[PATCH v13 0/4] support large align and nid in Rust allocators

[PATCH v13 0/4] support large align and nid in Rust allocators

[Vitaly Wool]

The coming patches provide the ability for Rust allocators to set
NUMA node and large alignment.

Changelog:
v2 -> v3:
* fixed the build breakage for non-MMU configs
v3 -> v4:
* added NUMA node support for k[v]realloc (patch #2)
* removed extra logic in Rust helpers
* patch for Rust allocators split into 2 (align: patch #3 and
  NUMA ids: patch #4)
v4 -> v5:
* reworked NUMA node support for k[v]realloc for all 3 <alloc>_node
  functions to have the same signature
* all 3 <alloc>_node slab/vmalloc functions now support alignment
  specification
* Rust helpers are extended with new functions, the old ones are left
  intact
* Rust support for NUMA nodes comes first now (as patch #3)
v5 -> v6:
* added <alloc>_node_align functions to keep the existing interfaces
  intact
* clearer separation for Rust support of MUNA ids and large alignments
v6 -> v7:
* NUMA identifier as a new Rust type (NumaNode)
* better documentation for changed and new functions and constants
v7 -> v8:
* removed NumaError
* small cleanups per reviewers' comments
v8 -> v9:
* realloc functions can now reallocate memory for a different NUMA
  node
* better comments/explanations in the Rust part
v9 -> v10:
* refined behavior when memory is being reallocated for a different
  NUMA node, comments added
* cleanups in the Rust part, rustfmt ran
* typos corrected
v10 -> v11:
* added documentation for the NO_NODE constant
* added node parameter to Allocator's alloc/realloc instead of adding
  separate alloc_node resp. realloc_node functions, modified users of
  alloc/realloc in accordance with that
v11 -> v12:
* some redundant _noprof functions removed in patch 2/4
* c'n'p error fixed in patch 2/4 (vmalloc_to_page -> virt_to_page)
* some typo corrections and documentation updates, primarily in patch
  3/4
v12 -> v13:
* fixed wording in comments (patches 1, 3)
* fixed bigger alignment handling in krealloc (patch 2)
* removed pr_warn import (patch 4)

Signed-off-by: Vitaly Wool <vitaly.wool@konsulko.se>
--
 fs/bcachefs/darray.c           |    2 -
 fs/bcachefs/util.h             |    2 -
 include/linux/bpfptr.h         |    2 -
 include/linux/slab.h           |   39 ++++++++++++++++++++++---------------
 include/linux/vmalloc.h        |   12 ++++++++---
 lib/rhashtable.c               |    4 +--
 mm/nommu.c                     |    3 +-
 mm/slub.c                      |   64 +++++++++++++++++++++++++++++++++++++++++--------------------
 mm/vmalloc.c                   |   29 ++++++++++++++++++++++-----
 rust/helpers/slab.c            |   10 +++++----
 rust/helpers/vmalloc.c         |    5 ++--
 rust/kernel/alloc.rs           |   54 ++++++++++++++++++++++++++++++++++++++++++++++-----
 rust/kernel/alloc/allocator.rs |   49 +++++++++++++++++++++-------------------------
 rust/kernel/alloc/kbox.rs      |    4 +--
 rust/kernel/alloc/kvec.rs      |   11 ++++++++--
 15 files changed, 200 insertions(+), 90 deletions(-)

[PATCH v13 1/4] :mm/vmalloc: allow to set node and align in vrealloc

[Vitaly Wool]

Reimplement vrealloc() to be able to set node and alignment should
a user need to do so. Rename the function to vrealloc_node_align()
to better match what it actually does now and introduce macros for
vrealloc() and friends for backward compatibility.

With that change we also provide the ability for the Rust part of
the kernel to set node and alignment in its allocations.

Signed-off-by: Vitaly Wool <vitaly.wool@konsulko.se>
Reviewed-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
---
 include/linux/vmalloc.h | 12 +++++++++---
 mm/nommu.c              |  3 ++-
 mm/vmalloc.c            | 29 ++++++++++++++++++++++++-----
 3 files changed, 35 insertions(+), 9 deletions(-)

diff --git a/include/linux/vmalloc.h b/include/linux/vmalloc.h
index fdc9aeb74a44..68791f7cb3ba 100644
--- a/include/linux/vmalloc.h
+++ b/include/linux/vmalloc.h
@@ -197,9 +197,15 @@ extern void *__vcalloc_noprof(size_t n, size_t size, gfp_t flags) __alloc_size(1
 extern void *vcalloc_noprof(size_t n, size_t size) __alloc_size(1, 2);
 #define vcalloc(...)		alloc_hooks(vcalloc_noprof(__VA_ARGS__))
 
-void * __must_check vrealloc_noprof(const void *p, size_t size, gfp_t flags)
-		__realloc_size(2);
-#define vrealloc(...)		alloc_hooks(vrealloc_noprof(__VA_ARGS__))
+void *__must_check vrealloc_node_align_noprof(const void *p, size_t size,
+		unsigned long align, gfp_t flags, int nid) __realloc_size(2);
+#define vrealloc_node_noprof(_p, _s, _f, _nid)	\
+	vrealloc_node_align_noprof(_p, _s, 1, _f, _nid)
+#define vrealloc_noprof(_p, _s, _f)		\
+	vrealloc_node_align_noprof(_p, _s, 1, _f, NUMA_NO_NODE)
+#define vrealloc_node_align(...)		alloc_hooks(vrealloc_node_align_noprof(__VA_ARGS__))
+#define vrealloc_node(...)			alloc_hooks(vrealloc_node_noprof(__VA_ARGS__))
+#define vrealloc(...)				alloc_hooks(vrealloc_noprof(__VA_ARGS__))
 
 extern void vfree(const void *addr);
 extern void vfree_atomic(const void *addr);
diff --git a/mm/nommu.c b/mm/nommu.c
index b624acec6d2e..afde6c626b07 100644
--- a/mm/nommu.c
+++ b/mm/nommu.c
@@ -119,7 +119,8 @@ void *__vmalloc_noprof(unsigned long size, gfp_t gfp_mask)
 }
 EXPORT_SYMBOL(__vmalloc_noprof);
 
-void *vrealloc_noprof(const void *p, size_t size, gfp_t flags)
+void *vrealloc_node_align_noprof(const void *p, size_t size, unsigned long align,
+				 gfp_t flags, int node)
 {
 	return krealloc_noprof(p, size, (flags | __GFP_COMP) & ~__GFP_HIGHMEM);
 }
diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index ab986dd09b6a..e0a593651d96 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -4081,19 +4081,29 @@ void *vzalloc_node_noprof(unsigned long size, int node)
 EXPORT_SYMBOL(vzalloc_node_noprof);
 
 /**
- * vrealloc - reallocate virtually contiguous memory; contents remain unchanged
+ * vrealloc_node_align_noprof - reallocate virtually contiguous memory; contents
+ * remain unchanged
  * @p: object to reallocate memory for
  * @size: the size to reallocate
+ * @align: requested alignment
  * @flags: the flags for the page level allocator
+ * @nid: node number of the target node
+ *
+ * If @p is %NULL, vrealloc_XXX() behaves exactly like vmalloc_XXX(). If @size
+ * is 0 and @p is not a %NULL pointer, the object pointed to is freed.
  *
- * If @p is %NULL, vrealloc() behaves exactly like vmalloc(). If @size is 0 and
- * @p is not a %NULL pointer, the object pointed to is freed.
+ * If the caller wants the new memory to be on specific node *only*,
+ * __GFP_THISNODE flag should be set, otherwise the function will try to avoid
+ * reallocation and possibly disregard the specified @nid.
  *
  * If __GFP_ZERO logic is requested, callers must ensure that, starting with the
  * initial memory allocation, every subsequent call to this API for the same
  * memory allocation is flagged with __GFP_ZERO. Otherwise, it is possible that
  * __GFP_ZERO is not fully honored by this API.
  *
+ * Requesting an alignment that is bigger than the alignment of the existing
+ * allocation will fail.
+ *
  * In any case, the contents of the object pointed to are preserved up to the
  * lesser of the new and old sizes.
  *
@@ -4103,7 +4113,8 @@ EXPORT_SYMBOL(vzalloc_node_noprof);
  * Return: pointer to the allocated memory; %NULL if @size is zero or in case of
  *         failure
  */
-void *vrealloc_noprof(const void *p, size_t size, gfp_t flags)
+void *vrealloc_node_align_noprof(const void *p, size_t size, unsigned long align,
+				 gfp_t flags, int nid)
 {
 	struct vm_struct *vm = NULL;
 	size_t alloced_size = 0;
@@ -4127,6 +4138,12 @@ void *vrealloc_noprof(const void *p, size_t size, gfp_t flags)
 		if (WARN(alloced_size < old_size,
 			 "vrealloc() has mismatched area vs requested sizes (%p)\n", p))
 			return NULL;
+		if (WARN(!IS_ALIGNED((unsigned long)p, align),
+			 "will not reallocate with a bigger alignment (0x%lx)\n", align))
+			return NULL;
+		if (unlikely(flags & __GFP_THISNODE) && nid != NUMA_NO_NODE &&
+			     nid != page_to_nid(vmalloc_to_page(p)))
+			goto need_realloc;
 	}
 
 	/*
@@ -4157,8 +4174,10 @@ void *vrealloc_noprof(const void *p, size_t size, gfp_t flags)
 		return (void *)p;
 	}
 
+need_realloc:
 	/* TODO: Grow the vm_area, i.e. allocate and map additional pages. */
-	n = __vmalloc_noprof(size, flags);
+	n = __vmalloc_node_noprof(size, align, flags, nid, __builtin_return_address(0));
+
 	if (!n)
 		return NULL;
 
-- 
2.39.2

[PATCH v13 1/4] mm/vmalloc: allow to set node and align in vrealloc

[Vitaly Wool]

Reimplement vrealloc() to be able to set node and alignment should
a user need to do so. Rename the function to vrealloc_node_align()
to better match what it actually does now and introduce macros for
vrealloc() and friends for backward compatibility.

With that change we also provide the ability for the Rust part of
the kernel to set node and alignment in its allocations.

Signed-off-by: Vitaly Wool <vitaly.wool@konsulko.se>
Reviewed-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
---
 include/linux/vmalloc.h | 12 +++++++++---
 mm/nommu.c              |  3 ++-
 mm/vmalloc.c            | 29 ++++++++++++++++++++++++-----
 3 files changed, 35 insertions(+), 9 deletions(-)

diff --git a/include/linux/vmalloc.h b/include/linux/vmalloc.h
index fdc9aeb74a44..68791f7cb3ba 100644
--- a/include/linux/vmalloc.h
+++ b/include/linux/vmalloc.h
@@ -197,9 +197,15 @@ extern void *__vcalloc_noprof(size_t n, size_t size, gfp_t flags) __alloc_size(1
 extern void *vcalloc_noprof(size_t n, size_t size) __alloc_size(1, 2);
 #define vcalloc(...)		alloc_hooks(vcalloc_noprof(__VA_ARGS__))
 
-void * __must_check vrealloc_noprof(const void *p, size_t size, gfp_t flags)
-		__realloc_size(2);
-#define vrealloc(...)		alloc_hooks(vrealloc_noprof(__VA_ARGS__))
+void *__must_check vrealloc_node_align_noprof(const void *p, size_t size,
+		unsigned long align, gfp_t flags, int nid) __realloc_size(2);
+#define vrealloc_node_noprof(_p, _s, _f, _nid)	\
+	vrealloc_node_align_noprof(_p, _s, 1, _f, _nid)
+#define vrealloc_noprof(_p, _s, _f)		\
+	vrealloc_node_align_noprof(_p, _s, 1, _f, NUMA_NO_NODE)
+#define vrealloc_node_align(...)		alloc_hooks(vrealloc_node_align_noprof(__VA_ARGS__))
+#define vrealloc_node(...)			alloc_hooks(vrealloc_node_noprof(__VA_ARGS__))
+#define vrealloc(...)				alloc_hooks(vrealloc_noprof(__VA_ARGS__))
 
 extern void vfree(const void *addr);
 extern void vfree_atomic(const void *addr);
diff --git a/mm/nommu.c b/mm/nommu.c
index b624acec6d2e..afde6c626b07 100644
--- a/mm/nommu.c
+++ b/mm/nommu.c
@@ -119,7 +119,8 @@ void *__vmalloc_noprof(unsigned long size, gfp_t gfp_mask)
 }
 EXPORT_SYMBOL(__vmalloc_noprof);
 
