On Sun, Sep 27, 2026 at 7:10 PM Hao Ge <[email protected]> wrote:
>
> Hi Suren
>
> On 2026/9/25 14:14, Suren Baghdasaryan wrote:
> > On Thu, Sep 24, 2026 at 12:21 AM Hao Ge <[email protected]> wrote:
> >>
> >> Hi Abhishek
> >>
> >> Sorry, I merged and tested your patches today and found a few details,
> >> so I've started a separate mail thread for this review.
> >>
> >> On 2026/9/22 05:26, Abhishek Bapat wrote:
> >>> The memory allocation profiling framework intercepts allocations across
> >>> the core subsystems, but currently lacks runtime tracing hooks for
> >>> standard observability tools to dynamically track the context (stack
> >>> traces and lifecycles of the individual memory chunks) of the
> >>> allocations made.
> >>>
> >>> Introduce three standard trace events to allow this tracking:
> >>>
> >>> 1. `alloc_tag_hit`: Fired at the exact call site. This allows userspace
> >>> tools to trigger and capture a call stack.
> >>> 2. `alloc_tag_mem_alloced`: Fired in alloc_tag_add upon successful
> >>> allocation. It records the allocated size, the tag, and the uniquely
> >>> generated codetag_ref metadata pointer.
> >>> 3. `alloc_tag_mem_freed`: Fired in alloc_tag_sub right before memory is
> >>> freed, yielding the same codetag_ref to allow tracing tools to find
> >>> the corresponding allocation.
> >>>
> >>> Because the introduced trace events occur at different stages in the
> >>> call stack, userspace tracing tools must stitch them together to form a
> >>> complete picture of a buffer's lifetime. Here's an example of how
> >>> userspace correlates these three events:
> >>>
> >>> 1. On `alloc_tag_hit`: The tool captures the stack trace and caches it,
> >>> keyed by the combination of the current thread's PID and the `tag`.
> >>> 2. On `alloc_tag_mem_alloced`: The tool extracts the PID and `tag` from
> >>> the event and looks up the stack trace cached in step 1. It creates a
> >>> new active allocation record, mapping the new provided `codetag_ref`
> >>> to this cached stack trace and the newly returned allocation size.
> >>> 3. On `alloc_tag_mem_freed`: When the memory is freed, the event yields
> >>> the same `codetag_ref`. The tool uses this reference to look up the
> >>> original allocation record, correlates the free, and safely retires
> >>> the tracking entry.
> >>>
> >>> Also, introduce `alloc_tag_trace_key` static key to minimize the
> >>> overhead when no tags are being traced (the usual case). Once tracing
> >>> for any tag is requested, the key is set, opening the path to check
> >>> whether tracing is enabled for the current tag.
> >>> Nore that the mechanism to enabl tag tracing is implemented in the next
> >>> patch, therefore for now, `alloc_tag_trace_key` stays always unset.
> >>>
> >>> Signed-off-by: Abhishek Bapat <[email protected]>
> >>> ---
> >>> MAINTAINERS | 1 +
> >>> include/linux/alloc_tag.h | 57 ++++++++++++---
> >>> include/trace/events/alloc_tag.h | 122 +++++++++++++++++++++++++++++++
> >>> mm/alloc_tag.c | 28 +++++++
> >>> 4 files changed, 196 insertions(+), 12 deletions(-)
> >>> create mode 100644 include/trace/events/alloc_tag.h
> >>>
> >>> diff --git a/MAINTAINERS b/MAINTAINERS
> >>> index 24420a8c06d0..29e1f7915cb9 100644
> >>> --- a/MAINTAINERS
> >>> +++ b/MAINTAINERS
> >>> @@ -17096,6 +17096,7 @@ S: Maintained
