[PATCH bpf v2 0/2] Fix bpf_link grace period wait for tracepoints

[PATCH bpf v2 0/2] Fix bpf_link grace period wait for tracepoints

[Kumar Kartikeya Dwivedi]

A recent change to non-faultable tracepoints switched from
preempt-disabled critical sections to SRCU-fast, which breaks
assumptions in the bpf_link_free() path. Use call_srcu() to fix the
breakage. While at it, retire rcu_trace_implies_rcu_gp() and clean up
all in-tree users.

Changelog:
----------
v1 -> v2
v1: https://lore.kernel.org/bpf/20260330032124.3141001-1-memxor@gmail.com

 * Add Reviewed-by tags. (Paul, Puranjay)
 * Adjust commit descriptions and comments to clarify intent. (Puranjay)

Kumar Kartikeya Dwivedi (2):
  bpf: Fix grace period wait for tracepoint bpf_link
  bpf: Retire rcu_trace_implies_rcu_gp()

 include/linux/rcupdate.h   | 12 -----------
 include/linux/tracepoint.h |  8 +++++++
 kernel/bpf/core.c          | 10 ++++-----
 kernel/bpf/helpers.c       |  2 +-
 kernel/bpf/memalloc.c      | 33 +++++++++-------------------
 kernel/bpf/syscall.c       | 44 +++++++++++++++++++++++---------------
 6 files changed, 50 insertions(+), 59 deletions(-)


base-commit: c369299895a591d96745d6492d4888259b004a9e
-- 
2.52.0

[PATCH bpf v2 1/2] bpf: Fix grace period wait for tracepoint bpf_link

[Kumar Kartikeya Dwivedi]

Recently, tracepoints were switched from using disabled preemption
(which acts as RCU read section) to SRCU-fast when they are not
faultable. This means that to do a proper grace period wait for programs
running in such tracepoints, we must use SRCU's grace period wait.
This is only for non-faultable tracepoints, faultable ones continue
using RCU Tasks Trace.

However, bpf_link_free() currently does call_rcu() for all cases when
the link is non-sleepable (hence, for tracepoints, non-faultable). Fix
this by doing a call_srcu() grace period wait.

As far RCU Tasks Trace gp -> RCU gp chaining is concerned, it is deemed
unnecessary for tracepoint programs. The link and program are either
accessed under RCU Tasks Trace protection, or SRCU-fast protection now.

The earlier logic of chaining both RCU Tasks Trace and RCU gp waits was
to generalize the logic, even if it conceded an extra RCU gp wait,
however that is unnecessary for tracepoints even before this change.
In practice no cost was paid since rcu_trace_implies_rcu_gp() was always
true. Hence we need not chaing any RCU gp after the SRCU gp.

For instance, in the non-faultable raw tracepoint, the RCU read section
of the program in __bpf_trace_run() is enclosed in the SRCU gp, likewise
for faultable raw tracepoint, the program is under the RCU Tasks Trace
protection. Hence, the outermost scope can be waited upon to ensure
correctness.

Fixes: a46023d5616e ("tracing: Guard __DECLARE_TRACE() use of __DO_TRACE_CALL() with SRCU-fast")
Reviewed-by: Puranjay Mohan <puranjay@kernel.org>
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
---
 include/linux/tracepoint.h |  8 ++++++++
 kernel/bpf/syscall.c       | 26 ++++++++++++++++++++++++--
 2 files changed, 32 insertions(+), 2 deletions(-)

diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h
index 22ca1c8b54f3..8227102a771f 100644
--- a/include/linux/tracepoint.h
+++ b/include/linux/tracepoint.h
@@ -113,6 +113,10 @@ void for_each_tracepoint_in_module(struct module *mod,
  */
 #ifdef CONFIG_TRACEPOINTS
 extern struct srcu_struct tracepoint_srcu;
+static inline struct srcu_struct *tracepoint_srcu_ptr(void)
+{
+	return &tracepoint_srcu;
+}
 static inline void tracepoint_synchronize_unregister(void)
 {
 	synchronize_rcu_tasks_trace();
@@ -123,6 +127,10 @@ static inline bool tracepoint_is_faultable(struct tracepoint *tp)
 	return tp->ext && tp->ext->faultable;
 }
 #else
+static inline struct srcu_struct *tracepoint_srcu_ptr(void)
+{
+	return NULL;
+}
 static inline void tracepoint_synchronize_unregister(void)
 { }
 static inline bool tracepoint_is_faultable(struct tracepoint *tp)
diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 274039e36465..89fa8f00adfa 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -3261,6 +3261,17 @@ static void bpf_link_defer_dealloc_rcu_gp(struct rcu_head *rcu)
 	bpf_link_dealloc(link);
 }
 