-void *vrealloc_noprof(const void *p, size_t size, gfp_t flags)
+void *vrealloc_node_align_noprof(const void *p, size_t size, unsigned long align,
+				 gfp_t flags, int node)
 {
 	return krealloc_noprof(p, size, (flags | __GFP_COMP) & ~__GFP_HIGHMEM);
 }
diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index ab986dd09b6a..e0a593651d96 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -4081,19 +4081,29 @@ void *vzalloc_node_noprof(unsigned long size, int node)
 EXPORT_SYMBOL(vzalloc_node_noprof);
 
 /**
- * vrealloc - reallocate virtually contiguous memory; contents remain unchanged
+ * vrealloc_node_align_noprof - reallocate virtually contiguous memory; contents
+ * remain unchanged
  * @p: object to reallocate memory for
  * @size: the size to reallocate
+ * @align: requested alignment
  * @flags: the flags for the page level allocator
+ * @nid: node number of the target node
+ *
+ * If @p is %NULL, vrealloc_XXX() behaves exactly like vmalloc_XXX(). If @size
+ * is 0 and @p is not a %NULL pointer, the object pointed to is freed.
  *
- * If @p is %NULL, vrealloc() behaves exactly like vmalloc(). If @size is 0 and
- * @p is not a %NULL pointer, the object pointed to is freed.
+ * If the caller wants the new memory to be on specific node *only*,
+ * __GFP_THISNODE flag should be set, otherwise the function will try to avoid
+ * reallocation and possibly disregard the specified @nid.
  *
  * If __GFP_ZERO logic is requested, callers must ensure that, starting with the
  * initial memory allocation, every subsequent call to this API for the same
  * memory allocation is flagged with __GFP_ZERO. Otherwise, it is possible that
  * __GFP_ZERO is not fully honored by this API.
  *
+ * Requesting an alignment that is bigger than the alignment of the existing
+ * allocation will fail.
+ *
  * In any case, the contents of the object pointed to are preserved up to the
  * lesser of the new and old sizes.
  *
@@ -4103,7 +4113,8 @@ EXPORT_SYMBOL(vzalloc_node_noprof);
  * Return: pointer to the allocated memory; %NULL if @size is zero or in case of
  *         failure
  */
-void *vrealloc_noprof(const void *p, size_t size, gfp_t flags)
+void *vrealloc_node_align_noprof(const void *p, size_t size, unsigned long align,
+				 gfp_t flags, int nid)
 {
 	struct vm_struct *vm = NULL;
 	size_t alloced_size = 0;
@@ -4127,6 +4138,12 @@ void *vrealloc_noprof(const void *p, size_t size, gfp_t flags)
 		if (WARN(alloced_size < old_size,
 			 "vrealloc() has mismatched area vs requested sizes (%p)\n", p))
 			return NULL;
+		if (WARN(!IS_ALIGNED((unsigned long)p, align),
+			 "will not reallocate with a bigger alignment (0x%lx)\n", align))
+			return NULL;
+		if (unlikely(flags & __GFP_THISNODE) && nid != NUMA_NO_NODE &&
+			     nid != page_to_nid(vmalloc_to_page(p)))
+			goto need_realloc;
 	}
 
 	/*
@@ -4157,8 +4174,10 @@ void *vrealloc_noprof(const void *p, size_t size, gfp_t flags)
 		return (void *)p;
 	}
 
+need_realloc:
 	/* TODO: Grow the vm_area, i.e. allocate and map additional pages. */
-	n = __vmalloc_noprof(size, flags);
+	n = __vmalloc_node_noprof(size, align, flags, nid, __builtin_return_address(0));
+
 	if (!n)
 		return NULL;
 
-- 
2.39.2

[PATCH v13 2/4] mm/slub: allow to set node and align in k[v]realloc

[Vitaly Wool]

Reimplement k[v]realloc_node() to be able to set node and
alignment should a user need to do so. In order to do that while
retaining the maximal backward compatibility, add
k[v]realloc_node_align() functions and redefine the rest of API
using these new ones.

While doing that, we also keep the number of  _noprof variants to a
minimum, which implies some changes to the existing users of older
_noprof functions, that basically being bcachefs.

With that change we also provide the ability for the Rust part of
the kernel to set node and alignment in its K[v]xxx
[re]allocations.

Signed-off-by: Vitaly Wool <vitaly.wool@konsulko.se>
---
 fs/bcachefs/darray.c   |  2 +-
 fs/bcachefs/util.h     |  2 +-
 include/linux/bpfptr.h |  2 +-
 include/linux/slab.h   | 39 +++++++++++++++----------
 lib/rhashtable.c       |  4 +--
 mm/slub.c              | 64 +++++++++++++++++++++++++++++-------------
 6 files changed, 73 insertions(+), 40 deletions(-)

diff --git a/fs/bcachefs/darray.c b/fs/bcachefs/darray.c
index e86d36d23e9e..928e83a1ce42 100644
--- a/fs/bcachefs/darray.c
+++ b/fs/bcachefs/darray.c
@@ -21,7 +21,7 @@ int __bch2_darray_resize_noprof(darray_char *d, size_t element_size, size_t new_
 			return -ENOMEM;
 
 		void *data = likely(bytes < INT_MAX)
-			? kvmalloc_noprof(bytes, gfp)
+			? kvmalloc_node_align_noprof(bytes, 1, gfp, NUMA_NO_NODE)
 			: vmalloc_noprof(bytes);
 		if (!data)
 			return -ENOMEM;
diff --git a/fs/bcachefs/util.h b/fs/bcachefs/util.h
index 0a4b1d433621..2d6d4b547db8 100644
--- a/fs/bcachefs/util.h
+++ b/fs/bcachefs/util.h
@@ -61,7 +61,7 @@ static inline void *bch2_kvmalloc_noprof(size_t n, gfp_t flags)
 {
 	void *p = unlikely(n >= INT_MAX)
 		? vmalloc_noprof(n)
-		: kvmalloc_noprof(n, flags & ~__GFP_ZERO);
+		: kvmalloc_node_align_noprof(n, 1, flags & ~__GFP_ZERO, NUMA_NO_NODE);
 	if (p && (flags & __GFP_ZERO))
 		memset(p, 0, n);
 	return p;
diff --git a/include/linux/bpfptr.h b/include/linux/bpfptr.h
index 1af241525a17..f6e0795db484 100644
--- a/include/linux/bpfptr.h
+++ b/include/linux/bpfptr.h
@@ -67,7 +67,7 @@ static inline int copy_to_bpfptr_offset(bpfptr_t dst, size_t offset,
 
 static inline void *kvmemdup_bpfptr_noprof(bpfptr_t src, size_t len)
 {
-	void *p = kvmalloc_noprof(len, GFP_USER | __GFP_NOWARN);
+	void *p = kvmalloc_node_align_noprof(len, 1, GFP_USER | __GFP_NOWARN, NUMA_NO_NODE);
 
 	if (!p)
 		return ERR_PTR(-ENOMEM);
diff --git a/include/linux/slab.h b/include/linux/slab.h
index d5a8ab98035c..6dc300bac2a1 100644
--- a/include/linux/slab.h
+++ b/include/linux/slab.h
@@ -465,9 +465,13 @@ int kmem_cache_shrink(struct kmem_cache *s);
 /*
  * Common kmalloc functions provided by all allocators
  */
-void * __must_check krealloc_noprof(const void *objp, size_t new_size,
-				    gfp_t flags) __realloc_size(2);
-#define krealloc(...)				alloc_hooks(krealloc_noprof(__VA_ARGS__))
+void * __must_check krealloc_node_align_noprof(const void *objp, size_t new_size,
+					       unsigned long align,
+					       gfp_t flags, int nid) __realloc_size(2);
+#define krealloc_noprof(_o, _s, _f)	krealloc_node_align_noprof(_o, _s, 1, _f, NUMA_NO_NODE)
+#define krealloc_node_align(...)	alloc_hooks(krealloc_node_align_noprof(__VA_ARGS__))
+#define krealloc_node(_o, _s, _f, _n)	krealloc_node_align(_o, _s, 1, _f, _n)
+#define krealloc(...)			krealloc_node(__VA_ARGS__, NUMA_NO_NODE)
 
 void kfree(const void *objp);
 void kfree_sensitive(const void *objp);
@@ -1041,18 +1045,20 @@ static inline __alloc_size(1) void *kzalloc_noprof(size_t size, gfp_t flags)
 #define kzalloc(...)				alloc_hooks(kzalloc_noprof(__VA_ARGS__))
 #define kzalloc_node(_size, _flags, _node)	kmalloc_node(_size, (_flags)|__GFP_ZERO, _node)
 
-void *__kvmalloc_node_noprof(DECL_BUCKET_PARAMS(size, b), gfp_t flags, int node) __alloc_size(1);
-#define kvmalloc_node_noprof(size, flags, node)	\
-	__kvmalloc_node_noprof(PASS_BUCKET_PARAMS(size, NULL), flags, node)
-#define kvmalloc_node(...)			alloc_hooks(kvmalloc_node_noprof(__VA_ARGS__))
-
-#define kvmalloc(_size, _flags)			kvmalloc_node(_size, _flags, NUMA_NO_NODE)
-#define kvmalloc_noprof(_size, _flags)		kvmalloc_node_noprof(_size, _flags, NUMA_NO_NODE)
+void *__kvmalloc_node_noprof(DECL_BUCKET_PARAMS(size, b), unsigned long align,
+			     gfp_t flags, int node) __alloc_size(1);
+#define kvmalloc_node_align_noprof(_size, _align, _flags, _node)	\
+	__kvmalloc_node_noprof(PASS_BUCKET_PARAMS(_size, NULL), _align, _flags, _node)
+#define kvmalloc_node_align(...)		\
+	alloc_hooks(kvmalloc_node_align_noprof(__VA_ARGS__))
+#define kvmalloc_node(_s, _f, _n)		kvmalloc_node_align(_s, 1, _f, _n)
+#define kvmalloc(...)				kvmalloc_node(__VA_ARGS__, NUMA_NO_NODE)
 #define kvzalloc(_size, _flags)			kvmalloc(_size, (_flags)|__GFP_ZERO)
 
 #define kvzalloc_node(_size, _flags, _node)	kvmalloc_node(_size, (_flags)|__GFP_ZERO, _node)
+
 #define kmem_buckets_valloc(_b, _size, _flags)	\
-	alloc_hooks(__kvmalloc_node_noprof(PASS_BUCKET_PARAMS(_size, _b), _flags, NUMA_NO_NODE))
+	alloc_hooks(__kvmalloc_node_noprof(PASS_BUCKET_PARAMS(_size, _b), 1, _flags, NUMA_NO_NODE))
 
 static inline __alloc_size(1, 2) void *
 kvmalloc_array_node_noprof(size_t n, size_t size, gfp_t flags, int node)
@@ -1062,7 +1068,7 @@ kvmalloc_array_node_noprof(size_t n, size_t size, gfp_t flags, int node)
 	if (unlikely(check_mul_overflow(n, size, &bytes)))
 		return NULL;
 
-	return kvmalloc_node_noprof(bytes, flags, node);
+	return kvmalloc_node_align_noprof(bytes, 1, flags, node);
 }
 
 #define kvmalloc_array_noprof(...)		kvmalloc_array_node_noprof(__VA_ARGS__, NUMA_NO_NODE)
@@ -1073,9 +1079,12 @@ kvmalloc_array_node_noprof(size_t n, size_t size, gfp_t flags, int node)
 #define kvcalloc_node(...)			alloc_hooks(kvcalloc_node_noprof(__VA_ARGS__))
 #define kvcalloc(...)				alloc_hooks(kvcalloc_noprof(__VA_ARGS__))
 
