* [PATCH 0/3] sched/preempt: Check notrace preempt-count updates
@ 2026-11-16 7:54 Karl Mehltretter
2026-11-16 7:54 ` [PATCH 1/3] arm64: Always inline raw preempt count updates Karl Mehltretter
` (2 more replies)
0 siblings, 3 replies; 4+ messages in thread
From: Karl Mehltretter @ 2026-11-16 7:54 UTC (permalink / raw)
To: Peter Zijlstra, Ingo Molnar, Catalin Marinas, Will Deacon
Cc: Karl Mehltretter, Mark Rutland, Hongyan Xia, Song Liu,
Steven Rostedt, Juri Lelli, Vincent Guittot, Dietmar Eggemann,
Valentin Schneider, linux-arm-kernel, linux-trace-kernel,
linux-kernel
With CONFIG_DEBUG_PREEMPT, the notrace preemption macros bypass the
bounds checks used by ordinary operations. Add checked noinstr updates
for their existing callers and share validation with
preempt_count_add() and preempt_count_sub().
Patch 1 forces the arm64 raw updates called by the new helpers inline.
Patch 2 does the same for preempt_count(), which the helpers use, for
should_resched(), reached by the debug notrace enable path, and for the
existing non-debug __preempt_count_dec_and_test() path. The corresponding
x86, s390 and asm-generic operations are already __always_inline.
Patch 1 is extracted from Hongyan Xia's entry RFC. Song Liu also posted
notrace annotations for the same arm64 helpers:
https://lore.kernel.org/r/20261023011049.1164625-10-song@kernel.org
Patch 3 disables branch profiling for core.o when DEBUG_PREEMPT and
TRACE_BRANCH_PROFILING are both enabled. DEBUG_LOCKS_WARN_ON() evaluates
its condition before entering its instrumentation region, so profiling
that condition would call ftrace_likely_update() from noinstr text. The
scheduler Makefile already uses this per-object exclusion for
build_policy.o and build_utility.o.
The cost is loss of branch-profile data for the rest of core.o in that
configuration. Scheduler tracepoints and function tracing are unchanged.
A dedicated non-profiled object is the fallback if this object-wide cost
is considered too broad.
This series is independently useful. It also supplies the checked count
update planned for follow-up IRQ-entry work; it does not change IRQ entry,
trace-recursion lifetime or TRACE_CTX_TRANSITION.
This addresses Peter's suggestion here:
https://lore.kernel.org/r/20261029093219.GH1779021@noisy.programming.kicks-ass.net
The series is based on v7.4-rc2.
Testing:
- Clang 19.1.7 x86-64: a W=1 defconfig-derived build with
DEBUG_PREEMPT, annotated branch profiling, DEBUG_ENTRY and noinstr
validation passed. A nested-KVM boot passed the built-in and external
module checks and all five tracer workloads. A DEBUG_PREEMPT=n control
retained branch profiling for core.o and emitted no new helper symbols.
- GCC 14.2 arm64: a W=1 PREEMPT_RT build and allmodconfig passed. The
latter's nVHE object had no undefined references to the new helpers.
A native Axion boot passed the built-in and module checks and all five
tracers.
- GCC 14.2 and 16.2 ARC, C-SKY and MIPS: performance, size, all-branches
and forced-noinline core.o inspections passed without prohibited
noinstr relocations or branch-profile records. Full baseline and patched
kernels plus the external debug module built. Matching QEMU boots reached
the expected balanced and invalid-count markers on ARC and C-SKY UP and
MIPS UP/SMP. The patched kernels diagnosed every invalid operation;
rejected decrements retained their counts. An ARC DEBUG_PREEMPT=n
control retained core.o profiling without references to the new helpers.
The MIPS UP baseline and patched positive boots emitted the same three
existing smp_processor_id() diagnostics; the SMP boots emitted none.
I used an LLM to help extract and review the changes, write the test
harnesses and prepare the commit messages.