+static bool bpf_link_is_tracepoint(struct bpf_link *link)
+{
+	/*
+	 * Only these combinations support a tracepoint bpf_link.
+	 * BPF_LINK_TYPE_TRACING raw_tp progs are hardcoded to use
+	 * bpf_raw_tp_link_lops, see bpf_raw_tp_link_attach().
+	 */
+	return link->type == BPF_LINK_TYPE_RAW_TRACEPOINT ||
+	       (link->type == BPF_LINK_TYPE_TRACING && link->attach_type == BPF_TRACE_RAW_TP);
+}
+
 static void bpf_link_defer_dealloc_mult_rcu_gp(struct rcu_head *rcu)
 {
 	if (rcu_trace_implies_rcu_gp())
@@ -3279,16 +3290,27 @@ static void bpf_link_free(struct bpf_link *link)
 	if (link->prog)
 		ops->release(link);
 	if (ops->dealloc_deferred) {
-		/* Schedule BPF link deallocation, which will only then
+		struct srcu_struct *tp_srcu = tracepoint_srcu_ptr();
+
+		/*
+		 * Schedule BPF link deallocation, which will only then
 		 * trigger putting BPF program refcount.
 		 * If underlying BPF program is sleepable or BPF link's target
 		 * attach hookpoint is sleepable or otherwise requires RCU GPs
 		 * to ensure link and its underlying BPF program is not
 		 * reachable anymore, we need to first wait for RCU tasks
-		 * trace sync, and then go through "classic" RCU grace period
+		 * trace sync, and then go through "classic" RCU grace period.
+		 *
+		 * For tracepoint BPF links, we need to go through SRCU grace
+		 * period wait instead when non-faultable tracepoint is used. We
+		 * don't need to chain SRCU grace period waits, however, for the
+		 * faultable case, since it exclusively uses RCU Tasks Trace.
 		 */
 		if (link->sleepable || (link->prog && link->prog->sleepable))
 			call_rcu_tasks_trace(&link->rcu, bpf_link_defer_dealloc_mult_rcu_gp);
+		/* We need to do a SRCU grace period wait for tracepoint-based BPF links. */
+		else if (bpf_link_is_tracepoint(link) && tp_srcu)
+			call_srcu(tp_srcu, &link->rcu, bpf_link_defer_dealloc_rcu_gp);
 		else
 			call_rcu(&link->rcu, bpf_link_defer_dealloc_rcu_gp);
 	} else if (ops->dealloc) {
-- 
2.52.0

Re: [PATCH bpf v2 1/2] bpf: Fix grace period wait for tracepoint bpf_link

[Andrii Nakryiko]