-void *kvrealloc_noprof(const void *p, size_t size, gfp_t flags)
-		__realloc_size(2);
-#define kvrealloc(...)				alloc_hooks(kvrealloc_noprof(__VA_ARGS__))
+void *kvrealloc_node_align_noprof(const void *p, size_t size, unsigned long align,
+				  gfp_t flags, int nid) __realloc_size(2);
+#define kvrealloc_node_align(...)		\
+	alloc_hooks(kvrealloc_node_align_noprof(__VA_ARGS__))
+#define kvrealloc_node(_p, _s, _f, _n)		kvrealloc_node_align(_p, _s, 1, _f, _n)
+#define kvrealloc(...)				kvrealloc_node(__VA_ARGS__, NUMA_NO_NODE)
 
 extern void kvfree(const void *addr);
 DEFINE_FREE(kvfree, void *, if (!IS_ERR_OR_NULL(_T)) kvfree(_T))
diff --git a/lib/rhashtable.c b/lib/rhashtable.c
index 3e555d012ed6..fde0f0e556f8 100644
--- a/lib/rhashtable.c
+++ b/lib/rhashtable.c
@@ -184,8 +184,8 @@ static struct bucket_table *bucket_table_alloc(struct rhashtable *ht,
 	static struct lock_class_key __key;
 
 	tbl = alloc_hooks_tag(ht->alloc_tag,
-			kvmalloc_node_noprof(struct_size(tbl, buckets, nbuckets),
-					     gfp|__GFP_ZERO, NUMA_NO_NODE));
+			kvmalloc_node_align_noprof(struct_size(tbl, buckets, nbuckets),
+					     1, gfp|__GFP_ZERO, NUMA_NO_NODE));
 
 	size = nbuckets;
 
diff --git a/mm/slub.c b/mm/slub.c
index 31e11ef256f9..80f7b609a18d 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -4844,7 +4844,7 @@ void kfree(const void *object)
 EXPORT_SYMBOL(kfree);
 
 static __always_inline __realloc_size(2) void *
-__do_krealloc(const void *p, size_t new_size, gfp_t flags)
+__do_krealloc(const void *p, size_t new_size, unsigned long align, gfp_t flags, int nid)
 {
 	void *ret;
 	size_t ks = 0;
@@ -4858,6 +4858,16 @@ __do_krealloc(const void *p, size_t new_size, gfp_t flags)
 	if (!kasan_check_byte(p))
 		return NULL;
 
+	/*
+	 * If reallocation is not necessary (e. g. the new size is less
+	 * than the current allocated size), the current allocation will be
+	 * preserved unless __GFP_THISNODE is set. In the latter case a new
+	 * allocation on the requested node will be attempted.
+	 */
+	if (unlikely(flags & __GFP_THISNODE) && nid != NUMA_NO_NODE &&
+		     nid != page_to_nid(virt_to_page(p)))
+		goto alloc_new;
+
 	if (is_kfence_address(p)) {
 		ks = orig_size = kfence_ksize(p);
 	} else {
@@ -4880,6 +4890,10 @@ __do_krealloc(const void *p, size_t new_size, gfp_t flags)
 	if (new_size > ks)
 		goto alloc_new;
 
+	/* If the old object doesn't satisfy the new alignment, allocate a new one */
+	if (!IS_ALIGNED((unsigned long)p, align))
+		goto alloc_new;
+
 	/* Zero out spare memory. */
 	if (want_init_on_alloc(flags)) {
 		kasan_disable_current();
@@ -4902,7 +4916,7 @@ __do_krealloc(const void *p, size_t new_size, gfp_t flags)
 	return (void *)p;
 
 alloc_new:
-	ret = kmalloc_node_track_caller_noprof(new_size, flags, NUMA_NO_NODE, _RET_IP_);
+	ret = kmalloc_node_track_caller_noprof(new_size, flags, nid, _RET_IP_);
 	if (ret && p) {
 		/* Disable KASAN checks as the object's redzone is accessed. */
 		kasan_disable_current();
@@ -4914,10 +4928,12 @@ __do_krealloc(const void *p, size_t new_size, gfp_t flags)
 }
 
 /**
- * krealloc - reallocate memory. The contents will remain unchanged.
+ * krealloc_node_align - reallocate memory. The contents will remain unchanged.
  * @p: object to reallocate memory for.
  * @new_size: how many bytes of memory are required.
+ * @align: desired alignment.
  * @flags: the type of memory to allocate.
+ * @nid: NUMA node or NUMA_NO_NODE
  *
  * If @p is %NULL, krealloc() behaves exactly like kmalloc().  If @new_size
  * is 0 and @p is not a %NULL pointer, the object pointed to is freed.
@@ -4946,7 +4962,8 @@ __do_krealloc(const void *p, size_t new_size, gfp_t flags)
  *
  * Return: pointer to the allocated memory or %NULL in case of error
  */
-void *krealloc_noprof(const void *p, size_t new_size, gfp_t flags)
+void *krealloc_node_align_noprof(const void *p, size_t new_size, unsigned long align,
+				 gfp_t flags, int nid)
 {
 	void *ret;
 
@@ -4955,13 +4972,13 @@ void *krealloc_noprof(const void *p, size_t new_size, gfp_t flags)
 		return ZERO_SIZE_PTR;
 	}
 
-	ret = __do_krealloc(p, new_size, flags);
+	ret = __do_krealloc(p, new_size, align, flags, nid);
 	if (ret && kasan_reset_tag(p) != kasan_reset_tag(ret))
 		kfree(p);
 
 	return ret;
 }
-EXPORT_SYMBOL(krealloc_noprof);
+EXPORT_SYMBOL(krealloc_node_align_noprof);
 
 static gfp_t kmalloc_gfp_adjust(gfp_t flags, size_t size)
 {
@@ -4992,6 +5009,7 @@ static gfp_t kmalloc_gfp_adjust(gfp_t flags, size_t size)
  * failure, fall back to non-contiguous (vmalloc) allocation.
  * @size: size of the request.
  * @b: which set of kmalloc buckets to allocate from.
+ * @align: desired alignment.
  * @flags: gfp mask for the allocation - must be compatible (superset) with GFP_KERNEL.
  * @node: numa node to allocate from
  *
@@ -5004,19 +5022,22 @@ static gfp_t kmalloc_gfp_adjust(gfp_t flags, size_t size)
  *
  * Return: pointer to the allocated memory of %NULL in case of failure
  */
-void *__kvmalloc_node_noprof(DECL_BUCKET_PARAMS(size, b), gfp_t flags, int node)
+void *__kvmalloc_node_noprof(DECL_BUCKET_PARAMS(size, b), unsigned long align,
+			     gfp_t flags, int node)
 {
 	void *ret;
 
 	/*
 	 * It doesn't really make sense to fallback to vmalloc for sub page
-	 * requests
+	 * requests and small alignments
 	 */
-	ret = __do_kmalloc_node(size, PASS_BUCKET_PARAM(b),
-				kmalloc_gfp_adjust(flags, size),
-				node, _RET_IP_);
-	if (ret || size <= PAGE_SIZE)
-		return ret;
+	if (size >= align) {
+		ret = __do_kmalloc_node(size, PASS_BUCKET_PARAM(b),
+					kmalloc_gfp_adjust(flags, size),
+					node, _RET_IP_);
+		if (ret || size <= PAGE_SIZE)
+			return ret;
+	}
 
 	/* non-sleeping allocations are not supported by vmalloc */
 	if (!gfpflags_allow_blocking(flags))
@@ -5034,7 +5055,7 @@ void *__kvmalloc_node_noprof(DECL_BUCKET_PARAMS(size, b), gfp_t flags, int node)
 	 * about the resulting pointer, and cannot play
 	 * protection games.
 	 */
-	return __vmalloc_node_range_noprof(size, 1, VMALLOC_START, VMALLOC_END,
+	return __vmalloc_node_range_noprof(size, align, VMALLOC_START, VMALLOC_END,
 			flags, PAGE_KERNEL, VM_ALLOW_HUGE_VMAP,
 			node, __builtin_return_address(0));
 }
@@ -5078,10 +5099,12 @@ void kvfree_sensitive(const void *addr, size_t len)
 EXPORT_SYMBOL(kvfree_sensitive);
 
 /**
- * kvrealloc - reallocate memory; contents remain unchanged
+ * kvrealloc_node_align - reallocate memory; contents remain unchanged
  * @p: object to reallocate memory for
  * @size: the size to reallocate
+ * @align: desired alignment
  * @flags: the flags for the page level allocator
+ * @nid: NUMA node id
  *
  * If @p is %NULL, kvrealloc() behaves exactly like kvmalloc(). If @size is 0
  * and @p is not a %NULL pointer, the object pointed to is freed.
@@ -5099,17 +5122,18 @@ EXPORT_SYMBOL(kvfree_sensitive);
  *
  * Return: pointer to the allocated memory or %NULL in case of error
  */
-void *kvrealloc_noprof(const void *p, size_t size, gfp_t flags)
+void *kvrealloc_node_align_noprof(const void *p, size_t size, unsigned long align,
+				  gfp_t flags, int nid)
 {
 	void *n;
 
 	if (is_vmalloc_addr(p))
-		return vrealloc_noprof(p, size, flags);
+		return vrealloc_node_align_noprof(p, size, align, flags, nid);
 
-	n = krealloc_noprof(p, size, kmalloc_gfp_adjust(flags, size));
+	n = krealloc_node_align_noprof(p, size, align, kmalloc_gfp_adjust(flags, size), nid);
 	if (!n) {
 		/* We failed to krealloc(), fall back to kvmalloc(). */
-		n = kvmalloc_noprof(size, flags);
+		n = kvmalloc_node_align_noprof(size, align, flags, nid);
 		if (!n)
 			return NULL;
 
@@ -5125,7 +5149,7 @@ void *kvrealloc_noprof(const void *p, size_t size, gfp_t flags)
 
 	return n;
 }
-EXPORT_SYMBOL(kvrealloc_noprof);
+EXPORT_SYMBOL(kvrealloc_node_align_noprof);
 
 struct detached_freelist {
 	struct slab *slab;
-- 
2.39.2

[PATCH v13 3/4] rust: add support for NUMA ids in allocations

[Vitaly Wool]

Add a new type to support specifying NUMA identifiers in Rust
allocators and extend the allocators to have NUMA id as a
parameter. Thus, modify ReallocFunc to use the new extended realloc
primitives from the C side of the kernel (i. e.
k[v]realloc_node_align/vrealloc_node_align) and add the new function
alloc_node to the Allocator trait while keeping the existing one
(alloc) for backward compatibility.

This will allow to specify node to use for allocation of e. g.
{KV}Box, as well as for future NUMA aware users of the API.

