hugetlb_cma[node])
+				break;
+
+			page = cma_alloc(hugetlb_cma[node], nr_pages,
+					 huge_page_order(h), true);
+			if (page)
+				return page;
+		}
+	}
+
 	return alloc_contig_pages(nr_pages, gfp_mask, nid, nodemask);
 }
 
@@ -1281,8 +1308,14 @@ static void update_and_free_page(struct hstate *h, struct page *page)
 	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
 	set_page_refcounted(page);
 	if (hstate_is_gigantic(h)) {
+		/*
+		 * Temporarily drop the hugetlb_lock, because
+		 * we might block in free_gigantic_page().
+		 */
+		spin_unlock(&hugetlb_lock);
 		destroy_compound_gigantic_page(page, huge_page_order(h));
 		free_gigantic_page(page, huge_page_order(h));
+		spin_lock(&hugetlb_lock);
 	} else {
 		__free_pages(page, huge_page_order(h));
 	}
@@ -2538,6 +2571,10 @@ static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
 
 	for (i = 0; i < h->max_huge_pages; ++i) {
 		if (hstate_is_gigantic(h)) {
+			if (IS_ENABLED(CONFIG_CMA) && hugetlb_cma[0]) {
+				pr_warn_once("HugeTLB: hugetlb_cma is enabled, skip boot time allocation\n");
+				break;
+			}
 			if (!alloc_bootmem_huge_page(h))
 				break;
 		} else if (!alloc_pool_huge_page(h,
@@ -3193,6 +3230,7 @@ static int __init hugetlb_init(void)
 			default_hstate.max_huge_pages = default_hstate_max_huge_pages;
 	}
 
+	hugetlb_cma_check();
 	hugetlb_init_hstates();
 	gather_bootmem_prealloc();
 	report_hugepages();
@@ -5505,3 +5543,74 @@ void move_hugetlb_state(struct page *oldpage, struct page *newpage, int reason)
 		spin_unlock(&hugetlb_lock);
 	}
 }
+
+#ifdef CONFIG_CMA
+static unsigned long hugetlb_cma_size __initdata;
+static bool cma_reserve_called __initdata;
+
+static int __init cmdline_parse_hugetlb_cma(char *p)
+{
+	hugetlb_cma_size = memparse(p, &p);
+	return 0;
+}
+
+early_param("hugetlb_cma", cmdline_parse_hugetlb_cma);
+
+void __init hugetlb_cma_reserve(int order)
+{
+	unsigned long size, reserved, per_node;
+	int nid;
+
+	cma_reserve_called = true;
+
+	if (!hugetlb_cma_size)
+		return;
+
+	if (hugetlb_cma_size < (PAGE_SIZE << order)) {
+		pr_warn("hugetlb_cma: cma area should be at least %lu MiB\n",
+			(PAGE_SIZE << order) / SZ_1M);
+		return;
+	}
+
+	/*
+	 * If 3 GB area is requested on a machine with 4 numa nodes,
+	 * let's allocate 1 GB on first three nodes and ignore the last one.
+	 */
+	per_node = DIV_ROUND_UP(hugetlb_cma_size, nr_online_nodes);
+	pr_info("hugetlb_cma: reserve %lu MiB, up to %lu MiB per node\n",
+		hugetlb_cma_size / SZ_1M, per_node / SZ_1M);
+
+	reserved = 0;
+	for_each_node_state(nid, N_ONLINE) {
+		int res;
+
+		size = min(per_node, hugetlb_cma_size - reserved);
+		size = round_up(size, PAGE_SIZE << order);
+
+		res = cma_declare_contiguous_nid(0, size, 0, PAGE_SIZE << order,
+						 0, false, "hugetlb",
+						 &hugetlb_cma[nid], nid);
+		if (res) {
+			pr_warn("hugetlb_cma: reservation failed: err %d, node %d",
+				res, nid);
+			break;
+		}
+
+		reserved += size;
+		pr_info("hugetlb_cma: reserved %lu MiB on node %d\n",
+			size / SZ_1M, nid);
+
+		if (reserved >= hugetlb_cma_size)
+			break;
+	}
+}
+
+void __init hugetlb_cma_check(void)
+{
+	if (!hugetlb_cma_size || cma_reserve_called)
+		return;
+
+	pr_warn("hugetlb_cma: the option isn't supported by current arch\n");
+}
+
+#endif /* CONFIG_CMA */
-- 
2.25.1

Re: [PATCH v4 2/2] mm: hugetlb: optionally allocate gigantic hugepages using cma

[Michal Hocko]

On Mon 06-04-20 18:04:31, Roman Gushchin wrote:
[...]
My ack still applies but I have only noticed two minor things now.