On Mon, Mar 30, 2026 at 7:31 AM Kumar Kartikeya Dwivedi
<memxor@gmail.com> wrote:
>
> Recently, tracepoints were switched from using disabled preemption
> (which acts as RCU read section) to SRCU-fast when they are not
> faultable. This means that to do a proper grace period wait for programs
> running in such tracepoints, we must use SRCU's grace period wait.
> This is only for non-faultable tracepoints, faultable ones continue
> using RCU Tasks Trace.
>
> However, bpf_link_free() currently does call_rcu() for all cases when
> the link is non-sleepable (hence, for tracepoints, non-faultable). Fix
> this by doing a call_srcu() grace period wait.
>
> As far RCU Tasks Trace gp -> RCU gp chaining is concerned, it is deemed
> unnecessary for tracepoint programs. The link and program are either
> accessed under RCU Tasks Trace protection, or SRCU-fast protection now.
>
> The earlier logic of chaining both RCU Tasks Trace and RCU gp waits was
> to generalize the logic, even if it conceded an extra RCU gp wait,
> however that is unnecessary for tracepoints even before this change.
> In practice no cost was paid since rcu_trace_implies_rcu_gp() was always
> true. Hence we need not chaing any RCU gp after the SRCU gp.
>
> For instance, in the non-faultable raw tracepoint, the RCU read section
> of the program in __bpf_trace_run() is enclosed in the SRCU gp, likewise
> for faultable raw tracepoint, the program is under the RCU Tasks Trace
> protection. Hence, the outermost scope can be waited upon to ensure
> correctness.
>
> Fixes: a46023d5616e ("tracing: Guard __DECLARE_TRACE() use of __DO_TRACE_CALL() with SRCU-fast")
> Reviewed-by: Puranjay Mohan <puranjay@kernel.org>
> Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
> ---
>  include/linux/tracepoint.h |  8 ++++++++
>  kernel/bpf/syscall.c       | 26 ++++++++++++++++++++++++--
>  2 files changed, 32 insertions(+), 2 deletions(-)
>
> diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h
> index 22ca1c8b54f3..8227102a771f 100644
> --- a/include/linux/tracepoint.h
> +++ b/include/linux/tracepoint.h
> @@ -113,6 +113,10 @@ void for_each_tracepoint_in_module(struct module *mod,
>   */
>  #ifdef CONFIG_TRACEPOINTS
>  extern struct srcu_struct tracepoint_srcu;
> +static inline struct srcu_struct *tracepoint_srcu_ptr(void)
> +{
> +       return &tracepoint_srcu;
> +}
>  static inline void tracepoint_synchronize_unregister(void)
>  {
>         synchronize_rcu_tasks_trace();
> @@ -123,6 +127,10 @@ static inline bool tracepoint_is_faultable(struct tracepoint *tp)
>         return tp->ext && tp->ext->faultable;
>  }
>  #else
> +static inline struct srcu_struct *tracepoint_srcu_ptr(void)
> +{
> +       return NULL;
> +}
>  static inline void tracepoint_synchronize_unregister(void)
>  { }
>  static inline bool tracepoint_is_faultable(struct tracepoint *tp)
> diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> index 274039e36465..89fa8f00adfa 100644
> --- a/kernel/bpf/syscall.c
> +++ b/kernel/bpf/syscall.c
> @@ -3261,6 +3261,17 @@ static void bpf_link_defer_dealloc_rcu_gp(struct rcu_head *rcu)
>         bpf_link_dealloc(link);
>  }
>
> +static bool bpf_link_is_tracepoint(struct bpf_link *link)
> +{
> +       /*
> +        * Only these combinations support a tracepoint bpf_link.
> +        * BPF_LINK_TYPE_TRACING raw_tp progs are hardcoded to use
> +        * bpf_raw_tp_link_lops, see bpf_raw_tp_link_attach().
> +        */
> +       return link->type == BPF_LINK_TYPE_RAW_TRACEPOINT ||
> +              (link->type == BPF_LINK_TYPE_TRACING && link->attach_type == BPF_TRACE_RAW_TP);
> +}
> +
>  static void bpf_link_defer_dealloc_mult_rcu_gp(struct rcu_head *rcu)
>  {
>         if (rcu_trace_implies_rcu_gp())
> @@ -3279,16 +3290,27 @@ static void bpf_link_free(struct bpf_link *link)
>         if (link->prog)
>                 ops->release(link);
>         if (ops->dealloc_deferred) {
> -               /* Schedule BPF link deallocation, which will only then
> +               struct srcu_struct *tp_srcu = tracepoint_srcu_ptr();
> +
> +               /*
> +                * Schedule BPF link deallocation, which will only then
>                  * trigger putting BPF program refcount.
>                  * If underlying BPF program is sleepable or BPF link's target
>                  * attach hookpoint is sleepable or otherwise requires RCU GPs
>                  * to ensure link and its underlying BPF program is not
>                  * reachable anymore, we need to first wait for RCU tasks
> -                * trace sync, and then go through "classic" RCU grace period
> +                * trace sync, and then go through "classic" RCU grace period.
> +                *
> +                * For tracepoint BPF links, we need to go through SRCU grace
> +                * period wait instead when non-faultable tracepoint is used. We
> +                * don't need to chain SRCU grace period waits, however, for the
> +                * faultable case, since it exclusively uses RCU Tasks Trace.
>                  */
>                 if (link->sleepable || (link->prog && link->prog->sleepable))
>                         call_rcu_tasks_trace(&link->rcu, bpf_link_defer_dealloc_mult_rcu_gp);
> +               /* We need to do a SRCU grace period wait for tracepoint-based BPF links. */
> +               else if (bpf_link_is_tracepoint(link) && tp_srcu)
> +                       call_srcu(tp_srcu, &link->rcu, bpf_link_defer_dealloc_rcu_gp);