Signed-off-by: Vitaly Wool <vitaly.wool@konsulko.se>
Acked-by: Danilo Krummrich <dakr@kernel.org>
---
 rust/helpers/slab.c            |  8 ++---
 rust/helpers/vmalloc.c         |  4 +--
 rust/kernel/alloc.rs           | 54 ++++++++++++++++++++++++++++++----
 rust/kernel/alloc/allocator.rs | 35 ++++++++++++++--------
 rust/kernel/alloc/kbox.rs      |  4 +--
 rust/kernel/alloc/kvec.rs      | 11 +++++--
 6 files changed, 88 insertions(+), 28 deletions(-)

diff --git a/rust/helpers/slab.c b/rust/helpers/slab.c
index a842bfbddcba..8472370a4338 100644
--- a/rust/helpers/slab.c
+++ b/rust/helpers/slab.c
@@ -3,13 +3,13 @@
 #include <linux/slab.h>
 
 void * __must_check __realloc_size(2)
-rust_helper_krealloc(const void *objp, size_t new_size, gfp_t flags)
+rust_helper_krealloc_node(const void *objp, size_t new_size, gfp_t flags, int node)
 {
-	return krealloc(objp, new_size, flags);
+	return krealloc_node(objp, new_size, flags, node);
 }
 
 void * __must_check __realloc_size(2)
-rust_helper_kvrealloc(const void *p, size_t size, gfp_t flags)
+rust_helper_kvrealloc_node(const void *p, size_t size, gfp_t flags, int node)
 {
-	return kvrealloc(p, size, flags);
+	return kvrealloc_node(p, size, flags, node);
 }
diff --git a/rust/helpers/vmalloc.c b/rust/helpers/vmalloc.c
index 80d34501bbc0..62d30db9a1a6 100644
--- a/rust/helpers/vmalloc.c
+++ b/rust/helpers/vmalloc.c
@@ -3,7 +3,7 @@
 #include <linux/vmalloc.h>
 
 void * __must_check __realloc_size(2)
-rust_helper_vrealloc(const void *p, size_t size, gfp_t flags)
+rust_helper_vrealloc_node(const void *p, size_t size, gfp_t flags, int node)
 {
-	return vrealloc(p, size, flags);
+	return vrealloc_node(p, size, flags, node);
 }
diff --git a/rust/kernel/alloc.rs b/rust/kernel/alloc.rs
index a2c49e5494d3..b39c279236f5 100644
--- a/rust/kernel/alloc.rs
+++ b/rust/kernel/alloc.rs
@@ -28,6 +28,8 @@
 /// Indicates an allocation error.
 #[derive(Copy, Clone, PartialEq, Eq, Debug)]
 pub struct AllocError;
+
+use crate::error::{code::EINVAL, Result};
 use core::{alloc::Layout, ptr::NonNull};
 
 /// Flags to be used when allocating memory.
@@ -115,6 +117,31 @@ pub mod flags {
     pub const __GFP_NOWARN: Flags = Flags(bindings::__GFP_NOWARN);
 }
 
+/// Non Uniform Memory Access (NUMA) node identifier.
+#[derive(Clone, Copy, PartialEq)]
+pub struct NumaNode(i32);
+
+impl NumaNode {
+    /// Create a new NUMA node identifier (non-negative integer).
+    ///
+    /// Returns [`EINVAL`] if a negative id or an id exceeding [`bindings::MAX_NUMNODES`] is
+    /// specified.
+    pub fn new(node: i32) -> Result<Self> {
+        // MAX_NUMNODES never exceeds 2**10 because NODES_SHIFT is 0..10.
+        if node < 0 || node >= bindings::MAX_NUMNODES as i32 {
+            return Err(EINVAL);
+        }
+        Ok(Self(node))
+    }
+}
+
+/// Specify necessary constant to pass the information to Allocator that the caller doesn't care
+/// about the NUMA node to allocate memory from.
+impl NumaNode {
+    /// No node preference.
+    pub const NO_NODE: NumaNode = NumaNode(bindings::NUMA_NO_NODE);
+}
+
 /// The kernel's [`Allocator`] trait.
 ///
 /// An implementation of [`Allocator`] can allocate, re-allocate and free memory buffers described
@@ -137,7 +164,7 @@ pub mod flags {
 /// - Implementers must ensure that all trait functions abide by the guarantees documented in the
 ///   `# Guarantees` sections.
 pub unsafe trait Allocator {
-    /// Allocate memory based on `layout` and `flags`.
+    /// Allocate memory based on `layout`, `flags` and `nid`.
     ///
     /// On success, returns a buffer represented as `NonNull<[u8]>` that satisfies the layout
     /// constraints (i.e. minimum size and alignment as specified by `layout`).
@@ -153,13 +180,21 @@ pub unsafe trait Allocator {
     ///
     /// Additionally, `Flags` are honored as documented in
     /// <https://docs.kernel.org/core-api/mm-api.html#mm-api-gfp-flags>.
-    fn alloc(layout: Layout, flags: Flags) -> Result<NonNull<[u8]>, AllocError> {
+    fn alloc(layout: Layout, flags: Flags, nid: NumaNode) -> Result<NonNull<[u8]>, AllocError> {
         // SAFETY: Passing `None` to `realloc` is valid by its safety requirements and asks for a
         // new memory allocation.
-        unsafe { Self::realloc(None, layout, Layout::new::<()>(), flags) }
+        unsafe { Self::realloc(None, layout, Layout::new::<()>(), flags, nid) }
     }
 
-    /// Re-allocate an existing memory allocation to satisfy the requested `layout`.
+    /// Re-allocate an existing memory allocation to satisfy the requested `layout` and
+    /// a specific NUMA node request to allocate the memory for.
+    ///
+    /// Systems employing a Non Uniform Memory Access (NUMA) architecture contain collections of
+    /// hardware resources including processors, memory, and I/O buses, that comprise what is
+    /// commonly known as a NUMA node.
+    ///
+    /// `nid` stands for NUMA id, i. e. NUMA node identifier, which is a non-negative integer
+    /// if a node needs to be specified, or [`NumaNode::NO_NODE`] if the caller doesn't care.
     ///
     /// If the requested size is zero, `realloc` behaves equivalent to `free`.
     ///
@@ -196,6 +231,7 @@ unsafe fn realloc(
         layout: Layout,
         old_layout: Layout,
         flags: Flags,
+        nid: NumaNode,
     ) -> Result<NonNull<[u8]>, AllocError>;
 
     /// Free an existing memory allocation.
@@ -211,7 +247,15 @@ unsafe fn free(ptr: NonNull<u8>, layout: Layout) {
         // SAFETY: The caller guarantees that `ptr` points at a valid allocation created by this
         // allocator. We are passing a `Layout` with the smallest possible alignment, so it is
         // smaller than or equal to the alignment previously used with this allocation.
-        let _ = unsafe { Self::realloc(Some(ptr), Layout::new::<()>(), layout, Flags(0)) };
+        let _ = unsafe {
+            Self::realloc(
+                Some(ptr),
+                Layout::new::<()>(),
+                layout,
+                Flags(0),
+                NumaNode::NO_NODE,
+            )
+        };
     }
 }
 
diff --git a/rust/kernel/alloc/allocator.rs b/rust/kernel/alloc/allocator.rs
index aa2dfa9dca4c..8af7e04e3cc6 100644
--- a/rust/kernel/alloc/allocator.rs
+++ b/rust/kernel/alloc/allocator.rs
@@ -13,7 +13,7 @@
 use core::ptr;
 use core::ptr::NonNull;
 
-use crate::alloc::{AllocError, Allocator};
+use crate::alloc::{AllocError, Allocator, NumaNode};
 use crate::bindings;
 use crate::pr_warn;
 
@@ -56,20 +56,25 @@ fn aligned_size(new_layout: Layout) -> usize {
 
 /// # Invariants
 ///
-/// One of the following: `krealloc`, `vrealloc`, `kvrealloc`.
+/// One of the following: `krealloc_node`, `vrealloc_node`, `kvrealloc_node`.
 struct ReallocFunc(
-    unsafe extern "C" fn(*const crate::ffi::c_void, usize, u32) -> *mut crate::ffi::c_void,
+    unsafe extern "C" fn(
+        *const crate::ffi::c_void,
+        usize,
+        u32,
+        crate::ffi::c_int,
+    ) -> *mut crate::ffi::c_void,
 );
 
 impl ReallocFunc {
-    // INVARIANT: `krealloc` satisfies the type invariants.
-    const KREALLOC: Self = Self(bindings::krealloc);
+    // INVARIANT: `krealloc_node` satisfies the type invariants.
+    const KREALLOC: Self = Self(bindings::krealloc_node);
 
-    // INVARIANT: `vrealloc` satisfies the type invariants.
-    const VREALLOC: Self = Self(bindings::vrealloc);
+    // INVARIANT: `vrealloc_node` satisfies the type invariants.
+    const VREALLOC: Self = Self(bindings::vrealloc_node);
 
-    // INVARIANT: `kvrealloc` satisfies the type invariants.
-    const KVREALLOC: Self = Self(bindings::kvrealloc);
+    // INVARIANT: `kvrealloc_node` satisfies the type invariants.
+    const KVREALLOC: Self = Self(bindings::kvrealloc_node);
 
     /// # Safety
     ///
@@ -87,6 +92,7 @@ unsafe fn call(
         layout: Layout,
         old_layout: Layout,
         flags: Flags,
+        nid: NumaNode,
     ) -> Result<NonNull<[u8]>, AllocError> {
         let size = aligned_size(layout);
         let ptr = match ptr {
@@ -110,7 +116,7 @@ unsafe fn call(
         // - Those functions provide the guarantees of this function.
         let raw_ptr = unsafe {
             // If `size == 0` and `ptr != NULL` the memory behind the pointer is freed.
-            self.0(ptr.cast(), size, flags.0).cast()
+            self.0(ptr.cast(), size, flags.0, nid.0).cast()
         };
 
         let ptr = if size == 0 {
@@ -134,9 +140,10 @@ unsafe fn realloc(
         layout: Layout,
         old_layout: Layout,
         flags: Flags,
+        nid: NumaNode,
     ) -> Result<NonNull<[u8]>, AllocError> {
         // SAFETY: `ReallocFunc::call` has the same safety requirements as `Allocator::realloc`.
-        unsafe { ReallocFunc::KREALLOC.call(ptr, layout, old_layout, flags) }
+        unsafe { ReallocFunc::KREALLOC.call(ptr, layout, old_layout, flags, nid) }
     }
 }
 
@@ -151,6 +158,7 @@ unsafe fn realloc(
         layout: Layout,
         old_layout: Layout,
         flags: Flags,
+        nid: NumaNode,
     ) -> Result<NonNull<[u8]>, AllocError> {
         // TODO: Support alignments larger than PAGE_SIZE.
         if layout.align() > bindings::PAGE_SIZE {
@@ -160,7 +168,7 @@ unsafe fn realloc(
 
         // SAFETY: If not `None`, `ptr` is guaranteed to point to valid memory, which was previously
         // allocated with this `Allocator`.
-        unsafe { ReallocFunc::VREALLOC.call(ptr, layout, old_layout, flags) }
+        unsafe { ReallocFunc::VREALLOC.call(ptr, layout, old_layout, flags, nid) }
     }
 }
 
@@ -175,6 +183,7 @@ unsafe fn realloc(
         layout: Layout,
         old_layout: Layout,
         flags: Flags,
+        nid: NumaNode,
     ) -> Result<NonNull<[u8]>, AllocError> {
         // TODO: Support alignments larger than PAGE_SIZE.
         if layout.align() > bindings::PAGE_SIZE {
@@ -184,6 +193,6 @@ unsafe fn realloc(
 
         // SAFETY: If not `None`, `ptr` is guaranteed to point to valid memory, which was previously
         // allocated with this `Allocator`.
-        unsafe { ReallocFunc::KVREALLOC.call(ptr, layout, old_layout, flags) }
+        unsafe { ReallocFunc::KVREALLOC.call(ptr, layout, old_layout, flags, nid) }
     }
 }
diff --git a/rust/kernel/alloc/kbox.rs b/rust/kernel/alloc/kbox.rs
index c386ff771d50..5c0b020fb2a4 100644
--- a/rust/kernel/alloc/kbox.rs
+++ b/rust/kernel/alloc/kbox.rs
@@ -4,7 +4,7 @@
 
 #[allow(unused_imports)] // Used in doc comments.
 use super::allocator::{KVmalloc, Kmalloc, Vmalloc};
-use super::{AllocError, Allocator, Flags};
+use super::{AllocError, Allocator, Flags, NumaNode};
 use core::alloc::Layout;
 use core::fmt;
 use core::marker::PhantomData;
@@ -271,7 +271,7 @@ pub fn new(x: T, flags: Flags) -> Result<Self, AllocError> {
     /// ```
     pub fn new_uninit(flags: Flags) -> Result<Box<MaybeUninit<T>, A>, AllocError> {
         let layout = Layout::new::<MaybeUninit<T>>();
-        let ptr = A::alloc(layout, flags)?;
+        let ptr = A::alloc(layout, flags, NumaNode::NO_NODE)?;
 
         // INVARIANT: `ptr` is either a dangling pointer or points to memory allocated with `A`,
         // which is sufficient in size and alignment for storing a `T`.
diff --git a/rust/kernel/alloc/kvec.rs b/rust/kernel/alloc/kvec.rs
index 1a0dd852a468..aa5d27176d9c 100644
--- a/rust/kernel/alloc/kvec.rs
+++ b/rust/kernel/alloc/kvec.rs
@@ -5,7 +5,7 @@
 use super::{
     allocator::{KVmalloc, Kmalloc, Vmalloc},
     layout::ArrayLayout,
-    AllocError, Allocator, Box, Flags,
+    AllocError, Allocator, Box, Flags, NumaNode,
 };
 use core::{
     fmt,
@@ -633,6 +633,7 @@ pub fn reserve(&mut self, additional: usize, flags: Flags) -> Result<(), AllocEr
                 layout.into(),
                 self.layout.into(),
                 flags,
+                NumaNode::NO_NODE,
             )?
         };
 
@@ -1058,7 +1059,13 @@ pub fn collect(self, flags: Flags) -> Vec<T, A> {
             // the type invariant to be smaller than `cap`. Depending on `realloc` this operation
             // may shrink the buffer or leave it as it is.
             ptr = match unsafe {
-                A::realloc(Some(buf.cast()), layout.into(), old_layout.into(), flags)
+                A::realloc(
+                    Some(buf.cast()),
+                    layout.into(),
+                    old_layout.into(),
+                    flags,
+                    NumaNode::NO_NODE,
+                )
             } {
                 // If we fail to shrink, which likely can't even happen, continue with the existing
                 // buffer.
-- 
2.39.2

[PATCH v13 4/4] rust: support large alignments in allocations

[Vitaly Wool]

Add support for large (> PAGE_SIZE) alignments in Rust allocators.
All the preparations on the C side are already done, we just need
to add bindings for <alloc>_node_align() functions and start
using those.