Hongyan Xia (1):
arm64: Always inline raw preempt count updates
Karl Mehltretter (2):
arm64: Always inline preempt count and reschedule checks
sched/preempt: Check preempt counts in notrace helpers
arch/arm64/include/asm/preempt.h | 10 ++++----
include/linux/preempt.h | 22 ++++++++++++++----
kernel/sched/Makefile | 3 +++
kernel/sched/core.c | 40 ++++++++++++++++++++++++++++----
4 files changed, 62 insertions(+), 13 deletions(-)
base-commit: 7475ea8b58616b33c04afb976cf0fbaad9dd63fb
^ permalink raw reply [flat|nested] 4+ messages in thread* [PATCH 1/3] arm64: Always inline raw preempt count updates
2026-11-16 7:54 [PATCH 0/3] sched/preempt: Check notrace preempt-count updates Karl Mehltretter
@ 2026-11-16 7:54 ` Karl Mehltretter
2026-11-16 7:54 ` [PATCH 2/3] arm64: Always inline preempt count and reschedule checks Karl Mehltretter
2026-11-16 7:54 ` [PATCH 3/3] sched/preempt: Check preempt counts in notrace helpers Karl Mehltretter
2 siblings, 0 replies; 4+ messages in thread
From: Karl Mehltretter @ 2026-11-16 7:54 UTC (permalink / raw)
To: Peter Zijlstra, Ingo Molnar, Catalin Marinas, Will Deacon
Cc: Hongyan Xia, Mark Rutland, Song Liu, Steven Rostedt, Juri Lelli,
Vincent Guittot, Dietmar Eggemann, Valentin Schneider,
linux-arm-kernel, linux-trace-kernel, linux-kernel,
Karl Mehltretter
From: Hongyan Xia <hongyan.xia@transsion.com>
Raw preempt count updates can be emitted out of line by the compiler.
For a noinstr caller, that puts the update in ordinary instrumentable
text. Force __preempt_count_add() and __preempt_count_sub() inline so
their code stays in the caller's section.
Link: https://lore.kernel.org/all/c3a90368dd6c096af9d918e036a27728696fb8a4.1786603168.git.hongyan.xia@transsion.com/
Signed-off-by: Hongyan Xia <hongyan.xia@transsion.com>
[kmehltretter: split out the arm64 preempt.h hunk and rewrite the changelog]
Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
---
arch/arm64/include/asm/preempt.h | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/arch/arm64/include/asm/preempt.h b/arch/arm64/include/asm/preempt.h
index 75a592b7f47f7..f3a5c2cd1936d 100644
--- a/arch/arm64/include/asm/preempt.h
+++ b/arch/arm64/include/asm/preempt.h
@@ -41,14 +41,14 @@ static inline bool test_preempt_need_resched(void)
return !current_thread_info()->preempt.need_resched;
}
-static inline void __preempt_count_add(int val)
+static __always_inline void __preempt_count_add(int val)
{
u32 pc = READ_ONCE(current_thread_info()->preempt.count);
pc += val;
WRITE_ONCE(current_thread_info()->preempt.count, pc);
}
-static inline void __preempt_count_sub(int val)
+static __always_inline void __preempt_count_sub(int val)
{
u32 pc = READ_ONCE(current_thread_info()->preempt.count);
pc -= val;
^ permalink raw reply [flat|nested] 4+ messages in thread* [PATCH 2/3] arm64: Always inline preempt count and reschedule checks
2026-11-16 7:54 [PATCH 0/3] sched/preempt: Check notrace preempt-count updates Karl Mehltretter
2026-11-16 7:54 ` [PATCH 1/3] arm64: Always inline raw preempt count updates Karl Mehltretter
@ 2026-11-16 7:54 ` Karl Mehltretter
2026-11-16 7:54 ` [PATCH 3/3] sched/preempt: Check preempt counts in notrace helpers Karl Mehltretter
2 siblings, 0 replies; 4+ messages in thread
From: Karl Mehltretter @ 2026-11-16 7:54 UTC (permalink / raw)
To: Peter Zijlstra, Ingo Molnar, Catalin Marinas, Will Deacon
Cc: Karl Mehltretter, Mark Rutland, Hongyan Xia, Song Liu,
Steven Rostedt, Juri Lelli, Vincent Guittot, Dietmar Eggemann,
Valentin Schneider, linux-arm-kernel, linux-trace-kernel,
linux-kernel
The preempt count and reschedule checks are plain inline functions,
which permits out-of-line copies in instrumentable text. The checked
notrace helpers added by the following patch use preempt_count(), while
preempt_enable_notrace() uses should_resched() with
CONFIG_DEBUG_PREEMPT and __preempt_count_dec_and_test() without it.
Force all three operations inline so noinstr callers keep them in their
own section.
Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
---
arch/arm64/include/asm/preempt.h | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/arch/arm64/include/asm/preempt.h b/arch/arm64/include/asm/preempt.h
index f3a5c2cd1936d..805f7b33baf09 100644
--- a/arch/arm64/include/asm/preempt.h
+++ b/arch/arm64/include/asm/preempt.h
@@ -7,7 +7,7 @@
#define PREEMPT_NEED_RESCHED BIT(32)
#define PREEMPT_ENABLED (PREEMPT_NEED_RESCHED)
-static inline int preempt_count(void)
+static __always_inline int preempt_count(void)
{
return READ_ONCE(current_thread_info()->preempt.count);
}
@@ -75,13 +75,13 @@ static inline int __preempt_count_sub_return(int val)
return pc;
}
-static inline bool should_resched(int preempt_offset)
+static __always_inline bool should_resched(int preempt_offset)
{
u64 pc = READ_ONCE(current_thread_info()->preempt_count);
return unlikely(pc == preempt_offset);
}
-static inline bool __preempt_count_dec_and_test(void)
+static __always_inline bool __preempt_count_dec_and_test(void)
{
/*
* We must load the combined 'preempt_count' after decrementing
^ permalink raw reply [flat|nested] 4+ messages in thread* [PATCH 3/3] sched/preempt: Check preempt counts in notrace helpers
2026-11-16 7:54 [PATCH 0/3] sched/preempt: Check notrace preempt-count updates Karl Mehltretter
2026-11-16 7:54 ` [PATCH 1/3] arm64: Always inline raw preempt count updates Karl Mehltretter
2026-11-16 7:54 ` [PATCH 2/3] arm64: Always inline preempt count and reschedule checks Karl Mehltretter
@ 2026-11-16 7:54 ` Karl Mehltretter
2 siblings, 0 replies; 4+ messages in thread
From: Karl Mehltretter @ 2026-11-16 7:54 UTC (permalink / raw)
To: Peter Zijlstra, Ingo Molnar, Catalin Marinas, Will Deacon
Cc: Karl Mehltretter, Mark Rutland, Hongyan Xia, Song Liu,
Steven Rostedt, Juri Lelli, Vincent Guittot, Dietmar Eggemann,
Valentin Schneider, linux-arm-kernel, linux-trace-kernel,
linux-kernel
The notrace preemption macros use raw count updates even with
CONFIG_DEBUG_PREEMPT enabled. They therefore skip the bounds checks
which ordinary preemption operations provide.
Add checked noinstr preempt-count helpers and use them in
preempt_disable_notrace(), preempt_enable_notrace() and
preempt_enable_no_resched_notrace(). Export the out-of-line helpers
for module callers. Keep raw updates when CONFIG_DEBUG_PREEMPT is
disabled.
Share the validation with preempt_count_add() and preempt_count_sub(),
while leaving their preemption-latency tracing and ordering unchanged.
DEBUG_LOCKS_WARN_ON() brackets its diagnostic work with
instrumentation_begin() and instrumentation_end(), but evaluates its
condition before that region. Branch profiling would therefore call
ftrace_likely_update() from noinstr text.
Compile core.o with branch profiling disabled when DEBUG_PREEMPT and
TRACE_BRANCH_PROFILING are both enabled, as the scheduler Makefile already
does for build_policy.o and build_utility.o. This keeps valid notrace count
updates free of profiler calls while retaining the existing warning path.
The tradeoff is loss of branch-profile data for the rest of core.o in that
configuration; scheduler tracepoints and function tracing are unchanged.
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20261029093219.GH1779021@noisy.programming.kicks-ass.net
Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
---
include/linux/preempt.h | 22 ++++++++++++++++++----
kernel/sched/Makefile | 3 +++
kernel/sched/core.c | 40 ++++++++++++++++++++++++++++++++++++----
3 files changed, 57 insertions(+), 8 deletions(-)
diff --git a/include/linux/preempt.h b/include/linux/preempt.h
index 06ff44a4b8b6a..42c585fd91d45 100644
--- a/include/linux/preempt.h
+++ b/include/linux/preempt.h
@@ -217,6 +217,17 @@ static __always_inline unsigned char interrupt_context_level(void)
*/
#define in_atomic_preempt_off() (preempt_count() != PREEMPT_DISABLE_OFFSET)
+#ifdef CONFIG_DEBUG_PREEMPT
+void preempt_count_add_notrace(int val);
+void preempt_count_sub_notrace(int val);
+#define preempt_count_dec_and_test_notrace() \
+ ({ preempt_count_sub_notrace(1); should_resched(0); })
+#else
+#define preempt_count_add_notrace(val) __preempt_count_add(val)
+#define preempt_count_sub_notrace(val) __preempt_count_sub(val)
+#define preempt_count_dec_and_test_notrace() __preempt_count_dec_and_test()
+#endif
+
#if defined(CONFIG_DEBUG_PREEMPT) || defined(CONFIG_TRACE_PREEMPT_TOGGLE)