[...]
> @@ -1281,8 +1308,14 @@ static void update_and_free_page(struct hstate *h, struct page *page)
>  	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
>  	set_page_refcounted(page);
>  	if (hstate_is_gigantic(h)) {
> +		/*
> +		 * Temporarily drop the hugetlb_lock, because
> +		 * we might block in free_gigantic_page().
> +		 */
> +		spin_unlock(&hugetlb_lock);
>  		destroy_compound_gigantic_page(page, huge_page_order(h));
>  		free_gigantic_page(page, huge_page_order(h));
> +		spin_lock(&hugetlb_lock);

This is OK with the current code because existing paths do not have to
revalidate the state AFAICS but it is a bit subtle. I have checked the
cma_free path and it can only sleep on the cma->lock unless I am missing
something. This lock is only used for cma bitmap manipulation and the
mutex sounds like an overkill there and it can be replaced by a
spinlock.

Sounds like a follow up patch material to me.

[...]
> +	for_each_node_state(nid, N_ONLINE) {
> +		int res;
> +
> +		size = min(per_node, hugetlb_cma_size - reserved);
> +		size = round_up(size, PAGE_SIZE << order);
> +
> +		res = cma_declare_contiguous_nid(0, size, 0, PAGE_SIZE << order,
> +						 0, false, "hugetlb",
> +						 &hugetlb_cma[nid], nid);
> +		if (res) {
> +			pr_warn("hugetlb_cma: reservation failed: err %d, node %d",
> +				res, nid);
> +			break;

Do we really have to break out after a single node failure? There might
be other nodes that can satisfy the allocation. You are not cleaning up
previous allocations so there is a partial state and then it would make
more sense to me to simply s@break@continue@ here.

> +		}
> +
> +		reserved += size;
> +		pr_info("hugetlb_cma: reserved %lu MiB on node %d\n",
> +			size / SZ_1M, nid);
> +
> +		if (reserved >= hugetlb_cma_size)
> +			break;
> +	}
> +}
-- 
Michal Hocko
SUSE Labs

Re: [PATCH v4 2/2] mm: hugetlb: optionally allocate gigantic hugepages using cma

[Roman Gushchin]

On Tue, Apr 07, 2020 at 09:03:31AM +0200, Michal Hocko wrote:
> On Mon 06-04-20 18:04:31, Roman Gushchin wrote:
> [...]
> My ack still applies but I have only noticed two minor things now.

Hello, Michal!

> 
> [...]
> > @@ -1281,8 +1308,14 @@ static void update_and_free_page(struct hstate *h, struct page *page)
> >  	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
> >  	set_page_refcounted(page);
> >  	if (hstate_is_gigantic(h)) {
> > +		/*
> > +		 * Temporarily drop the hugetlb_lock, because
> > +		 * we might block in free_gigantic_page().
> > +		 */
> > +		spin_unlock(&hugetlb_lock);
> >  		destroy_compound_gigantic_page(page, huge_page_order(h));
> >  		free_gigantic_page(page, huge_page_order(h));
> > +		spin_lock(&hugetlb_lock);
> 
> This is OK with the current code because existing paths do not have to
> revalidate the state AFAICS but it is a bit subtle. I have checked the
> cma_free path and it can only sleep on the cma->lock unless I am missing
> something. This lock is only used for cma bitmap manipulation and the
> mutex sounds like an overkill there and it can be replaced by a
> spinlock.
> 
> Sounds like a follow up patch material to me.

I had the same idea and even posted a patch:
https://lore.kernel.org/linux-mm/20200403174559.GC220160@carbon.lan/T/#m87be98bdacda02cea3dd6759b48a28bd23f29ff0

However, Joonsoo pointed out that in some cases the bitmap operation might
be too long for a spinlock.

Alternatively, we can implement an asynchronous delayed release on the cma side,
I just don't know if it's worth it (I mean adding code/complexity).

> 
> [...]
> > +	for_each_node_state(nid, N_ONLINE) {
> > +		int res;
> > +
> > +		size = min(per_node, hugetlb_cma_size - reserved);
> > +		size = round_up(size, PAGE_SIZE << order);
> > +
> > +		res = cma_declare_contiguous_nid(0, size, 0, PAGE_SIZE << order,
> > +						 0, false, "hugetlb",
> > +						 &hugetlb_cma[nid], nid);
> > +		if (res) {
> > +			pr_warn("hugetlb_cma: reservation failed: err %d, node %d",
> > +				res, nid);
> > +			break;
> 
> Do we really have to break out after a single node failure? There might
> be other nodes that can satisfy the allocation. You are not cleaning up
> previous allocations so there is a partial state and then it would make
> more sense to me to simply s@break@continue@ here.