Signed-off-by: Vitaly Wool <vitaly.wool@konsulko.se>
Acked-by: Danilo Krummrich <dakr@kernel.org>
---
 rust/helpers/slab.c            | 10 ++++++----
 rust/helpers/vmalloc.c         |  5 +++--
 rust/kernel/alloc/allocator.rs | 30 +++++++++---------------------
 3 files changed, 18 insertions(+), 27 deletions(-)

diff --git a/rust/helpers/slab.c b/rust/helpers/slab.c
index 8472370a4338..7fac958907b0 100644
--- a/rust/helpers/slab.c
+++ b/rust/helpers/slab.c
@@ -3,13 +3,15 @@
 #include <linux/slab.h>
 
 void * __must_check __realloc_size(2)
-rust_helper_krealloc_node(const void *objp, size_t new_size, gfp_t flags, int node)
+rust_helper_krealloc_node_align(const void *objp, size_t new_size, unsigned long align,
+				gfp_t flags, int node)
 {
-	return krealloc_node(objp, new_size, flags, node);
+	return krealloc_node_align(objp, new_size, align, flags, node);
 }
 
 void * __must_check __realloc_size(2)
-rust_helper_kvrealloc_node(const void *p, size_t size, gfp_t flags, int node)
+rust_helper_kvrealloc_node_align(const void *p, size_t size, unsigned long align,
+				 gfp_t flags, int node)
 {
-	return kvrealloc_node(p, size, flags, node);
+	return kvrealloc_node_align(p, size, align, flags, node);
 }
diff --git a/rust/helpers/vmalloc.c b/rust/helpers/vmalloc.c
index 62d30db9a1a6..7d7f7336b3d2 100644
--- a/rust/helpers/vmalloc.c
+++ b/rust/helpers/vmalloc.c
@@ -3,7 +3,8 @@
 #include <linux/vmalloc.h>
 
 void * __must_check __realloc_size(2)
-rust_helper_vrealloc_node(const void *p, size_t size, gfp_t flags, int node)
+rust_helper_vrealloc_node_align(const void *p, size_t size, unsigned long align,
+				gfp_t flags, int node)
 {
-	return vrealloc_node(p, size, flags, node);
+	return vrealloc_node_align(p, size, align, flags, node);
 }
diff --git a/rust/kernel/alloc/allocator.rs b/rust/kernel/alloc/allocator.rs
index 8af7e04e3cc6..63f271624428 100644
--- a/rust/kernel/alloc/allocator.rs
+++ b/rust/kernel/alloc/allocator.rs
@@ -15,7 +15,6 @@
 
 use crate::alloc::{AllocError, Allocator, NumaNode};
 use crate::bindings;
-use crate::pr_warn;
 
 /// The contiguous kernel allocator.
 ///
@@ -56,25 +55,26 @@ fn aligned_size(new_layout: Layout) -> usize {
 
 /// # Invariants
 ///
-/// One of the following: `krealloc_node`, `vrealloc_node`, `kvrealloc_node`.
+/// One of the following: `krealloc_node_align`, `vrealloc_node_align`, `kvrealloc_node_align`.
 struct ReallocFunc(
     unsafe extern "C" fn(
         *const crate::ffi::c_void,
         usize,
+        crate::ffi::c_ulong,
         u32,
         crate::ffi::c_int,
     ) -> *mut crate::ffi::c_void,
 );
 
 impl ReallocFunc {
-    // INVARIANT: `krealloc_node` satisfies the type invariants.
-    const KREALLOC: Self = Self(bindings::krealloc_node);
+    // INVARIANT: `krealloc_node_align` satisfies the type invariants.
+    const KREALLOC: Self = Self(bindings::krealloc_node_align);
 
-    // INVARIANT: `vrealloc_node` satisfies the type invariants.
-    const VREALLOC: Self = Self(bindings::vrealloc_node);
+    // INVARIANT: `vrealloc_node_align` satisfies the type invariants.
+    const VREALLOC: Self = Self(bindings::vrealloc_node_align);
 
-    // INVARIANT: `kvrealloc_node` satisfies the type invariants.
-    const KVREALLOC: Self = Self(bindings::kvrealloc_node);
+    // INVARIANT: `kvrealloc_node_align` satisfies the type invariants.
+    const KVREALLOC: Self = Self(bindings::kvrealloc_node_align);
 
     /// # Safety
     ///
@@ -116,7 +116,7 @@ unsafe fn call(
         // - Those functions provide the guarantees of this function.
         let raw_ptr = unsafe {
             // If `size == 0` and `ptr != NULL` the memory behind the pointer is freed.
-            self.0(ptr.cast(), size, flags.0, nid.0).cast()
+            self.0(ptr.cast(), size, layout.align(), flags.0, nid.0).cast()
         };
 
         let ptr = if size == 0 {
@@ -160,12 +160,6 @@ unsafe fn realloc(
         flags: Flags,
         nid: NumaNode,
     ) -> Result<NonNull<[u8]>, AllocError> {
-        // TODO: Support alignments larger than PAGE_SIZE.
-        if layout.align() > bindings::PAGE_SIZE {
-            pr_warn!("Vmalloc does not support alignments larger than PAGE_SIZE yet.\n");
-            return Err(AllocError);
-        }
-
         // SAFETY: If not `None`, `ptr` is guaranteed to point to valid memory, which was previously
         // allocated with this `Allocator`.
         unsafe { ReallocFunc::VREALLOC.call(ptr, layout, old_layout, flags, nid) }
@@ -185,12 +179,6 @@ unsafe fn realloc(
         flags: Flags,
         nid: NumaNode,
     ) -> Result<NonNull<[u8]>, AllocError> {
-        // TODO: Support alignments larger than PAGE_SIZE.
-        if layout.align() > bindings::PAGE_SIZE {
-            pr_warn!("KVmalloc does not support alignments larger than PAGE_SIZE yet.\n");
-            return Err(AllocError);
-        }
-
         // SAFETY: If not `None`, `ptr` is guaranteed to point to valid memory, which was previously
         // allocated with this `Allocator`.
         unsafe { ReallocFunc::KVREALLOC.call(ptr, layout, old_layout, flags, nid) }
-- 
2.39.2

Re: [PATCH v13 1/4] mm/vmalloc: allow to set node and align in vrealloc

[Liam R. Howlett]

* Vitaly Wool <vitaly.wool@konsulko.se> [250715 09:58]:
> Reimplement vrealloc() to be able to set node and alignment should
> a user need to do so. Rename the function to vrealloc_node_align()
> to better match what it actually does now and introduce macros for
> vrealloc() and friends for backward compatibility.
> 
> With that change we also provide the ability for the Rust part of
> the kernel to set node and alignment in its allocations.
> 
> Signed-off-by: Vitaly Wool <vitaly.wool@konsulko.se>
> Reviewed-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
> Reviewed-by: Vlastimil Babka <vbabka@suse.cz>

Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>

> ---
>  include/linux/vmalloc.h | 12 +++++++++---
>  mm/nommu.c              |  3 ++-
>  mm/vmalloc.c            | 29 ++++++++++++++++++++++++-----
>  3 files changed, 35 insertions(+), 9 deletions(-)
> 
> diff --git a/include/linux/vmalloc.h b/include/linux/vmalloc.h
> index fdc9aeb74a44..68791f7cb3ba 100644
> --- a/include/linux/vmalloc.h
> +++ b/include/linux/vmalloc.h
> @@ -197,9 +197,15 @@ extern void *__vcalloc_noprof(size_t n, size_t size, gfp_t flags) __alloc_size(1
>  extern void *vcalloc_noprof(size_t n, size_t size) __alloc_size(1, 2);
>  #define vcalloc(...)		alloc_hooks(vcalloc_noprof(__VA_ARGS__))
>  
> -void * __must_check vrealloc_noprof(const void *p, size_t size, gfp_t flags)
> -		__realloc_size(2);
> -#define vrealloc(...)		alloc_hooks(vrealloc_noprof(__VA_ARGS__))
> +void *__must_check vrealloc_node_align_noprof(const void *p, size_t size,
> +		unsigned long align, gfp_t flags, int nid) __realloc_size(2);
> +#define vrealloc_node_noprof(_p, _s, _f, _nid)	\
> +	vrealloc_node_align_noprof(_p, _s, 1, _f, _nid)
> +#define vrealloc_noprof(_p, _s, _f)		\
> +	vrealloc_node_align_noprof(_p, _s, 1, _f, NUMA_NO_NODE)
> +#define vrealloc_node_align(...)		alloc_hooks(vrealloc_node_align_noprof(__VA_ARGS__))
> +#define vrealloc_node(...)			alloc_hooks(vrealloc_node_noprof(__VA_ARGS__))
> +#define vrealloc(...)				alloc_hooks(vrealloc_noprof(__VA_ARGS__))
>  
>  extern void vfree(const void *addr);
>  extern void vfree_atomic(const void *addr);
> diff --git a/mm/nommu.c b/mm/nommu.c
> index b624acec6d2e..afde6c626b07 100644
> --- a/mm/nommu.c
> +++ b/mm/nommu.c
> @@ -119,7 +119,8 @@ void *__vmalloc_noprof(unsigned long size, gfp_t gfp_mask)
>  }
>  EXPORT_SYMBOL(__vmalloc_noprof);
>  
> -void *vrealloc_noprof(const void *p, size_t size, gfp_t flags)
> +void *vrealloc_node_align_noprof(const void *p, size_t size, unsigned long align,
> +				 gfp_t flags, int node)
>  {
>  	return krealloc_noprof(p, size, (flags | __GFP_COMP) & ~__GFP_HIGHMEM);
>  }
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index ab986dd09b6a..e0a593651d96 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -4081,19 +4081,29 @@ void *vzalloc_node_noprof(unsigned long size, int node)
>  EXPORT_SYMBOL(vzalloc_node_noprof);
>  
>  /**
> - * vrealloc - reallocate virtually contiguous memory; contents remain unchanged
> + * vrealloc_node_align_noprof - reallocate virtually contiguous memory; contents
> + * remain unchanged
>   * @p: object to reallocate memory for
>   * @size: the size to reallocate
> + * @align: requested alignment
>   * @flags: the flags for the page level allocator
> + * @nid: node number of the target node
> + *
> + * If @p is %NULL, vrealloc_XXX() behaves exactly like vmalloc_XXX(). If @size
> + * is 0 and @p is not a %NULL pointer, the object pointed to is freed.
>   *
> - * If @p is %NULL, vrealloc() behaves exactly like vmalloc(). If @size is 0 and
> - * @p is not a %NULL pointer, the object pointed to is freed.
> + * If the caller wants the new memory to be on specific node *only*,
> + * __GFP_THISNODE flag should be set, otherwise the function will try to avoid
> + * reallocation and possibly disregard the specified @nid.
>   *
>   * If __GFP_ZERO logic is requested, callers must ensure that, starting with the
>   * initial memory allocation, every subsequent call to this API for the same
>   * memory allocation is flagged with __GFP_ZERO. Otherwise, it is possible that
>   * __GFP_ZERO is not fully honored by this API.
>   *
> + * Requesting an alignment that is bigger than the alignment of the existing
> + * allocation will fail.
> + *
>   * In any case, the contents of the object pointed to are preserved up to the
>   * lesser of the new and old sizes.
>   *
> @@ -4103,7 +4113,8 @@ EXPORT_SYMBOL(vzalloc_node_noprof);
>   * Return: pointer to the allocated memory; %NULL if @size is zero or in case of
>   *         failure
>   */
> -void *vrealloc_noprof(const void *p, size_t size, gfp_t flags)
> +void *vrealloc_node_align_noprof(const void *p, size_t size, unsigned long align,
> +				 gfp_t flags, int nid)
>  {
>  	struct vm_struct *vm = NULL;
>  	size_t alloced_size = 0;
> @@ -4127,6 +4138,12 @@ void *vrealloc_noprof(const void *p, size_t size, gfp_t flags)
>  		if (WARN(alloced_size < old_size,
>  			 "vrealloc() has mismatched area vs requested sizes (%p)\n", p))
>  			return NULL;
> +		if (WARN(!IS_ALIGNED((unsigned long)p, align),
> +			 "will not reallocate with a bigger alignment (0x%lx)\n", align))
> +			return NULL;
> +		if (unlikely(flags & __GFP_THISNODE) && nid != NUMA_NO_NODE &&
> +			     nid != page_to_nid(vmalloc_to_page(p)))
> +			goto need_realloc;
>  	}
>  
>  	/*
> @@ -4157,8 +4174,10 @@ void *vrealloc_noprof(const void *p, size_t size, gfp_t flags)
>  		return (void *)p;
>  	}
>  
> +need_realloc:
>  	/* TODO: Grow the vm_area, i.e. allocate and map additional pages. */
> -	n = __vmalloc_noprof(size, flags);
> +	n = __vmalloc_node_noprof(size, align, flags, nid, __builtin_return_address(0));
> +
>  	if (!n)
>  		return NULL;
>  
> -- 
> 2.39.2
> 

