Most of sched_clock()'s implementation is ineligible for instrumentation
due to relying on sched_clock_noinstr().

Split the implementation off into an __always_inline function
__sched_clock(), which is then used by the noinstr and instrumentable
version, to allow more of sched_clock() to be covered by various
instrumentation.

This will allow instrumentation with the various sanitizers (KASAN,
KCSAN, KMSAN, UBSAN). For KCSAN, we know that raw seqcount_latch usage
without annotations will result in false positive reports: tell it that
all of __sched_clock() is "atomic" for the latch writer; later changes
in this series will take care of the readers.

Link: https://lore.kernel.org/all/20241030204815.GQ14555@noisy.programming.kicks-ass.net/
Co-developed-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Marco Elver <elver@google.com>
---
v2:
* New patch.
---
 kernel/time/sched_clock.c | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/kernel/time/sched_clock.c b/kernel/time/sched_clock.c
index 85595fcf6aa2..29bdf309dae8 100644
--- a/kernel/time/sched_clock.c
+++ b/kernel/time/sched_clock.c
@@ -80,7 +80,7 @@ notrace int sched_clock_read_retry(unsigned int seq)
 	return raw_read_seqcount_latch_retry(&cd.seq, seq);
 }
 
-unsigned long long noinstr sched_clock_noinstr(void)
+static __always_inline unsigned long long __sched_clock(void)
 {
 	struct clock_read_data *rd;
 	unsigned int seq;
@@ -98,11 +98,23 @@ unsigned long long noinstr sched_clock_noinstr(void)
 	return res;
 }
 
+unsigned long long noinstr sched_clock_noinstr(void)
+{
+	return __sched_clock();
+}
+
 unsigned long long notrace sched_clock(void)
 {
 	unsigned long long ns;
 	preempt_disable_notrace();
-	ns = sched_clock_noinstr();
+	/*
+	 * All of __sched_clock() is a seqcount_latch reader critical section,
+	 * but relies on the raw helpers which are uninstrumented. For KCSAN,
+	 * mark all accesses in __sched_clock() as atomic.
+	 */
+	kcsan_nestable_atomic_begin();
+	ns = __sched_clock();
+	kcsan_nestable_atomic_end();
 	preempt_enable_notrace();
 	return ns;
 }
-- 
2.47.0.163.g1226f6d8fa-goog

Oops, typo'd the commit message:

On Mon, 4 Nov 2024 at 17:19, Marco Elver <elver@google.com> wrote:
>
> Most of sched_clock()'s implementation is ineligible for instrumentation
> due to relying on sched_clock_noinstr().
>
> Split the implementation off into an __always_inline function
> __sched_clock(), which is then used by the noinstr and instrumentable
> version, to allow more of sched_clock() to be covered by various
> instrumentation.
>
> This will allow instrumentation with the various sanitizers (KASAN,
> KCSAN, KMSAN, UBSAN). For KCSAN, we know that raw seqcount_latch usage
> without annotations will result in false positive reports: tell it that
> all of __sched_clock() is "atomic" for the latch writer; later changes

s/writer/reader/

> in this series will take care of the readers.

s/readers/writers/

... might be less confusing. If you apply, kindly fix up the commit
message, so that future people will be less confused. The code comment
is correct.

Thanks,
-- Marco

On Tue, Nov 05, 2024 at 10:22:51AM +0100, Marco Elver wrote:
> Oops, typo'd the commit message:
> 
> On Mon, 4 Nov 2024 at 17:19, Marco Elver <elver@google.com> wrote:
> >
> > Most of sched_clock()'s implementation is ineligible for instrumentation
> > due to relying on sched_clock_noinstr().
> >
> > Split the implementation off into an __always_inline function
> > __sched_clock(), which is then used by the noinstr and instrumentable
> > version, to allow more of sched_clock() to be covered by various
> > instrumentation.
> >
> > This will allow instrumentation with the various sanitizers (KASAN,
> > KCSAN, KMSAN, UBSAN). For KCSAN, we know that raw seqcount_latch usage
> > without annotations will result in false positive reports: tell it that
> > all of __sched_clock() is "atomic" for the latch writer; later changes
> 
> s/writer/reader/
> 
> > in this series will take care of the readers.
> 
> s/readers/writers/
> 
> ... might be less confusing. If you apply, kindly fix up the commit
> message, so that future people will be less confused. The code comment
> is correct.

So done. Thanks!

The following commit has been merged into the locking/core branch of tip:

Commit-ID:     8ab40fc2b9086b915e46890bb9252dc7692f1da0
Gitweb:        https://git.kernel.org/tip/8ab40fc2b9086b915e46890bb9252dc7692f1da0
Author:        Marco Elver <elver@google.com>
AuthorDate:    Mon, 04 Nov 2024 16:43:06 +01:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Tue, 05 Nov 2024 12:55:35 +01:00

time/sched_clock: Broaden sched_clock()'s instrumentation coverage

Most of sched_clock()'s implementation is ineligible for instrumentation
due to relying on sched_clock_noinstr().

Split the implementation off into an __always_inline function
__sched_clock(), which is then used by the noinstr and instrumentable
version, to allow more of sched_clock() to be covered by various
instrumentation.

This will allow instrumentation with the various sanitizers (KASAN,
KCSAN, KMSAN, UBSAN). For KCSAN, we know that raw seqcount_latch usage
without annotations will result in false positive reports: tell it that
all of __sched_clock() is "atomic" for the latch reader; later changes
in this series will take care of the writers.

Co-developed-by: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Signed-off-by: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Signed-off-by: Marco Elver <elver@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20241104161910.780003-3-elver@google.com
---
 kernel/time/sched_clock.c | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/kernel/time/sched_clock.c b/kernel/time/sched_clock.c
index 85595fc..29bdf30 100644
--- a/kernel/time/sched_clock.c
+++ b/kernel/time/sched_clock.c
@@ -80,7 +80,7 @@ notrace int sched_clock_read_retry(unsigned int seq)
 	return raw_read_seqcount_latch_retry(&cd.seq, seq);
 }
 
-unsigned long long noinstr sched_clock_noinstr(void)
+static __always_inline unsigned long long __sched_clock(void)
 {
 	struct clock_read_data *rd;
 	unsigned int seq;
@@ -98,11 +98,23 @@ unsigned long long noinstr sched_clock_noinstr(void)
 	return res;
 }
 
+unsigned long long noinstr sched_clock_noinstr(void)
+{
+	return __sched_clock();
+}
+
 unsigned long long notrace sched_clock(void)
 {
 	unsigned long long ns;
 	preempt_disable_notrace();
-	ns = sched_clock_noinstr();
+	/*
+	 * All of __sched_clock() is a seqcount_latch reader critical section,
+	 * but relies on the raw helpers which are uninstrumented. For KCSAN,
+	 * mark all accesses in __sched_clock() as atomic.
+	 */
+	kcsan_nestable_atomic_begin();
+	ns = __sched_clock();
+	kcsan_nestable_atomic_end();
 	preempt_enable_notrace();
 	return ns;
 }