> >>> F: Documentation/mm/allocation-profiling.rst
> >>> F: include/linux/alloc_tag.h
> >>> F: include/linux/pgalloc_tag.h
> >>> +F: include/trace/events/alloc_tag.h
> >>> F: include/uapi/linux/alloc_tag.h
> >>> F: mm/alloc_tag.c
> >>> F: tools/testing/selftests/alloc_tag/
> >>> diff --git a/include/linux/alloc_tag.h b/include/linux/alloc_tag.h
> >>> index 7f2d80a59792..2994934cf44a 100644
> >>> --- a/include/linux/alloc_tag.h
> >>> +++ b/include/linux/alloc_tag.h
> >>> @@ -128,12 +128,33 @@ DECLARE_PER_CPU(struct alloc_tag_counters,
> >>> _shared_alloc_tag);
> >>> DECLARE_STATIC_KEY_MAYBE(CONFIG_MEM_ALLOC_PROFILING_ENABLED_BY_DEFAULT,
> >>> mem_alloc_profiling_key);
> >>>
> >>> +DECLARE_STATIC_KEY_FALSE(alloc_tag_trace_key);
> >>> +
> >>> static inline bool mem_alloc_profiling_enabled(void)
> >>> {
> >>> return
> >>> static_branch_maybe(CONFIG_MEM_ALLOC_PROFILING_ENABLED_BY_DEFAULT,
> >>> &mem_alloc_profiling_key);
> >>> }
> >>>
> >>> +static inline bool alloc_tag_trace_enabled(const struct alloc_tag *tag)
> >>> +{
> >>> + return static_branch_unlikely(&alloc_tag_trace_key);
> >>> +}
> >>> +
> >>> +void alloc_tag_trace_mem_alloc(union codetag_ref *ref, struct alloc_tag
> >>> *tag,
> >>> + size_t bytes);
> >>> +
> >>> +void alloc_tag_trace_mem_free(union codetag_ref *ref, struct alloc_tag
> >>> *tag,
> >>> + size_t bytes);
> >>> +
> >>> +void __alloc_tag_trace_hit(struct alloc_tag *tag);
> >>> +
> >>> +static inline void alloc_tag_trace_hit(struct alloc_tag *tag)
> >>> +{
> >>> + if (alloc_tag_trace_enabled(tag))
> >>> + __alloc_tag_trace_hit(tag);
> >>> +}
> >>> +
> >>> bool mem_alloc_profiling_permanently_disabled(void);
> >>>
> >>> static inline struct alloc_tag_counters alloc_tag_read(struct alloc_tag
> >>> *tag)
> >>> @@ -200,8 +221,13 @@ static inline bool alloc_tag_ref_set(union
> >>> codetag_ref *ref, struct alloc_tag *t
> >>>
> >>> static inline void alloc_tag_add(union codetag_ref *ref, struct
> >>> alloc_tag *tag, size_t bytes)
> >>> {
> >>> - if (likely(alloc_tag_ref_set(ref, tag)))
> >>> + if (likely(alloc_tag_ref_set(ref, tag))) {
> >>> this_cpu_add(tag->counters->bytes, bytes);
> >>> +
> >>> + if (alloc_tag_trace_enabled(tag))
> >>> + /* Trace successful allocs with their unique ref */
> >>> + alloc_tag_trace_mem_alloc(ref, tag, bytes);
> >>> + }
> >>> }
> >>>
> >>> static inline void alloc_tag_sub(union codetag_ref *ref, size_t bytes)
> >>> @@ -222,6 +248,10 @@ static inline void alloc_tag_sub(union codetag_ref
> >>> *ref, size_t bytes)
> >>> this_cpu_sub(tag->counters->bytes, bytes);
> >>> this_cpu_dec(tag->counters->calls);
> >>>
> >>> + if (alloc_tag_trace_enabled(tag))
> >>> + /* Trace frees with their unique ref */
> >>> + alloc_tag_trace_mem_free(ref, tag, bytes);
> >>> +
> >>> ref->ct = NULL;
> >>> }
> >>>
> >>> @@ -247,21 +277,24 @@ static inline void alloc_tag_add(union codetag_ref
> >>> *ref, struct alloc_tag *tag,
> >>> static inline void alloc_tag_sub(union codetag_ref *ref, size_t bytes) {}
> >>> static inline void alloc_tag_set_inaccurate(struct alloc_tag *tag) {}
> >>> static inline bool alloc_tag_is_inaccurate(struct alloc_tag *tag) {
> >>> return false; }
> >>> +#define alloc_tag_trace_hit(_tag) /* NOOP */
> >>
> >> I'd prefer we use `do {} while (0)` here, following the same pattern
> >> as alloc_tag_record.