It bothers me that we have a special rule here just for tracepoints
(and non-sleepable ones only), and fetch that srcu_struct...  it feels
like we are missing some more fundamental link-type-specific grace
period callback that could be overridden by link ops. E.g., perhaps we
should be having something like


if (ops->call_rcu)
    ops->call_rcu(link, bpf_link_defer_dealloc_rcu_gp);
else
    /* current sleepable or non-sleepable RCU stuff */

and for non-sleepable rawtps we'll have link_ops->call_rcu set to
something that would fetch that tracepoint_srcu_ptr() and call
call_srcu().

I don't think we need to solve this right now, but if we ever have
second case where link's attach hook has some extra lifetime nuances
like this, we should generalize this. For now:

Acked-by: Andrii Nakryiko <andrii@kernel.org>


>                 else
>                         call_rcu(&link->rcu, bpf_link_defer_dealloc_rcu_gp);
>         } else if (ops->dealloc) {
> --
> 2.52.0
>

Re: [PATCH bpf v2 1/2] bpf: Fix grace period wait for tracepoint bpf_link

[Andrii Nakryiko]

On Mon, Mar 30, 2026 at 4:09 PM Andrii Nakryiko
<andrii.nakryiko@gmail.com> wrote:
>
> On Mon, Mar 30, 2026 at 7:31 AM Kumar Kartikeya Dwivedi
> <memxor@gmail.com> wrote:
> >
> > Recently, tracepoints were switched from using disabled preemption
> > (which acts as RCU read section) to SRCU-fast when they are not
> > faultable. This means that to do a proper grace period wait for programs
> > running in such tracepoints, we must use SRCU's grace period wait.
> > This is only for non-faultable tracepoints, faultable ones continue
> > using RCU Tasks Trace.
> >
> > However, bpf_link_free() currently does call_rcu() for all cases when
> > the link is non-sleepable (hence, for tracepoints, non-faultable). Fix
> > this by doing a call_srcu() grace period wait.
> >
> > As far RCU Tasks Trace gp -> RCU gp chaining is concerned, it is deemed
> > unnecessary for tracepoint programs. The link and program are either
> > accessed under RCU Tasks Trace protection, or SRCU-fast protection now.
> >
> > The earlier logic of chaining both RCU Tasks Trace and RCU gp waits was
> > to generalize the logic, even if it conceded an extra RCU gp wait,
> > however that is unnecessary for tracepoints even before this change.
> > In practice no cost was paid since rcu_trace_implies_rcu_gp() was always
> > true. Hence we need not chaing any RCU gp after the SRCU gp.
> >
> > For instance, in the non-faultable raw tracepoint, the RCU read section
> > of the program in __bpf_trace_run() is enclosed in the SRCU gp, likewise
> > for faultable raw tracepoint, the program is under the RCU Tasks Trace
> > protection. Hence, the outermost scope can be waited upon to ensure
> > correctness.
> >
> > Fixes: a46023d5616e ("tracing: Guard __DECLARE_TRACE() use of __DO_TRACE_CALL() with SRCU-fast")
> > Reviewed-by: Puranjay Mohan <puranjay@kernel.org>
> > Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
> > ---
> >  include/linux/tracepoint.h |  8 ++++++++
> >  kernel/bpf/syscall.c       | 26 ++++++++++++++++++++++++--
> >  2 files changed, 32 insertions(+), 2 deletions(-)
> >
> > diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h
> > index 22ca1c8b54f3..8227102a771f 100644
> > --- a/include/linux/tracepoint.h
> > +++ b/include/linux/tracepoint.h
> > @@ -113,6 +113,10 @@ void for_each_tracepoint_in_module(struct module *mod,
> >   */
> >  #ifdef CONFIG_TRACEPOINTS
> >  extern struct srcu_struct tracepoint_srcu;
> > +static inline struct srcu_struct *tracepoint_srcu_ptr(void)
> > +{
> > +       return &tracepoint_srcu;
> > +}
> >  static inline void tracepoint_synchronize_unregister(void)
> >  {
> >         synchronize_rcu_tasks_trace();
> > @@ -123,6 +127,10 @@ static inline bool tracepoint_is_faultable(struct tracepoint *tp)
> >         return tp->ext && tp->ext->faultable;
> >  }
> >  #else