Re: [PATCH v13 2/4] mm/slub: allow to set node and align in k[v]realloc

[Danilo Krummrich]

Hi Kent,

On Tue Jul 15, 2025 at 3:58 PM CEST, Vitaly Wool wrote:
> diff --git a/fs/bcachefs/darray.c b/fs/bcachefs/darray.c
> index e86d36d23e9e..928e83a1ce42 100644
> --- a/fs/bcachefs/darray.c
> +++ b/fs/bcachefs/darray.c
> @@ -21,7 +21,7 @@ int __bch2_darray_resize_noprof(darray_char *d, size_t element_size, size_t new_
>  			return -ENOMEM;
>  
>  		void *data = likely(bytes < INT_MAX)
> -			? kvmalloc_noprof(bytes, gfp)
> +			? kvmalloc_node_align_noprof(bytes, 1, gfp, NUMA_NO_NODE)
>  			: vmalloc_noprof(bytes);
>  		if (!data)
>  			return -ENOMEM;
> diff --git a/fs/bcachefs/util.h b/fs/bcachefs/util.h
> index 0a4b1d433621..2d6d4b547db8 100644
> --- a/fs/bcachefs/util.h
> +++ b/fs/bcachefs/util.h
> @@ -61,7 +61,7 @@ static inline void *bch2_kvmalloc_noprof(size_t n, gfp_t flags)
>  {
>  	void *p = unlikely(n >= INT_MAX)
>  		? vmalloc_noprof(n)
> -		: kvmalloc_noprof(n, flags & ~__GFP_ZERO);
> +		: kvmalloc_node_align_noprof(n, 1, flags & ~__GFP_ZERO, NUMA_NO_NODE);
>  	if (p && (flags & __GFP_ZERO))
>  		memset(p, 0, n);
>  	return p;

I assume this is because kvmalloc(), and hence kvrealloc(), does this:

	/* Don't even allow crazy sizes */
	if (unlikely(size > INT_MAX)) {
		WARN_ON_ONCE(!(flags & __GFP_NOWARN));
		return NULL;
	}

Do we still consider this a "crazy size"? :)

Re: [PATCH v13 1/4] :mm/vmalloc: allow to set node and align in vrealloc

[Matthew Wilcox]

On Tue, Jul 15, 2025 at 03:57:24PM +0200, Vitaly Wool wrote:
> +void *__must_check vrealloc_node_align_noprof(const void *p, size_t size,
> +		unsigned long align, gfp_t flags, int nid) __realloc_size(2);
> +#define vrealloc_node_noprof(_p, _s, _f, _nid)	\
> +	vrealloc_node_align_noprof(_p, _s, 1, _f, _nid)
> +#define vrealloc_noprof(_p, _s, _f)		\
> +	vrealloc_node_align_noprof(_p, _s, 1, _f, NUMA_NO_NODE)
> +#define vrealloc_node_align(...)		alloc_hooks(vrealloc_node_align_noprof(__VA_ARGS__))
> +#define vrealloc_node(...)			alloc_hooks(vrealloc_node_noprof(__VA_ARGS__))
> +#define vrealloc(...)				alloc_hooks(vrealloc_noprof(__VA_ARGS__))

I think we can simplify all of this.

void *__must_check vrealloc_noprof(const void *p, size_t size,
		unsigned long align, gfp_t flags, int nid) __realloc_size(2);
#define vrealloc_node_align(...) \
	alloc_hooks(vrealloc_noprof(__VA_ARGS__))
#define vrealloc_node(p, s, f, nid) \
	alloc_hooks(vrealloc_noprof(p, s, 1, f, nid))
#define vrealloc(p, s, f) \
	alloc_hooks(vrealloc_noprof(p, s, 1, f, NUMA_NO_NODE))

Re: [PATCH v13 2/4] mm/slub: allow to set node and align in k[v]realloc

[Vlastimil Babka]

On 7/15/25 16:33, Danilo Krummrich wrote:
> On Tue Jul 15, 2025 at 3:58 PM CEST, Vitaly Wool wrote:
>> diff --git a/fs/bcachefs/darray.c b/fs/bcachefs/darray.c
>> index e86d36d23e9e..928e83a1ce42 100644
>> --- a/fs/bcachefs/darray.c
>> +++ b/fs/bcachefs/darray.c
>> @@ -21,7 +21,7 @@ int __bch2_darray_resize_noprof(darray_char *d, size_t element_size, size_t new_
>>  			return -ENOMEM;
>>  
>>  		void *data = likely(bytes < INT_MAX)
>> -			? kvmalloc_noprof(bytes, gfp)
>> +			? kvmalloc_node_align_noprof(bytes, 1, gfp, NUMA_NO_NODE)
>>  			: vmalloc_noprof(bytes);
>>  		if (!data)
>>  			return -ENOMEM;
>> diff --git a/fs/bcachefs/util.h b/fs/bcachefs/util.h
>> index 0a4b1d433621..2d6d4b547db8 100644
>> --- a/fs/bcachefs/util.h
>> +++ b/fs/bcachefs/util.h
>> @@ -61,7 +61,7 @@ static inline void *bch2_kvmalloc_noprof(size_t n, gfp_t flags)
>>  {
>>  	void *p = unlikely(n >= INT_MAX)
>>  		? vmalloc_noprof(n)
>> -		: kvmalloc_noprof(n, flags & ~__GFP_ZERO);
>> +		: kvmalloc_node_align_noprof(n, 1, flags & ~__GFP_ZERO, NUMA_NO_NODE);
>>  	if (p && (flags & __GFP_ZERO))
>>  		memset(p, 0, n);
>>  	return p;
> 
> I assume this is because kvmalloc(), and hence kvrealloc(), does this:
> 
> 	/* Don't even allow crazy sizes */
> 	if (unlikely(size > INT_MAX)) {
> 		WARN_ON_ONCE(!(flags & __GFP_NOWARN));
> 		return NULL;
> 	}
> 
> Do we still consider this a "crazy size"? :)

Yeah, with "we" including Linus:
https://lore.kernel.org/all/CAHk-=wi=PrbZnwnvhKEF6UUQNCZdNsUbr+hk-jOWGr-q4Mmz=Q@mail.gmail.com/

Re: [PATCH v13 2/4] mm/slub: allow to set node and align in k[v]realloc

[Danilo Krummrich]

On Tue Jul 15, 2025 at 5:34 PM CEST, Vlastimil Babka wrote:
> On 7/15/25 16:33, Danilo Krummrich wrote:
>> On Tue Jul 15, 2025 at 3:58 PM CEST, Vitaly Wool wrote:
>>> diff --git a/fs/bcachefs/darray.c b/fs/bcachefs/darray.c
>>> index e86d36d23e9e..928e83a1ce42 100644
>>> --- a/fs/bcachefs/darray.c
>>> +++ b/fs/bcachefs/darray.c
>>> @@ -21,7 +21,7 @@ int __bch2_darray_resize_noprof(darray_char *d, size_t element_size, size_t new_
>>>  			return -ENOMEM;
>>>  
>>>  		void *data = likely(bytes < INT_MAX)
>>> -			? kvmalloc_noprof(bytes, gfp)
>>> +			? kvmalloc_node_align_noprof(bytes, 1, gfp, NUMA_NO_NODE)
>>>  			: vmalloc_noprof(bytes);
>>>  		if (!data)
>>>  			return -ENOMEM;
>>> diff --git a/fs/bcachefs/util.h b/fs/bcachefs/util.h
>>> index 0a4b1d433621..2d6d4b547db8 100644
>>> --- a/fs/bcachefs/util.h
>>> +++ b/fs/bcachefs/util.h
>>> @@ -61,7 +61,7 @@ static inline void *bch2_kvmalloc_noprof(size_t n, gfp_t flags)
>>>  {
>>>  	void *p = unlikely(n >= INT_MAX)
>>>  		? vmalloc_noprof(n)
>>> -		: kvmalloc_noprof(n, flags & ~__GFP_ZERO);
>>> +		: kvmalloc_node_align_noprof(n, 1, flags & ~__GFP_ZERO, NUMA_NO_NODE);
>>>  	if (p && (flags & __GFP_ZERO))
>>>  		memset(p, 0, n);
>>>  	return p;
>> 
>> I assume this is because kvmalloc(), and hence kvrealloc(), does this:
>> 
>> 	/* Don't even allow crazy sizes */
>> 	if (unlikely(size > INT_MAX)) {
>> 		WARN_ON_ONCE(!(flags & __GFP_NOWARN));
>> 		return NULL;
>> 	}
>> 
>> Do we still consider this a "crazy size"? :)
>
> Yeah, with "we" including Linus:
> https://lore.kernel.org/all/CAHk-=wi=PrbZnwnvhKEF6UUQNCZdNsUbr+hk-jOWGr-q4Mmz=Q@mail.gmail.com/

I don't know why bcachefs needs this, hence my question. But I agree that this
clearly raises an eyebrow. :)

Re: [PATCH v13 2/4] mm/slub: allow to set node and align in k[v]realloc

[Danilo Krummrich]

On Tue Jul 15, 2025 at 5:48 PM CEST, Danilo Krummrich wrote:
> On Tue Jul 15, 2025 at 5:34 PM CEST, Vlastimil Babka wrote:
>> On 7/15/25 16:33, Danilo Krummrich wrote:
>>> On Tue Jul 15, 2025 at 3:58 PM CEST, Vitaly Wool wrote:
>>>> diff --git a/fs/bcachefs/darray.c b/fs/bcachefs/darray.c
>>>> index e86d36d23e9e..928e83a1ce42 100644
>>>> --- a/fs/bcachefs/darray.c
>>>> +++ b/fs/bcachefs/darray.c
>>>> @@ -21,7 +21,7 @@ int __bch2_darray_resize_noprof(darray_char *d, size_t element_size, size_t new_
>>>>  			return -ENOMEM;
>>>>  
>>>>  		void *data = likely(bytes < INT_MAX)
>>>> -			? kvmalloc_noprof(bytes, gfp)
>>>> +			? kvmalloc_node_align_noprof(bytes, 1, gfp, NUMA_NO_NODE)
>>>>  			: vmalloc_noprof(bytes);
>>>>  		if (!data)
>>>>  			return -ENOMEM;
>>>> diff --git a/fs/bcachefs/util.h b/fs/bcachefs/util.h
>>>> index 0a4b1d433621..2d6d4b547db8 100644
>>>> --- a/fs/bcachefs/util.h
>>>> +++ b/fs/bcachefs/util.h
>>>> @@ -61,7 +61,7 @@ static inline void *bch2_kvmalloc_noprof(size_t n, gfp_t flags)
>>>>  {
>>>>  	void *p = unlikely(n >= INT_MAX)
>>>>  		? vmalloc_noprof(n)
>>>> -		: kvmalloc_noprof(n, flags & ~__GFP_ZERO);
>>>> +		: kvmalloc_node_align_noprof(n, 1, flags & ~__GFP_ZERO, NUMA_NO_NODE);
>>>>  	if (p && (flags & __GFP_ZERO))
>>>>  		memset(p, 0, n);
>>>>  	return p;
>>> 
>>> I assume this is because kvmalloc(), and hence kvrealloc(), does this:
>>> 
>>> 	/* Don't even allow crazy sizes */
>>> 	if (unlikely(size > INT_MAX)) {
>>> 		WARN_ON_ONCE(!(flags & __GFP_NOWARN));
>>> 		return NULL;
>>> 	}
>>> 
>>> Do we still consider this a "crazy size"? :)
>>
>> Yeah, with "we" including Linus:
>> https://lore.kernel.org/all/CAHk-=wi=PrbZnwnvhKEF6UUQNCZdNsUbr+hk-jOWGr-q4Mmz=Q@mail.gmail.com/
>
> I don't know why bcachefs needs this, hence my question. But I agree that this
> clearly raises an eyebrow. :)

I.e. this is me noting that we're considering anything larger to be unreasonable
while having an in-tree user bypassing this check intentionally.