> >>
> >>> #define alloc_tag_record(p) do {} while (0)
> >>>
> >>> #endif /* CONFIG_MEM_ALLOC_PROFILING */
> >>>
> >>> -#define alloc_hooks_tag(_tag, _do_alloc) \
> >>> -({ \
> >>> - typeof(_do_alloc) _res; \
> >>> - if (mem_alloc_profiling_enabled()) { \
> >>> - struct alloc_tag * __maybe_unused _old; \
> >>> - _old = alloc_tag_save(_tag); \
> >>> - _res = _do_alloc; \
> >>> - alloc_tag_restore(_tag, _old); \
> >>> - } else \
> >>> - _res = _do_alloc; \
> >>> - _res; \
> >>> +#define alloc_hooks_tag(_tag, _do_alloc)
> >>> \
> >>> +({
> >>> \
> >>> + typeof(_do_alloc) _res;
> >>> \
> >>> + if (mem_alloc_profiling_enabled()) {
> >>> \
> >>> + struct alloc_tag * __maybe_unused _old;
> >>> \
> >>> + /* Fired here to cleanly capture the caller's stack trace
> >>> */ \
> >>> + alloc_tag_trace_hit(_tag);
> >>> \
> >>> + _old = alloc_tag_save(_tag);
> >>> \
> >>> + _res = _do_alloc;
> >>> \
> >>> + alloc_tag_restore(_tag, _old);
> >>> \
> >>> + } else
> >>> \
> >>> + _res = _do_alloc;
> >>> \
> >>> + _res;
> >>> \
> >>> })
> >>>
> >>> #define alloc_hooks(_do_alloc)
> >>> \
> >>> diff --git a/include/trace/events/alloc_tag.h
> >>> b/include/trace/events/alloc_tag.h
> >>> new file mode 100644
> >>> index 000000000000..af2182501864
> >>> --- /dev/null
> >>> +++ b/include/trace/events/alloc_tag.h
> >>> @@ -0,0 +1,122 @@
> >>> +/* SPDX-License-Identifier: GPL-2.0 */
> >>> +#undef TRACE_SYSTEM
> >>> +#define TRACE_SYSTEM alloc_tag
> >>> +
> >>> +#if !defined(_TRACE_ALLOC_TAG_H) || defined(TRACE_HEADER_MULTI_READ)
> >>> +#define _TRACE_ALLOC_TAG_H
> >>> +
> >>> +#include <linux/tracepoint.h>
> >>> +
> >>> +/*
> >>> + * alloc_tag_hit is generated at the exact allocation call site and can
> >>> be
> >>> + * used to capture a clean stack trace.
> >>> + *
> >>> + * To link this stack trace to the actual allocated memory chunk, tools
> >>> must
> >>> + * correlate this event with the resulting alloc_tag_mem_alloced event.
> >>> Since
> >>> + * multiple threads can hit the same tag simultaneously, tools must
> >>> match BOTH
> >>> + * the `tag` field and the implicitly recorded PID provided by the core
> >>> + * tracing subsystem.
> >>> + */
> >>> +TRACE_EVENT(alloc_tag_hit,
> >>> +
> >>> + TP_PROTO(struct alloc_tag *tag),
> >>> +
> >>> + TP_ARGS(tag),
> >>> +
> >>> + TP_STRUCT__entry(
> >>> + __field(struct alloc_tag *, tag)
> >>> + __string(modname, tag->ct.modname ? tag->ct.modname :
> >>> "NONE")
> >>> + __string(filename, tag->ct.filename)
> >>> + __string(function, tag->ct.function)
> >>> + __field(unsigned int, lineno)
> >>> + ),
> >>> +
> >>> + TP_fast_assign(
> >>> + __entry->tag = tag;
> >>> + __assign_str(modname);
> >>> + __assign_str(filename);
> >>> + __assign_str(function);
> >>> + __entry->lineno = tag->ct.lineno;
> >>> + ),
> >>> +
> >>> + TP_printk("tag %p, module: %s, filename: %s, function %s, lineno
> >>> %u",
> >>> + __entry->tag,
> >>> + __get_str(modname),
> >>> + __get_str(filename),
> >>> + __get_str(function),
> >>> + __entry->lineno
> >>> + )
> >>> +);
> >>>
> >>
> >> When I trace an interface provided by a kernel build-in such as shmem,
> >> this is what I see in the trace:
> >>
> >> trig-6838 [004] ..... 1309.321906: alloc_tag_hit: tag
> >> 000000007874a2c9, module: NONE, filename: mm/shmem.c, function
> >> shmem_alloc_folio, lineno 2138
> >>
> >> The module field shows NONE, which looks a bit odd.