> > +static inline struct srcu_struct *tracepoint_srcu_ptr(void)
> > +{
> > +       return NULL;
> > +}
> >  static inline void tracepoint_synchronize_unregister(void)
> >  { }
> >  static inline bool tracepoint_is_faultable(struct tracepoint *tp)
> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> > index 274039e36465..89fa8f00adfa 100644
> > --- a/kernel/bpf/syscall.c
> > +++ b/kernel/bpf/syscall.c
> > @@ -3261,6 +3261,17 @@ static void bpf_link_defer_dealloc_rcu_gp(struct rcu_head *rcu)
> >         bpf_link_dealloc(link);
> >  }
> >
> > +static bool bpf_link_is_tracepoint(struct bpf_link *link)
> > +{
> > +       /*
> > +        * Only these combinations support a tracepoint bpf_link.
> > +        * BPF_LINK_TYPE_TRACING raw_tp progs are hardcoded to use
> > +        * bpf_raw_tp_link_lops, see bpf_raw_tp_link_attach().
> > +        */
> > +       return link->type == BPF_LINK_TYPE_RAW_TRACEPOINT ||
> > +              (link->type == BPF_LINK_TYPE_TRACING && link->attach_type == BPF_TRACE_RAW_TP);
> > +}
> > +
> >  static void bpf_link_defer_dealloc_mult_rcu_gp(struct rcu_head *rcu)
> >  {
> >         if (rcu_trace_implies_rcu_gp())
> > @@ -3279,16 +3290,27 @@ static void bpf_link_free(struct bpf_link *link)
> >         if (link->prog)
> >                 ops->release(link);
> >         if (ops->dealloc_deferred) {
> > -               /* Schedule BPF link deallocation, which will only then
> > +               struct srcu_struct *tp_srcu = tracepoint_srcu_ptr();
> > +
> > +               /*
> > +                * Schedule BPF link deallocation, which will only then
> >                  * trigger putting BPF program refcount.
> >                  * If underlying BPF program is sleepable or BPF link's target
> >                  * attach hookpoint is sleepable or otherwise requires RCU GPs
> >                  * to ensure link and its underlying BPF program is not
> >                  * reachable anymore, we need to first wait for RCU tasks
> > -                * trace sync, and then go through "classic" RCU grace period
> > +                * trace sync, and then go through "classic" RCU grace period.
> > +                *
> > +                * For tracepoint BPF links, we need to go through SRCU grace
> > +                * period wait instead when non-faultable tracepoint is used. We
> > +                * don't need to chain SRCU grace period waits, however, for the
> > +                * faultable case, since it exclusively uses RCU Tasks Trace.
> >                  */
> >                 if (link->sleepable || (link->prog && link->prog->sleepable))
> >                         call_rcu_tasks_trace(&link->rcu, bpf_link_defer_dealloc_mult_rcu_gp);
> > +               /* We need to do a SRCU grace period wait for tracepoint-based BPF links. */
> > +               else if (bpf_link_is_tracepoint(link) && tp_srcu)
> > +                       call_srcu(tp_srcu, &link->rcu, bpf_link_defer_dealloc_rcu_gp);
>
> It bothers me that we have a special rule here just for tracepoints
> (and non-sleepable ones only), and fetch that srcu_struct...  it feels
> like we are missing some more fundamental link-type-specific grace
> period callback that could be overridden by link ops. E.g., perhaps we
> should be having something like
>
>
> if (ops->call_rcu)
>     ops->call_rcu(link, bpf_link_defer_dealloc_rcu_gp);
> else
>     /* current sleepable or non-sleepable RCU stuff */
>
> and for non-sleepable rawtps we'll have link_ops->call_rcu set to
> something that would fetch that tracepoint_srcu_ptr() and call
> call_srcu().
>
> I don't think we need to solve this right now, but if we ever have
> second case where link's attach hook has some extra lifetime nuances
> like this, we should generalize this. For now:
>
> Acked-by: Andrii Nakryiko <andrii@kernel.org>

there is also that comment in bpf_link_ops next to dealloc_deferred,
see if it should be updated a bit?