Re: [PATCH v13 3/4] rust: add support for NUMA ids in allocations

[Alice Ryhl]

On Tue, Jul 15, 2025 at 03:58:27PM +0200, Vitaly Wool wrote:
> Add a new type to support specifying NUMA identifiers in Rust
> allocators and extend the allocators to have NUMA id as a
> parameter. Thus, modify ReallocFunc to use the new extended realloc
> primitives from the C side of the kernel (i. e.
> k[v]realloc_node_align/vrealloc_node_align) and add the new function
> alloc_node to the Allocator trait while keeping the existing one
> (alloc) for backward compatibility.
> 
> This will allow to specify node to use for allocation of e. g.
> {KV}Box, as well as for future NUMA aware users of the API.
> 
> Signed-off-by: Vitaly Wool <vitaly.wool@konsulko.se>
> Acked-by: Danilo Krummrich <dakr@kernel.org>

Acked-by: Alice Ryhl <aliceryhl@google.com>

Re: [PATCH v13 4/4] rust: support large alignments in allocations

[Alice Ryhl]

On Tue, Jul 15, 2025 at 03:58:45PM +0200, Vitaly Wool wrote:
> Add support for large (> PAGE_SIZE) alignments in Rust allocators.
> All the preparations on the C side are already done, we just need
> to add bindings for <alloc>_node_align() functions and start
> using those.
> 
> Signed-off-by: Vitaly Wool <vitaly.wool@konsulko.se>
> Acked-by: Danilo Krummrich <dakr@kernel.org>

Acked-by: Alice Ryhl <aliceryhl@google.com>

Re: [PATCH v13 0/4] support large align and nid in Rust allocators

[Andrew Morton]

On Tue, 15 Jul 2025 15:56:45 +0200 Vitaly Wool <vitaly.wool@konsulko.se> wrote:

> The coming patches provide the ability for Rust allocators to set
> NUMA node and large alignment.
> 
> ...
>
>  fs/bcachefs/darray.c           |    2 -
>  fs/bcachefs/util.h             |    2 -
>  include/linux/bpfptr.h         |    2 -
>  include/linux/slab.h           |   39 ++++++++++++++++++++++---------------
>  include/linux/vmalloc.h        |   12 ++++++++---
>  lib/rhashtable.c               |    4 +--
>  mm/nommu.c                     |    3 +-
>  mm/slub.c                      |   64 +++++++++++++++++++++++++++++++++++++++++--------------------
>  mm/vmalloc.c                   |   29 ++++++++++++++++++++++-----
>  rust/helpers/slab.c            |   10 +++++----
>  rust/helpers/vmalloc.c         |    5 ++--
>  rust/kernel/alloc.rs           |   54 ++++++++++++++++++++++++++++++++++++++++++++++-----
>  rust/kernel/alloc/allocator.rs |   49 +++++++++++++++++++++-------------------------
>  rust/kernel/alloc/kbox.rs      |    4 +--
>  rust/kernel/alloc/kvec.rs      |   11 ++++++++--
>  15 files changed, 200 insertions(+), 90 deletions(-)

I assume we're looking for a merge into mm.git?

We're at -rc7 so let's target 6.17.  Please resend around the end of
the upcoming merge window?

Re: [PATCH v13 0/4] support large align and nid in Rust allocators

[Alice Ryhl]

On Thu, Jul 24, 2025 at 10:54 PM Andrew Morton
<akpm@linux-foundation.org> wrote:
>
> On Tue, 15 Jul 2025 15:56:45 +0200 Vitaly Wool <vitaly.wool@konsulko.se> wrote:
>
> > The coming patches provide the ability for Rust allocators to set
> > NUMA node and large alignment.
> >
> > ...
> >
> >  fs/bcachefs/darray.c           |    2 -
> >  fs/bcachefs/util.h             |    2 -
> >  include/linux/bpfptr.h         |    2 -
> >  include/linux/slab.h           |   39 ++++++++++++++++++++++---------------
> >  include/linux/vmalloc.h        |   12 ++++++++---
> >  lib/rhashtable.c               |    4 +--
> >  mm/nommu.c                     |    3 +-
> >  mm/slub.c                      |   64 +++++++++++++++++++++++++++++++++++++++++--------------------
> >  mm/vmalloc.c                   |   29 ++++++++++++++++++++++-----
> >  rust/helpers/slab.c            |   10 +++++----
> >  rust/helpers/vmalloc.c         |    5 ++--
> >  rust/kernel/alloc.rs           |   54 ++++++++++++++++++++++++++++++++++++++++++++++-----
> >  rust/kernel/alloc/allocator.rs |   49 +++++++++++++++++++++-------------------------
> >  rust/kernel/alloc/kbox.rs      |    4 +--
> >  rust/kernel/alloc/kvec.rs      |   11 ++++++++--
> >  15 files changed, 200 insertions(+), 90 deletions(-)
>
> I assume we're looking for a merge into mm.git?
>
> We're at -rc7 so let's target 6.17.  Please resend around the end of
> the upcoming merge window?

I think it would make sense for this to land through mm.git, so yes
that sounds like a good plan.

Alice

Re: [PATCH v13 0/4] support large align and nid in Rust allocators

[Vlastimil Babka]

On 7/25/25 09:14, Alice Ryhl wrote:
> On Thu, Jul 24, 2025 at 10:54 PM Andrew Morton
> <akpm@linux-foundation.org> wrote:
>>
>> On Tue, 15 Jul 2025 15:56:45 +0200 Vitaly Wool <vitaly.wool@konsulko.se> wrote:
>>
>> > The coming patches provide the ability for Rust allocators to set
>> > NUMA node and large alignment.
>> >
>> > ...
>> >
>> >  fs/bcachefs/darray.c           |    2 -
>> >  fs/bcachefs/util.h             |    2 -
>> >  include/linux/bpfptr.h         |    2 -
>> >  include/linux/slab.h           |   39 ++++++++++++++++++++++---------------
>> >  include/linux/vmalloc.h        |   12 ++++++++---
>> >  lib/rhashtable.c               |    4 +--
>> >  mm/nommu.c                     |    3 +-
>> >  mm/slub.c                      |   64 +++++++++++++++++++++++++++++++++++++++++--------------------
>> >  mm/vmalloc.c                   |   29 ++++++++++++++++++++++-----
>> >  rust/helpers/slab.c            |   10 +++++----
>> >  rust/helpers/vmalloc.c         |    5 ++--
>> >  rust/kernel/alloc.rs           |   54 ++++++++++++++++++++++++++++++++++++++++++++++-----
>> >  rust/kernel/alloc/allocator.rs |   49 +++++++++++++++++++++-------------------------
>> >  rust/kernel/alloc/kbox.rs      |    4 +--
>> >  rust/kernel/alloc/kvec.rs      |   11 ++++++++--
>> >  15 files changed, 200 insertions(+), 90 deletions(-)
>>
>> I assume we're looking for a merge into mm.git?
>>
>> We're at -rc7 so let's target 6.17.  Please resend around the end of
>> the upcoming merge window?
> 
> I think it would make sense for this to land through mm.git, so yes
> that sounds like a good plan.

Ack.

> Alice

Re: [PATCH v13 2/4] mm/slub: allow to set node and align in k[v]realloc

[Vlastimil Babka]

On 7/15/25 15:58, Vitaly Wool wrote:
> Reimplement k[v]realloc_node() to be able to set node and
> alignment should a user need to do so. In order to do that while
> retaining the maximal backward compatibility, add
> k[v]realloc_node_align() functions and redefine the rest of API
> using these new ones.
> 
> While doing that, we also keep the number of  _noprof variants to a
> minimum, which implies some changes to the existing users of older
> _noprof functions, that basically being bcachefs.
> 
> With that change we also provide the ability for the Rust part of
> the kernel to set node and alignment in its K[v]xxx
> [re]allocations.
> 
> Signed-off-by: Vitaly Wool <vitaly.wool@konsulko.se>
>  
>  /**
> - * krealloc - reallocate memory. The contents will remain unchanged.
> + * krealloc_node_align - reallocate memory. The contents will remain unchanged.
>   * @p: object to reallocate memory for.
>   * @new_size: how many bytes of memory are required.
> + * @align: desired alignment.

It should be better noted that only alignments up to those guaranteed by
kmalloc() (with a link to its doc where it's described) are expected and not
arbitrary ones. So we don't give the wrong impression here.

>   * @flags: the type of memory to allocate.
> + * @nid: NUMA node or NUMA_NO_NODE
>   *
>   * If @p is %NULL, krealloc() behaves exactly like kmalloc().  If @new_size
>   * is 0 and @p is not a %NULL pointer, the object pointed to is freed.
> @@ -4946,7 +4962,8 @@ __do_krealloc(const void *p, size_t new_size, gfp_t flags)
>   *
>   * Return: pointer to the allocated memory or %NULL in case of error
>   */
> -void *krealloc_noprof(const void *p, size_t new_size, gfp_t flags)
> +void *krealloc_node_align_noprof(const void *p, size_t new_size, unsigned long align,
> +				 gfp_t flags, int nid)
>  {
>  	void *ret;
>  
> @@ -4955,13 +4972,13 @@ void *krealloc_noprof(const void *p, size_t new_size, gfp_t flags)
>  		return ZERO_SIZE_PTR;
>  	}
>  
> -	ret = __do_krealloc(p, new_size, flags);
> +	ret = __do_krealloc(p, new_size, align, flags, nid);
>  	if (ret && kasan_reset_tag(p) != kasan_reset_tag(ret))
>  		kfree(p);
>  
>  	return ret;
>  }
> -EXPORT_SYMBOL(krealloc_noprof);
> +EXPORT_SYMBOL(krealloc_node_align_noprof);
>  
>  static gfp_t kmalloc_gfp_adjust(gfp_t flags, size_t size)
>  {
> @@ -4992,6 +5009,7 @@ static gfp_t kmalloc_gfp_adjust(gfp_t flags, size_t size)
>   * failure, fall back to non-contiguous (vmalloc) allocation.
>   * @size: size of the request.
>   * @b: which set of kmalloc buckets to allocate from.
> + * @align: desired alignment.

I guess here we should say something similar to what I suggested to
krealloc(). The "size >= align" check below doesn't mean the alignment can
be arbitrary. It would be incompatible with how
kvrealloc_node_align_noprof() works anyway.