But then we should iterate over all nodes in alloc_gigantic_page()?
Currently if hugetlb_cma[0] is NULL it will immediately switch back
to the fallback approach.

Actually, Idk how realistic are use cases with complex node configuration,
so that we can hugetlb_cma areas can be allocated only on some of them.
I'd leave it up to the moment when we'll have a real world example.
Then we probably want something more sophisticated anyway...

I have no strong opinion here, so if you really think we should s/break/continue,
I'm fine with it too.

Thanks!

Re: [PATCH v4 2/2] mm: hugetlb: optionally allocate gigantic hugepages using cma

[Michal Hocko]

On Tue 07-04-20 08:25:44, Roman Gushchin wrote:
> On Tue, Apr 07, 2020 at 09:03:31AM +0200, Michal Hocko wrote:
> > On Mon 06-04-20 18:04:31, Roman Gushchin wrote:
> > [...]
> > My ack still applies but I have only noticed two minor things now.
> 
> Hello, Michal!
> 
> > 
> > [...]
> > > @@ -1281,8 +1308,14 @@ static void update_and_free_page(struct hstate *h, struct page *page)
> > >  	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
> > >  	set_page_refcounted(page);
> > >  	if (hstate_is_gigantic(h)) {
> > > +		/*
> > > +		 * Temporarily drop the hugetlb_lock, because
> > > +		 * we might block in free_gigantic_page().
> > > +		 */
> > > +		spin_unlock(&hugetlb_lock);
> > >  		destroy_compound_gigantic_page(page, huge_page_order(h));
> > >  		free_gigantic_page(page, huge_page_order(h));
> > > +		spin_lock(&hugetlb_lock);
> > 
> > This is OK with the current code because existing paths do not have to
> > revalidate the state AFAICS but it is a bit subtle. I have checked the
> > cma_free path and it can only sleep on the cma->lock unless I am missing
> > something. This lock is only used for cma bitmap manipulation and the
> > mutex sounds like an overkill there and it can be replaced by a
> > spinlock.
> > 
> > Sounds like a follow up patch material to me.
> 
> I had the same idea and even posted a patch:
> https://lore.kernel.org/linux-mm/20200403174559.GC220160@carbon.lan/T/#m87be98bdacda02cea3dd6759b48a28bd23f29ff0
> 
> However, Joonsoo pointed out that in some cases the bitmap operation might
> be too long for a spinlock.

I was not aware of this email thread. I will have a look. Thanks!
 
> Alternatively, we can implement an asynchronous delayed release on the cma side,
> I just don't know if it's worth it (I mean adding code/complexity).
> 
> > 
> > [...]
> > > +	for_each_node_state(nid, N_ONLINE) {
> > > +		int res;
> > > +
> > > +		size = min(per_node, hugetlb_cma_size - reserved);
> > > +		size = round_up(size, PAGE_SIZE << order);
> > > +
> > > +		res = cma_declare_contiguous_nid(0, size, 0, PAGE_SIZE << order,
> > > +						 0, false, "hugetlb",
> > > +						 &hugetlb_cma[nid], nid);
> > > +		if (res) {
> > > +			pr_warn("hugetlb_cma: reservation failed: err %d, node %d",
> > > +				res, nid);
> > > +			break;
> > 
> > Do we really have to break out after a single node failure? There might
> > be other nodes that can satisfy the allocation. You are not cleaning up
> > previous allocations so there is a partial state and then it would make
> > more sense to me to simply s@break@continue@ here.
> 
> But then we should iterate over all nodes in alloc_gigantic_page()?

OK, I've managed to miss the early break on hugetlb_cma[node] == NULL
there as well. I do not think this makes much sense. Just consider a
setup with one node much smaller than others (not unseen on LPAR
configurations) and then you are potentially using CMA areas on some
nodes without a good reason.

> Currently if hugetlb_cma[0] is NULL it will immediately switch back
> to the fallback approach.
> 
> Actually, Idk how realistic are use cases with complex node configuration,
> so that we can hugetlb_cma areas can be allocated only on some of them.
> I'd leave it up to the moment when we'll have a real world example.
> Then we probably want something more sophisticated anyway...