>
>
> >                 else
> >                         call_rcu(&link->rcu, bpf_link_defer_dealloc_rcu_gp);
> >         } else if (ops->dealloc) {
> > --
> > 2.52.0
> >

[PATCH bpf v2 2/2] bpf: Retire rcu_trace_implies_rcu_gp()

[Kumar Kartikeya Dwivedi]

RCU Tasks Trace grace period implies RCU grace period, and this
guarantee is expected to remain in the future. Only BPF is the user of
this predicate, hence retire the API and clean up all in-tree users.

RCU Tasks Trace is now implemented on SRCU-fast and its grace period
mechanism always has at least one call to synchronize_rcu() as it is
required for SRCU-fast's correctness (it replaces the smp_mb() that
SRCU-fast readers skip). So, RCU-tt GP will always imply RCU GP.

Reviewed-by: Puranjay Mohan <puranjay@kernel.org>
Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
---
 include/linux/rcupdate.h | 12 ------------
 kernel/bpf/core.c        | 10 ++++------
 kernel/bpf/helpers.c     |  2 +-
 kernel/bpf/memalloc.c    | 33 ++++++++++-----------------------
 kernel/bpf/syscall.c     | 24 ++++++------------------
 5 files changed, 21 insertions(+), 60 deletions(-)

diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index 04f3f86a4145..bfa765132de8 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -205,18 +205,6 @@ static inline void exit_tasks_rcu_start(void) { }
 static inline void exit_tasks_rcu_finish(void) { }
 #endif /* #else #ifdef CONFIG_TASKS_RCU_GENERIC */
 
-/**
- * rcu_trace_implies_rcu_gp - does an RCU Tasks Trace grace period imply an RCU grace period?
- *
- * As an accident of implementation, an RCU Tasks Trace grace period also
- * acts as an RCU grace period.  However, this could change at any time.
- * Code relying on this accident must call this function to verify that
- * this accident is still happening.
- *
- * You have been warned!
- */
-static inline bool rcu_trace_implies_rcu_gp(void) { return true; }
-
 /**
  * cond_resched_tasks_rcu_qs - Report potential quiescent states to RCU
  *
diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
index 7b675a451ec8..1984f061dcf4 100644
--- a/kernel/bpf/core.c
+++ b/kernel/bpf/core.c
@@ -2641,14 +2641,12 @@ static void __bpf_prog_array_free_sleepable_cb(struct rcu_head *rcu)
 {
 	struct bpf_prog_array *progs;
 
-	/* If RCU Tasks Trace grace period implies RCU grace period, there is
-	 * no need to call kfree_rcu(), just call kfree() directly.
+	/*
+	 * RCU Tasks Trace grace period implies RCU grace period, there is no
+	 * need to call kfree_rcu(), just call kfree() directly.
 	 */
 	progs = container_of(rcu, struct bpf_prog_array, rcu);
-	if (rcu_trace_implies_rcu_gp())
-		kfree(progs);
-	else
-		kfree_rcu(progs, rcu);
+	kfree(progs);
 }
 
 void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs)
diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
index 6eb6c82ed2ee..a5a7961f68ec 100644
--- a/kernel/bpf/helpers.c
+++ b/kernel/bpf/helpers.c
@@ -1272,7 +1272,7 @@ static void bpf_async_cb_rcu_tasks_trace_free(struct rcu_head *rcu)
 		return;
 	}
 
-	/* rcu_trace_implies_rcu_gp() is true and will remain so */
+	/* RCU Tasks Trace grace period implies RCU grace period. */
 	bpf_async_cb_rcu_free(rcu);
 }
 
diff --git a/kernel/bpf/memalloc.c b/kernel/bpf/memalloc.c
index 682a9f34214b..e9662db7198f 100644
--- a/kernel/bpf/memalloc.c
+++ b/kernel/bpf/memalloc.c
@@ -284,17 +284,6 @@ static void __free_rcu(struct rcu_head *head)
 	atomic_set(&c->call_rcu_ttrace_in_progress, 0);
 }
 