>   * @flags: gfp mask for the allocation - must be compatible (superset) with GFP_KERNEL.
>   * @node: numa node to allocate from
>   *
> @@ -5004,19 +5022,22 @@ static gfp_t kmalloc_gfp_adjust(gfp_t flags, size_t size)
>   *
>   * Return: pointer to the allocated memory of %NULL in case of failure
>   */
> -void *__kvmalloc_node_noprof(DECL_BUCKET_PARAMS(size, b), gfp_t flags, int node)
> +void *__kvmalloc_node_noprof(DECL_BUCKET_PARAMS(size, b), unsigned long align,
> +			     gfp_t flags, int node)
>  {
>  	void *ret;
>  
>  	/*
>  	 * It doesn't really make sense to fallback to vmalloc for sub page
> -	 * requests
> +	 * requests and small alignments
>  	 */
> -	ret = __do_kmalloc_node(size, PASS_BUCKET_PARAM(b),
> -				kmalloc_gfp_adjust(flags, size),
> -				node, _RET_IP_);
> -	if (ret || size <= PAGE_SIZE)
> -		return ret;
> +	if (size >= align) {

I think it's similar to the check in __do_krealloc() in v12 and not
particularly useful. We don't need to support align > size for rust anyway?
Does vmalloc even honor that?

It would also be inconsistent that kvmalloc() would give you this
possibility of size<align, but starting with a small size and size>=align
and then kvrealloc() to size<align wouldn't actually work.

> +		ret = __do_kmalloc_node(size, PASS_BUCKET_PARAM(b),
> +					kmalloc_gfp_adjust(flags, size),
> +					node, _RET_IP_);
> +		if (ret || size <= PAGE_SIZE)
> +			return ret;
> +	}
>  
>  	/* non-sleeping allocations are not supported by vmalloc */
>  	if (!gfpflags_allow_blocking(flags))
> @@ -5034,7 +5055,7 @@ void *__kvmalloc_node_noprof(DECL_BUCKET_PARAMS(size, b), gfp_t flags, int node)
>  	 * about the resulting pointer, and cannot play
>  	 * protection games.
>  	 */
> -	return __vmalloc_node_range_noprof(size, 1, VMALLOC_START, VMALLOC_END,
> +	return __vmalloc_node_range_noprof(size, align, VMALLOC_START, VMALLOC_END,
>  			flags, PAGE_KERNEL, VM_ALLOW_HUGE_VMAP,
>  			node, __builtin_return_address(0));
>  }
> @@ -5078,10 +5099,12 @@ void kvfree_sensitive(const void *addr, size_t len)
>  EXPORT_SYMBOL(kvfree_sensitive);
>  
>  /**
> - * kvrealloc - reallocate memory; contents remain unchanged
> + * kvrealloc_node_align - reallocate memory; contents remain unchanged
>   * @p: object to reallocate memory for
>   * @size: the size to reallocate
> + * @align: desired alignment

Ditto.

>   * @flags: the flags for the page level allocator
> + * @nid: NUMA node id
>   *
>   * If @p is %NULL, kvrealloc() behaves exactly like kvmalloc(). If @size is 0
>   * and @p is not a %NULL pointer, the object pointed to is freed.
> @@ -5099,17 +5122,18 @@ EXPORT_SYMBOL(kvfree_sensitive);
>   *
>   * Return: pointer to the allocated memory or %NULL in case of error
>   */
> -void *kvrealloc_noprof(const void *p, size_t size, gfp_t flags)
> +void *kvrealloc_node_align_noprof(const void *p, size_t size, unsigned long align,
> +				  gfp_t flags, int nid)
>  {
>  	void *n;
>  
>  	if (is_vmalloc_addr(p))
> -		return vrealloc_noprof(p, size, flags);
> +		return vrealloc_node_align_noprof(p, size, align, flags, nid);
>  
> -	n = krealloc_noprof(p, size, kmalloc_gfp_adjust(flags, size));
> +	n = krealloc_node_align_noprof(p, size, align, kmalloc_gfp_adjust(flags, size), nid);
>  	if (!n) {
>  		/* We failed to krealloc(), fall back to kvmalloc(). */
> -		n = kvmalloc_noprof(size, flags);
> +		n = kvmalloc_node_align_noprof(size, align, flags, nid);
>  		if (!n)
>  			return NULL;
>  
> @@ -5125,7 +5149,7 @@ void *kvrealloc_noprof(const void *p, size_t size, gfp_t flags)
>  
>  	return n;
>  }
> -EXPORT_SYMBOL(kvrealloc_noprof);
> +EXPORT_SYMBOL(kvrealloc_node_align_noprof);
>  
>  struct detached_freelist {
>  	struct slab *slab;

Re: [PATCH v13 0/4] support large align and nid in Rust allocators

[Danilo Krummrich]

(Cc: Hui)

On Thu Jul 24, 2025 at 10:54 PM CEST, Andrew Morton wrote:
> On Tue, 15 Jul 2025 15:56:45 +0200 Vitaly Wool <vitaly.wool@konsulko.se> wrote:
>
>> The coming patches provide the ability for Rust allocators to set
>> NUMA node and large alignment.
>> 
>> ...
>>
>>  fs/bcachefs/darray.c           |    2 -
>>  fs/bcachefs/util.h             |    2 -
>>  include/linux/bpfptr.h         |    2 -
>>  include/linux/slab.h           |   39 ++++++++++++++++++++++---------------
>>  include/linux/vmalloc.h        |   12 ++++++++---
>>  lib/rhashtable.c               |    4 +--
>>  mm/nommu.c                     |    3 +-
>>  mm/slub.c                      |   64 +++++++++++++++++++++++++++++++++++++++++--------------------
>>  mm/vmalloc.c                   |   29 ++++++++++++++++++++++-----
>>  rust/helpers/slab.c            |   10 +++++----
>>  rust/helpers/vmalloc.c         |    5 ++--
>>  rust/kernel/alloc.rs           |   54 ++++++++++++++++++++++++++++++++++++++++++++++-----
>>  rust/kernel/alloc/allocator.rs |   49 +++++++++++++++++++++-------------------------
>>  rust/kernel/alloc/kbox.rs      |    4 +--
>>  rust/kernel/alloc/kvec.rs      |   11 ++++++++--
>>  15 files changed, 200 insertions(+), 90 deletions(-)
>
> I assume we're looking for a merge into mm.git?

Yes, I think that's what we agreed in v1 -- the bits I maintain should carry my
Acked-by already.

> We're at -rc7 so let's target 6.17.  Please resend around the end of
> the upcoming merge window?

Yes, this is too late for this cycle. Given that we target the next one, there's
a patch in my queue [1] that interacts with this series.

It would be good if you could pick up [1] (once ready) after this series lands.
I asked Hui to rebase onto this series and Cc you for subsequent submissions for
this purpose.

[1] https://lore.kernel.org/lkml/da9b2afca02124ec14fc9ac7f2a2a85e5be96bc7.1753423953.git.zhuhui@kylinos.cn/

Re: [PATCH v13 3/4] rust: add support for NUMA ids in allocations

[Danilo Krummrich]

On Tue Jul 15, 2025 at 3:58 PM CEST, Vitaly Wool wrote:
>  pub unsafe trait Allocator {
> -    /// Allocate memory based on `layout` and `flags`.
> +    /// Allocate memory based on `layout`, `flags` and `nid`.
>      ///
>      /// On success, returns a buffer represented as `NonNull<[u8]>` that satisfies the layout
>      /// constraints (i.e. minimum size and alignment as specified by `layout`).
> @@ -153,13 +180,21 @@ pub unsafe trait Allocator {
>      ///
>      /// Additionally, `Flags` are honored as documented in
>      /// <https://docs.kernel.org/core-api/mm-api.html#mm-api-gfp-flags>.
> -    fn alloc(layout: Layout, flags: Flags) -> Result<NonNull<[u8]>, AllocError> {
> +    fn alloc(layout: Layout, flags: Flags, nid: NumaNode) -> Result<NonNull<[u8]>, AllocError> {
>          // SAFETY: Passing `None` to `realloc` is valid by its safety requirements and asks for a
>          // new memory allocation.
> -        unsafe { Self::realloc(None, layout, Layout::new::<()>(), flags) }
> +        unsafe { Self::realloc(None, layout, Layout::new::<()>(), flags, nid) }
>      }
>  
> -    /// Re-allocate an existing memory allocation to satisfy the requested `layout`.
> +    /// Re-allocate an existing memory allocation to satisfy the requested `layout` and
> +    /// a specific NUMA node request to allocate the memory for.
> +    ///
> +    /// Systems employing a Non Uniform Memory Access (NUMA) architecture contain collections of
> +    /// hardware resources including processors, memory, and I/O buses, that comprise what is
> +    /// commonly known as a NUMA node.
> +    ///
> +    /// `nid` stands for NUMA id, i. e. NUMA node identifier, which is a non-negative integer
> +    /// if a node needs to be specified, or [`NumaNode::NO_NODE`] if the caller doesn't care.
>      ///
>      /// If the requested size is zero, `realloc` behaves equivalent to `free`.
>      ///
> @@ -196,6 +231,7 @@ unsafe fn realloc(
>          layout: Layout,
>          old_layout: Layout,
>          flags: Flags,
> +        nid: NumaNode,
>      ) -> Result<NonNull<[u8]>, AllocError>;
>  
>      /// Free an existing memory allocation.
> @@ -211,7 +247,15 @@ unsafe fn free(ptr: NonNull<u8>, layout: Layout) {
>          // SAFETY: The caller guarantees that `ptr` points at a valid allocation created by this
>          // allocator. We are passing a `Layout` with the smallest possible alignment, so it is
>          // smaller than or equal to the alignment previously used with this allocation.
> -        let _ = unsafe { Self::realloc(Some(ptr), Layout::new::<()>(), layout, Flags(0)) };
> +        let _ = unsafe {
> +            Self::realloc(
> +                Some(ptr),
> +                Layout::new::<()>(),
> +                layout,
> +                Flags(0),
> +                NumaNode::NO_NODE,
> +            )
> +        };
>      }
>  }

Regarding the change in the Allocator trait, we also have to consider
the Cmalloc allocator in rust/kernel/alloc/allocator_test.rs, which is there to
support userspace tests.

While we're planning to remove this (see also [1]), we still have to consider
it for now.

[1] https://lore.kernel.org/rust-for-linux/20250726180750.2735836-1-ojeda@kernel.org/

Re: [PATCH v13 1/4] :mm/vmalloc: allow to set node and align in vrealloc

[Vitaly Wool]

> On Jul 15, 2025, at 4:37 PM, Matthew Wilcox <willy@infradead.org> wrote:
> 
> On Tue, Jul 15, 2025 at 03:57:24PM +0200, Vitaly Wool wrote:
>> +void *__must_check vrealloc_node_align_noprof(const void *p, size_t size,
>> + unsigned long align, gfp_t flags, int nid) __realloc_size(2);
>> +#define vrealloc_node_noprof(_p, _s, _f, _nid) \
>> + vrealloc_node_align_noprof(_p, _s, 1, _f, _nid)
>> +#define vrealloc_noprof(_p, _s, _f) \
>> + vrealloc_node_align_noprof(_p, _s, 1, _f, NUMA_NO_NODE)
>> +#define vrealloc_node_align(...) alloc_hooks(vrealloc_node_align_noprof(__VA_ARGS__))
>> +#define vrealloc_node(...) alloc_hooks(vrealloc_node_noprof(__VA_ARGS__))
>> +#define vrealloc(...) alloc_hooks(vrealloc_noprof(__VA_ARGS__))
> 
> I think we can simplify all of this.
> 
> void *__must_check vrealloc_noprof(const void *p, size_t size,
> unsigned long align, gfp_t flags, int nid) __realloc_size(2);
> #define vrealloc_node_align(...) \
> alloc_hooks(vrealloc_noprof(__VA_ARGS__))
> #define vrealloc_node(p, s, f, nid) \
> alloc_hooks(vrealloc_noprof(p, s, 1, f, nid))
> #define vrealloc(p, s, f) \
> alloc_hooks(vrealloc_noprof(p, s, 1, f, NUMA_NO_NODE))
> 
> 

In this case, to keep things buildable an each step we will need to modify slub.c in this patch. Since we change slub.c in the next patch in the series I would suggest that we keep things simple (== as they are now, even if it means some redundant macros have to stay). I can come up with a macro simplification like yours when this series is accepted.

Thanks,
Vitaly