I do not follow. Isn't the s@break@continue@ in this and
alloc_gigantic_page path enough to make it work?
-- 
Michal Hocko
SUSE Labs

Re: [PATCH v4 2/2] mm: hugetlb: optionally allocate gigantic hugepages using cma

[Roman Gushchin]

On Tue, Apr 07, 2020 at 05:40:05PM +0200, Michal Hocko wrote:
> On Tue 07-04-20 08:25:44, Roman Gushchin wrote:
> > On Tue, Apr 07, 2020 at 09:03:31AM +0200, Michal Hocko wrote:
> > > On Mon 06-04-20 18:04:31, Roman Gushchin wrote:
> > > [...]
> > > My ack still applies but I have only noticed two minor things now.
> > 
> > Hello, Michal!
> > 
> > > 
> > > [...]
> > > > @@ -1281,8 +1308,14 @@ static void update_and_free_page(struct hstate *h, struct page *page)
> > > >  	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
> > > >  	set_page_refcounted(page);
> > > >  	if (hstate_is_gigantic(h)) {
> > > > +		/*
> > > > +		 * Temporarily drop the hugetlb_lock, because
> > > > +		 * we might block in free_gigantic_page().
> > > > +		 */
> > > > +		spin_unlock(&hugetlb_lock);
> > > >  		destroy_compound_gigantic_page(page, huge_page_order(h));
> > > >  		free_gigantic_page(page, huge_page_order(h));
> > > > +		spin_lock(&hugetlb_lock);
> > > 
> > > This is OK with the current code because existing paths do not have to
> > > revalidate the state AFAICS but it is a bit subtle. I have checked the
> > > cma_free path and it can only sleep on the cma->lock unless I am missing
> > > something. This lock is only used for cma bitmap manipulation and the
> > > mutex sounds like an overkill there and it can be replaced by a
> > > spinlock.
> > > 
> > > Sounds like a follow up patch material to me.
> > 
> > I had the same idea and even posted a patch:
> > https://lore.kernel.org/linux-mm/20200403174559.GC220160@carbon.lan/T/#m87be98bdacda02cea3dd6759b48a28bd23f29ff0
> > 
> > However, Joonsoo pointed out that in some cases the bitmap operation might
> > be too long for a spinlock.
> 
> I was not aware of this email thread. I will have a look. Thanks!
>  
> > Alternatively, we can implement an asynchronous delayed release on the cma side,
> > I just don't know if it's worth it (I mean adding code/complexity).
> > 
> > > 
> > > [...]
> > > > +	for_each_node_state(nid, N_ONLINE) {
> > > > +		int res;
> > > > +
> > > > +		size = min(per_node, hugetlb_cma_size - reserved);
> > > > +		size = round_up(size, PAGE_SIZE << order);
> > > > +
> > > > +		res = cma_declare_contiguous_nid(0, size, 0, PAGE_SIZE << order,
> > > > +						 0, false, "hugetlb",
> > > > +						 &hugetlb_cma[nid], nid);
> > > > +		if (res) {
> > > > +			pr_warn("hugetlb_cma: reservation failed: err %d, node %d",
> > > > +				res, nid);
> > > > +			break;
> > > 
> > > Do we really have to break out after a single node failure? There might
> > > be other nodes that can satisfy the allocation. You are not cleaning up
> > > previous allocations so there is a partial state and then it would make
> > > more sense to me to simply s@break@continue@ here.
> > 
> > But then we should iterate over all nodes in alloc_gigantic_page()?
> 
> OK, I've managed to miss the early break on hugetlb_cma[node] == NULL
> there as well. I do not think this makes much sense. Just consider a
> setup with one node much smaller than others (not unseen on LPAR
> configurations) and then you are potentially using CMA areas on some
> nodes without a good reason.
> 
> > Currently if hugetlb_cma[0] is NULL it will immediately switch back
> > to the fallback approach.
> > 
> > Actually, Idk how realistic are use cases with complex node configuration,
> > so that we can hugetlb_cma areas can be allocated only on some of them.
> > I'd leave it up to the moment when we'll have a real world example.
> > Then we probably want something more sophisticated anyway...
> 
> I do not follow. Isn't the s@break@continue@ in this and
> alloc_gigantic_page path enough to make it work?

Well, of course it will. But for a highly asymmetrical configuration
there is probably not much sense to try allocate cma areas of a similar
size on each node and rely on allocation failures on some of them.

But, again, if you strictly prefer s/break/continue, I can send a v5.
Just let me know.

Thanks!