> >>
> >>> +/*
> >>> + * alloc_tag_mem_alloced is generated after memory is successfully
> >>> allocated.
> >>> + * It captures the exact byte size.
> >>> + *
> >>> + * The `ref` pointer identifies the memory chunk for tracking its
> >>> lifecycle
> >>> + * (e.g., matching it with alloc_tag_mem_freed).
> >>> + *
> >>> + * Because the kernel isolates active allocations within the task struct
> >>> + * (current->alloc_tag), this even will always share the same implicit
> >>> PID as
> >>> + * its corresponding alloc_tag_hit event. Tools should use the
> >>> combination
> >>> + * PID + `tag` to correlate them.
> >>> + */
> >>> +TRACE_EVENT(alloc_tag_mem_alloced,
> >>> +
> >>> + TP_PROTO(union codetag_ref *ref, struct alloc_tag *tag, size_t
> >>> bytes),
> >>> +
> >>> + TP_ARGS(ref, tag, bytes),
> >>> +
> >>> + TP_STRUCT__entry(
> >>> + __field(union codetag_ref *, ref)
> >>> + __field(struct alloc_tag *, tag)
> >>> + __field(size_t, bytes)
> >>> + ),
> >>> +
> >>> + TP_fast_assign(
> >>> + __entry->ref = ref;
> >>> + __entry->tag = tag;
> >>> + __entry->bytes = bytes;
> >>> + ),
> >>> +
> >>> + TP_printk("reference %p, tag %p, bytes %zu",
> >>> + __entry->ref,
> >>> + __entry->tag,
> >>> + __entry->bytes
> >>> + )
> >>> +);
> >>> +
> >>> +/*
> >>> + * alloc_tag_mem_freed event is generated immediately before memory is
> >>> + * freed. The `ref` pointer matches the one emitted during allocation,
> >>> + * allowing tools to match it to it's corresponding allocation and
> >>> + * call stack.
> >>> + */
> >>
> >> But the thread for allocation and free may not be the same.
> >> For example, memory allocated in thread A could be freed by kswapd.
> >> In this case, would the ref fail to match and falsely report a memory leak?
> >
> > No, ref is unique within the system, so even if alloc and free happen
> > in different threads it still can be used to match them. The
> > thread+tag is used only when matching alloc_tag_hit with
> > alloc_tag_mem_alloced because at the time of alloc_tag_hit we don't
> > yet have the ref.
> >
>
> "The ref is unique within the system" is true of the obj_ext/page_ext slot,
> but for page allocations the event does not carry the slot's address,
> it carries the address of a stack temporary.
>
> /* Should be called only if mem_alloc_profiling_enabled() */
> static noinline
> void __pgalloc_tag_add(struct page *page, struct task_struct *task,
> unsigned int nr, unsigned int alloc_flags)
> {
> union pgtag_ref_handle handle;
> union codetag_ref ref; <- stack temporary
>
> if (likely(get_page_tag_ref(page, &ref, &handle))) {
> alloc_tag_add(&ref, task->alloc_tag, PAGE_SIZE * nr);
> update_page_tag_ref(handle, &ref);
> put_page_tag_ref(handle);
> } else {
> ......
> }
> }
>
> /* Should be called only if mem_alloc_profiling_enabled() */
> static noinline
> void __pgalloc_tag_sub(struct page *page, unsigned int nr)
> {
> union pgtag_ref_handle handle;
> union codetag_ref ref; <- stack temporary
>
> if (get_page_tag_ref(page, &ref, &handle)) {
> alloc_tag_sub(&ref, PAGE_SIZE * nr);
> update_page_tag_ref(handle, &ref);
> put_page_tag_ref(handle);
> }
> }
>
> Or am I missing / confusing something here?
>
This is a legitimate concern, I went through the code again, the
codetag_ref pointer approach works for percpu and slab allocators but
don't work for the page allocator.