-static void __free_rcu_tasks_trace(struct rcu_head *head)
-{
-	/* If RCU Tasks Trace grace period implies RCU grace period,
-	 * there is no need to invoke call_rcu().
-	 */
-	if (rcu_trace_implies_rcu_gp())
-		__free_rcu(head);
-	else
-		call_rcu(head, __free_rcu);
-}
-
 static void enque_to_free(struct bpf_mem_cache *c, void *obj)
 {
 	struct llist_node *llnode = obj;
@@ -326,12 +315,12 @@ static void do_call_rcu_ttrace(struct bpf_mem_cache *c)
 		return;
 	}
 
-	/* Use call_rcu_tasks_trace() to wait for sleepable progs to finish.
-	 * If RCU Tasks Trace grace period implies RCU grace period, free
-	 * these elements directly, else use call_rcu() to wait for normal
-	 * progs to finish and finally do free_one() on each element.
+	/*
+	 * Use call_rcu_tasks_trace() to wait for sleepable progs to finish.
+	 * RCU Tasks Trace grace period implies RCU grace period, so pass
+	 * __free_rcu directly as the callback.
 	 */
-	call_rcu_tasks_trace(&c->rcu_ttrace, __free_rcu_tasks_trace);
+	call_rcu_tasks_trace(&c->rcu_ttrace, __free_rcu);
 }
 
 static void free_bulk(struct bpf_mem_cache *c)
@@ -696,20 +685,18 @@ static void free_mem_alloc_no_barrier(struct bpf_mem_alloc *ma)
 
 static void free_mem_alloc(struct bpf_mem_alloc *ma)
 {
-	/* waiting_for_gp[_ttrace] lists were drained, but RCU callbacks
+	/*
+	 * waiting_for_gp[_ttrace] lists were drained, but RCU callbacks
 	 * might still execute. Wait for them.
 	 *
 	 * rcu_barrier_tasks_trace() doesn't imply synchronize_rcu_tasks_trace(),
 	 * but rcu_barrier_tasks_trace() and rcu_barrier() below are only used
-	 * to wait for the pending __free_rcu_tasks_trace() and __free_rcu(),
-	 * so if call_rcu(head, __free_rcu) is skipped due to
-	 * rcu_trace_implies_rcu_gp(), it will be OK to skip rcu_barrier() by
-	 * using rcu_trace_implies_rcu_gp() as well.
+	 * to wait for the pending __free_by_rcu(), and __free_rcu(). RCU Tasks
+	 * Trace grace period implies RCU grace period, so all __free_rcu don't
+	 * need extra call_rcu() (and thus extra rcu_barrier() here).
 	 */
 	rcu_barrier(); /* wait for __free_by_rcu */
 	rcu_barrier_tasks_trace(); /* wait for __free_rcu */
-	if (!rcu_trace_implies_rcu_gp())
-		rcu_barrier();
 	free_mem_alloc_no_barrier(ma);
 }
 
diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 89fa8f00adfa..e27573614f12 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -941,14 +941,6 @@ static void bpf_map_free_rcu_gp(struct rcu_head *rcu)
 	bpf_map_free_in_work(container_of(rcu, struct bpf_map, rcu));
 }
 
-static void bpf_map_free_mult_rcu_gp(struct rcu_head *rcu)
-{
-	if (rcu_trace_implies_rcu_gp())
-		bpf_map_free_rcu_gp(rcu);
-	else
-		call_rcu(rcu, bpf_map_free_rcu_gp);
-}
-
 /* decrement map refcnt and schedule it for freeing via workqueue
  * (underlying map implementation ops->map_free() might sleep)
  */
@@ -959,8 +951,9 @@ void bpf_map_put(struct bpf_map *map)
 		bpf_map_free_id(map);
 