extern void preempt_count_add(int val);
extern void preempt_count_sub(int val);
@@ -231,6 +242,9 @@ extern void preempt_count_sub(int val);
#define __preempt_count_inc() __preempt_count_add(1)
#define __preempt_count_dec() __preempt_count_sub(1)
+#define preempt_count_inc_notrace() preempt_count_add_notrace(1)
+#define preempt_count_dec_notrace() preempt_count_sub_notrace(1)
+
#define preempt_count_inc() preempt_count_add(1)
#define preempt_count_dec() preempt_count_sub(1)
@@ -263,7 +277,7 @@ do { \
#define preempt_enable_notrace() \
do { \
barrier(); \
- if (unlikely(__preempt_count_dec_and_test())) \
+ if (unlikely(preempt_count_dec_and_test_notrace())) \
__preempt_schedule_notrace(); \
} while (0)
@@ -283,7 +297,7 @@ do { \
#define preempt_enable_notrace() \
do { \
barrier(); \
- __preempt_count_dec(); \
+ preempt_count_dec_notrace(); \
} while (0)
#define preempt_check_resched() do { } while (0)
@@ -291,14 +305,14 @@ do { \
#define preempt_disable_notrace() \
do { \
- __preempt_count_inc(); \
+ preempt_count_inc_notrace(); \
barrier(); \
} while (0)
#define preempt_enable_no_resched_notrace() \
do { \
barrier(); \
- __preempt_count_dec(); \
+ preempt_count_dec_notrace(); \
} while (0)
#else /* !CONFIG_PREEMPT_COUNT */
diff --git a/kernel/sched/Makefile b/kernel/sched/Makefile
index b1f1a367034ff..d87f1bd3c28c3 100644
--- a/kernel/sched/Makefile
+++ b/kernel/sched/Makefile
@@ -27,6 +27,9 @@ endif
# Branch profiling isn't noinstr-safe
ifdef CONFIG_TRACE_BRANCH_PROFILING
+ifdef CONFIG_DEBUG_PREEMPT
+CFLAGS_core.o += -DDISABLE_BRANCH_PROFILING
+endif
CFLAGS_build_policy.o += -DDISABLE_BRANCH_PROFILING
CFLAGS_build_utility.o += -DDISABLE_BRANCH_PROFILING
endif
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 1a088d4576f59..98ca93bc60887 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -6040,7 +6040,7 @@ static inline void preempt_latency_start(int val)
}
}
-void preempt_count_add(int val)
+static __always_inline bool __preempt_count_add_checked(int val)
{
#ifdef CONFIG_DEBUG_PREEMPT
/*
@@ -6052,7 +6052,7 @@ void preempt_count_add(int val)
*/
if (!IS_ENABLED(CONFIG_HAS_SEPARATE_PREEMPT_RESCHED_BITS) &&
DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
- return;
+ return false;
#endif
__preempt_count_add(val);
#ifdef CONFIG_DEBUG_PREEMPT
@@ -6062,6 +6062,21 @@ void preempt_count_add(int val)
DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
PREEMPT_MASK - 10);
#endif
+ return true;
+}
+
+#ifdef CONFIG_DEBUG_PREEMPT
+void noinstr preempt_count_add_notrace(int val)
+{
+ __preempt_count_add_checked(val);
+}
+EXPORT_SYMBOL(preempt_count_add_notrace);
+#endif
+
+void preempt_count_add(int val)
+{
+ if (!__preempt_count_add_checked(val))
+ return;
preempt_latency_start(val);
}
EXPORT_SYMBOL(preempt_count_add);
@@ -6077,7 +6092,7 @@ static inline void preempt_latency_stop(int val)
trace_preempt_on(CALLER_ADDR0, get_lock_parent_ip());
}
-void preempt_count_sub(int val)
+static __always_inline bool preempt_count_sub_debug_checks(int val)
{
#ifdef CONFIG_DEBUG_PREEMPT
/*
@@ -6087,15 +6102,32 @@ void preempt_count_sub(int val)
unsigned int pc = preempt_count();
if (DEBUG_LOCKS_WARN_ON(pc - uval > pc))
- return;
+ return false;
/*
* Is the spinlock portion underflowing?
*/
if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
!(preempt_count() & PREEMPT_MASK)))
+ return false;
+#endif
+ return true;
+}
+
+#ifdef CONFIG_DEBUG_PREEMPT
+void noinstr preempt_count_sub_notrace(int val)
+{
+ if (!preempt_count_sub_debug_checks(val))
return;
+ __preempt_count_sub(val);
+}
+EXPORT_SYMBOL(preempt_count_sub_notrace);
#endif
+void preempt_count_sub(int val)
+{
+ if (!preempt_count_sub_debug_checks(val))
+ return;
+
preempt_latency_stop(val);
__preempt_count_sub(val);
}
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2026-11-16 7:54 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-16 7:54 [PATCH 0/3] sched/preempt: Check notrace preempt-count updates Karl Mehltretter
2026-11-16 7:54 ` [PATCH 1/3] arm64: Always inline raw preempt count updates Karl Mehltretter
2026-11-16 7:54 ` [PATCH 2/3] arm64: Always inline preempt count and reschedule checks Karl Mehltretter
2026-11-16 7:54 ` [PATCH 3/3] sched/preempt: Check preempt counts in notrace helpers Karl Mehltretter
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®