Suren, Hao, I could potentially pass in the struct page * to
__alloc_tag_add and __alloc_tag_sub from the page allocator and keep
the codepath for the percpu and slab allocators the same. What do you
think? I think that should solve this concern.
> Thanks
> Best Regards
> Hao
>
> >>
> >> Thanks
> >> Best Regards
> >> Hao
> >>
> >>> +TRACE_EVENT(alloc_tag_mem_freed,
> >>> +
> >>> + TP_PROTO(union codetag_ref *ref, struct alloc_tag *tag, size_t
> >>> bytes),
> >>> +
> >>> + TP_ARGS(ref, tag, bytes),
> >>> +
> >>> + TP_STRUCT__entry(
> >>> + __field(union codetag_ref *, ref)
> >>> + __field(struct alloc_tag *, tag)
> >>> + __field(size_t, bytes)
> >>> + ),
> >>> +
> >>> + TP_fast_assign(
> >>> + __entry->ref = ref;
> >>> + __entry->tag = tag;
> >>> + __entry->bytes = bytes;
> >>> + ),
> >>> +
> >>> + TP_printk("reference %p, tag %p, bytes %zu",
> >>> + __entry->ref,
> >>> + __entry->tag,
> >>> + __entry->bytes
> >>> + )
> >>> +);
> >>> +
> >>> +#endif /* _TRACE_ALLOC_TAG_H */
> >>> +
> >>> +/* This part must be outside protection */
> >>> +#include <trace/define_trace.h>
> >>> diff --git a/mm/alloc_tag.c b/mm/alloc_tag.c
> >>> index f30ef8dd24c7..a5339767efd5 100644
> >>> --- a/mm/alloc_tag.c
> >>> +++ b/mm/alloc_tag.c
> >>> @@ -19,6 +19,9 @@
> >>> #include <linux/kmemleak.h>
> >>> #include <uapi/linux/alloc_tag.h>
> >>>
> >>> +#define CREATE_TRACE_POINTS
> >>> +#include <trace/events/alloc_tag.h>
> >>> +
> >>> #include "internal.h"
> >>> #include "page_alloc.h"
> >>>
> >>> @@ -55,6 +58,9 @@ EXPORT_SYMBOL(mem_alloc_profiling_key);
> >>>
> >>> DEFINE_STATIC_KEY_FALSE(mem_profiling_compressed);
> >>>
> >>> +DEFINE_STATIC_KEY_FALSE(alloc_tag_trace_key);
> >>> +EXPORT_SYMBOL(alloc_tag_trace_key);
> >>> +
> >>> struct alloc_tag_kernel_section kernel_tags = { NULL, 0 };
> >>> unsigned long alloc_tag_ref_mask;
> >>> int alloc_tag_ref_offs;
> >>> @@ -485,6 +491,28 @@ static const struct proc_ops allocinfo_proc_ops = {
> >>> #endif
> >>> };
> >>>
> >>> +void __alloc_tag_trace_hit(struct alloc_tag *tag)
> >>> +{
> >>> + if (unlikely(!tag))
> >>> + return;
> >>> + trace_alloc_tag_hit(tag);
> >>> +}
> >>> +EXPORT_SYMBOL(__alloc_tag_trace_hit);
> >>> +
> >>> +void alloc_tag_trace_mem_alloc(union codetag_ref *ref, struct alloc_tag
> >>> *tag,
> >>> + size_t bytes)
> >>> +{
> >>> + trace_alloc_tag_mem_alloced(ref, tag, bytes);
> >>> +}
> >>> +EXPORT_SYMBOL(alloc_tag_trace_mem_alloc);
> >>> +
> >>> +void alloc_tag_trace_mem_free(union codetag_ref *ref, struct alloc_tag
> >>> *tag,
> >>> + size_t bytes)
> >>> +{
> >>> + trace_alloc_tag_mem_freed(ref, tag, bytes);
> >>> +}
> >>> +EXPORT_SYMBOL(alloc_tag_trace_mem_free);
> >>> +
> >>> size_t alloc_tag_top_users(struct codetag_bytes *tags, size_t count,
> >>> bool can_sleep)
> >>> {
> >>> struct codetag_iterator iter;