 		WARN_ON_ONCE(atomic64_read(&map->sleepable_refcnt));
+		/* RCU tasks trace grace period implies RCU grace period. */
 		if (READ_ONCE(map->free_after_mult_rcu_gp))
-			call_rcu_tasks_trace(&map->rcu, bpf_map_free_mult_rcu_gp);
+			call_rcu_tasks_trace(&map->rcu, bpf_map_free_rcu_gp);
 		else if (READ_ONCE(map->free_after_rcu_gp))
 			call_rcu(&map->rcu, bpf_map_free_rcu_gp);
 		else
@@ -3272,14 +3265,6 @@ static bool bpf_link_is_tracepoint(struct bpf_link *link)
 	       (link->type == BPF_LINK_TYPE_TRACING && link->attach_type == BPF_TRACE_RAW_TP);
 }
 
-static void bpf_link_defer_dealloc_mult_rcu_gp(struct rcu_head *rcu)
-{
-	if (rcu_trace_implies_rcu_gp())
-		bpf_link_defer_dealloc_rcu_gp(rcu);
-	else
-		call_rcu(rcu, bpf_link_defer_dealloc_rcu_gp);
-}
-
 /* bpf_link_free is guaranteed to be called from process context */
 static void bpf_link_free(struct bpf_link *link)
 {
@@ -3305,9 +3290,12 @@ static void bpf_link_free(struct bpf_link *link)
 		 * period wait instead when non-faultable tracepoint is used. We
 		 * don't need to chain SRCU grace period waits, however, for the
 		 * faultable case, since it exclusively uses RCU Tasks Trace.
+		 *
+		 * RCU Tasks Trace grace period implies RCU grace period, hence
+		 * pass bpf_link_defer_dealloc_rcu_gp as callback directly.
 		 */
 		if (link->sleepable || (link->prog && link->prog->sleepable))
-			call_rcu_tasks_trace(&link->rcu, bpf_link_defer_dealloc_mult_rcu_gp);
+			call_rcu_tasks_trace(&link->rcu, bpf_link_defer_dealloc_rcu_gp);
 		/* We need to do a SRCU grace period wait for tracepoint-based BPF links. */
 		else if (bpf_link_is_tracepoint(link) && tp_srcu)
 			call_srcu(tp_srcu, &link->rcu, bpf_link_defer_dealloc_rcu_gp);
-- 
2.52.0

Re: [PATCH bpf v2 2/2] bpf: Retire rcu_trace_implies_rcu_gp()

[Alexei Starovoitov]

On Mon, Mar 30, 2026 at 7:31 AM Kumar Kartikeya Dwivedi
<memxor@gmail.com> wrote:
>
> RCU Tasks Trace grace period implies RCU grace period, and this
> guarantee is expected to remain in the future. Only BPF is the user of
> this predicate, hence retire the API and clean up all in-tree users.
>
> RCU Tasks Trace is now implemented on SRCU-fast and its grace period
> mechanism always has at least one call to synchronize_rcu() as it is
> required for SRCU-fast's correctness (it replaces the smp_mb() that
> SRCU-fast readers skip). So, RCU-tt GP will always imply RCU GP.
>
> Reviewed-by: Puranjay Mohan <puranjay@kernel.org>
> Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
> Signed-off-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
> ---
>  include/linux/rcupdate.h | 12 ------------
>  kernel/bpf/core.c        | 10 ++++------
>  kernel/bpf/helpers.c     |  2 +-
>  kernel/bpf/memalloc.c    | 33 ++++++++++-----------------------
>  kernel/bpf/syscall.c     | 24 ++++++------------------
>  5 files changed, 21 insertions(+), 60 deletions(-)
>
> diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
> index 04f3f86a4145..bfa765132de8 100644
> --- a/include/linux/rcupdate.h
> +++ b/include/linux/rcupdate.h
> @@ -205,18 +205,6 @@ static inline void exit_tasks_rcu_start(void) { }
>  static inline void exit_tasks_rcu_finish(void) { }
>  #endif /* #else #ifdef CONFIG_TASKS_RCU_GENERIC */
>
> -/**
> - * rcu_trace_implies_rcu_gp - does an RCU Tasks Trace grace period imply an RCU grace period?
> - *
> - * As an accident of implementation, an RCU Tasks Trace grace period also
> - * acts as an RCU grace period.  However, this could change at any time.
> - * Code relying on this accident must call this function to verify that
> - * this accident is still happening.
> - *
> - * You have been warned!
> - */
> -static inline bool rcu_trace_implies_rcu_gp(void) { return true; }
> -

This is not a fix, so respin to bpf-next, and probably as separate patch.

pw-bot: cr