fixup! fix missing vmalloc.h includes
Subsystems (5)
Suren Baghdasaryan
maintainerPATCH 1/1 fixup! fix missing vmalloc.h includes
vmalloc.h header refactoring missed vmalloc.h inclusion required for __vmalloc_start_set symbol. Add missing inclusion.
--- arch/x86/mm/numa_32.c | 1 + 1 file changed, 1 insertion(+)
arch/x86/mm/numa_32.c +1 -0
@@ -24,6 +24,7 @@#include <linux/memblock.h> #include <linux/init.h> +#include <linux/vmalloc.h> #include "numa_internal.h"
base-commit: d4cd6840d1dc25963aa10ef5e5b1d01876baebf2
Signature · 1 line
2.44.0.478.gd926399ef9-goog
Suren Baghdasaryan
maintainerPATCH 1/1 lib: do limited memory accounting for modules with ARCH_NEEDS_WEAK_PER_CPU
ARCH_NEEDS_WEAK_PER_CPU does not allow percpu variable definitions inside a function, therefore memory allocation profiling can't use it. This definition is used only for modules, so we still can account core kernel allocations and for modules we can do limited allocation accounting by charging all of them to a single counter. This is not ideal but better than no accounting at all.
--- include/linux/alloc_tag.h | 14 ++++++++++++-- lib/alloc_tag.c | 3 +++ 2 files changed, 15 insertions(+), 2 deletions(-)
include/linux/alloc_tag.h +12 -2
@@ -68,9 +68,17 @@ static inline struct alloc_tag *ct_to_alloc_tag(struct codetag *ct)/* * When percpu variables are required to be defined as weak, static percpu * variables can't be used inside a function (see comments for DECLARE_PER_CPU_SECTION). +* Instead we will accound all module allocations to a single counter. */ -#error "Memory allocation profiling is incompatible with ARCH_NEEDS_WEAK_PER_CPU" -#endif +DECLARE_PER_CPU(struct alloc_tag_counters, _shared_alloc_tag); + +#define DEFINE_ALLOC_TAG (_alloc_tag) \ + static struct alloc_tag _alloc_tag __used __aligned(8) \ + __section("alloc_tags") = { \ + .ct = CODE_TAG_INIT, \ + .counters = &_shared_alloc_tag }; + +#else /* ARCH_NEEDS_WEAK_PER_CPU */ #define DEFINE_ALLOC_TAG (_alloc_tag) \ static DEFINE_PER_CPU(struct alloc_tag_counters, _alloc_tag_cntr); \
@@ -79,6 +87,8 @@ static inline struct alloc_tag *ct_to_alloc_tag(struct codetag *ct) .ct = CODE_TAG_INIT, \ .counters = &_alloc_tag_cntr }; +#endif /* ARCH_NEEDS_WEAK_PER_CPU */ +DECLARE_STATIC_KEY_MAYBE (CONFIG_MEM_ALLOC_PROFILING_ENABLED_BY_DEFAULT , mem_alloc_profiling_key);
lib/alloc_tag.c +3 -0
@@ -10,6 +10,9 @@static struct codetag_type *alloc_tag_cttype; +DEFINE_PER_CPU(struct alloc_tag_counters, _shared_alloc_tag); +EXPORT_SYMBOL (_shared_alloc_tag); +DEFINE_STATIC_KEY_MAYBE (CONFIG_MEM_ALLOC_PROFILING_ENABLED_BY_DEFAULT , mem_alloc_profiling_key);
base-commit: d4cd6840d1dc25963aa10ef5e5b1d01876baebf2
Signature · 1 line
2.44.0.478.gd926399ef9-goog
Andrew Morton
maintainerRe: [PATCH 1/1] lib: do limited memory accounting for modules with ARCH_NEEDS_WEAK_PER_CPU
Quoted 5 lines · On Tue, 2 Apr 2024 11:09:33 -0700 Suren Baghdasaryan <[email protected]> wrote:
ARCH_NEEDS_WEAK_PER_CPUdoes not allow percpu variable definitions inside a function, therefore memory allocation profiling can't use it. This definition is used only for modules, so we still can account core kernel allocations and for modules we can do limited allocation accounting by charging all of them to a single counter. This is not ideal but better
I'll queue this as a to-be-squashed fix against "lib: add allocation tagging support for memory allocation profiling", OK?
Suren Baghdasaryan
maintainerRe: [PATCH 1/1] lib: do limited memory accounting for modules with ARCH_NEEDS_WEAK_PER_CPU
Quoted 11 lines · Andrew Morton
Quoted 5 lines · On Tue, 2 Apr 2024 11:09:33 -0700 Suren Baghdasaryan <[email protected]> wrote:
ARCH_NEEDS_WEAK_PER_CPUdoes not allow percpu variable definitions inside a function, therefore memory allocation profiling can't use it. This definition is used only for modules, so we still can account core kernel allocations and for modules we can do limited allocation accounting by charging all of them to a single counter. This is not ideal but betterI'll queue this as a to-be-squashed fix against "lib: add allocation tagging support for memory allocation profiling", OK?
Yes, that would be ideal. Thank you!