Re: [PATCH v4 2/2] mm: hugetlb: optionally allocate gigantic hugepages using cma

[Michal Hocko]

On Tue 07-04-20 09:06:40, Roman Gushchin wrote:
> On Tue, Apr 07, 2020 at 05:40:05PM +0200, Michal Hocko wrote:
> > On Tue 07-04-20 08:25:44, Roman Gushchin wrote:
> > > On Tue, Apr 07, 2020 at 09:03:31AM +0200, Michal Hocko wrote:
> > > > On Mon 06-04-20 18:04:31, Roman Gushchin wrote:
> > > > [...]
> > > > My ack still applies but I have only noticed two minor things now.
> > > 
> > > Hello, Michal!
> > > 
> > > > 
> > > > [...]
> > > > > @@ -1281,8 +1308,14 @@ static void update_and_free_page(struct hstate *h, struct page *page)
> > > > >  	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
> > > > >  	set_page_refcounted(page);
> > > > >  	if (hstate_is_gigantic(h)) {
> > > > > +		/*
> > > > > +		 * Temporarily drop the hugetlb_lock, because
> > > > > +		 * we might block in free_gigantic_page().
> > > > > +		 */
> > > > > +		spin_unlock(&hugetlb_lock);
> > > > >  		destroy_compound_gigantic_page(page, huge_page_order(h));
> > > > >  		free_gigantic_page(page, huge_page_order(h));
> > > > > +		spin_lock(&hugetlb_lock);
> > > > 
> > > > This is OK with the current code because existing paths do not have to
> > > > revalidate the state AFAICS but it is a bit subtle. I have checked the
> > > > cma_free path and it can only sleep on the cma->lock unless I am missing
> > > > something. This lock is only used for cma bitmap manipulation and the
> > > > mutex sounds like an overkill there and it can be replaced by a
> > > > spinlock.
> > > > 
> > > > Sounds like a follow up patch material to me.
> > > 
> > > I had the same idea and even posted a patch:
> > > https://lore.kernel.org/linux-mm/20200403174559.GC220160@carbon.lan/T/#m87be98bdacda02cea3dd6759b48a28bd23f29ff0
> > > 
> > > However, Joonsoo pointed out that in some cases the bitmap operation might
> > > be too long for a spinlock.
> > 
> > I was not aware of this email thread. I will have a look. Thanks!
> >  
> > > Alternatively, we can implement an asynchronous delayed release on the cma side,
> > > I just don't know if it's worth it (I mean adding code/complexity).
> > > 
> > > > 
> > > > [...]
> > > > > +	for_each_node_state(nid, N_ONLINE) {
> > > > > +		int res;
> > > > > +
> > > > > +		size = min(per_node, hugetlb_cma_size - reserved);
> > > > > +		size = round_up(size, PAGE_SIZE << order);
> > > > > +
> > > > > +		res = cma_declare_contiguous_nid(0, size, 0, PAGE_SIZE << order,
> > > > > +						 0, false, "hugetlb",
> > > > > +						 &hugetlb_cma[nid], nid);
> > > > > +		if (res) {
> > > > > +			pr_warn("hugetlb_cma: reservation failed: err %d, node %d",
> > > > > +				res, nid);
> > > > > +			break;
> > > > 
> > > > Do we really have to break out after a single node failure? There might
> > > > be other nodes that can satisfy the allocation. You are not cleaning up
> > > > previous allocations so there is a partial state and then it would make
> > > > more sense to me to simply s@break@continue@ here.
> > > 
> > > But then we should iterate over all nodes in alloc_gigantic_page()?
> > 
> > OK, I've managed to miss the early break on hugetlb_cma[node] == NULL
> > there as well. I do not think this makes much sense. Just consider a
> > setup with one node much smaller than others (not unseen on LPAR
> > configurations) and then you are potentially using CMA areas on some
> > nodes without a good reason.
> > 
> > > Currently if hugetlb_cma[0] is NULL it will immediately switch back
> > > to the fallback approach.
> > > 
> > > Actually, Idk how realistic are use cases with complex node configuration,
> > > so that we can hugetlb_cma areas can be allocated only on some of them.
> > > I'd leave it up to the moment when we'll have a real world example.
> > > Then we probably want something more sophisticated anyway...
> > 
> > I do not follow. Isn't the s@break@continue@ in this and
> > alloc_gigantic_page path enough to make it work?
> 
> Well, of course it will. But for a highly asymmetrical configuration
> there is probably not much sense to try allocate cma areas of a similar
> size on each node and rely on allocation failures on some of them.
> 
> But, again, if you strictly prefer s/break/continue, I can send a v5.
> Just let me know.

There is no real reason to have such a restriction. I can follow up with
a separate patch if you want me but it should be "fixed".

Thanks

-- 
Michal Hocko
SUSE Labs