2023-12-12 17:40:08

by Neeraj upadhyay

[permalink] [raw]
Subject: [PATCH rcu 0/3] RCU misc fixes for v6.8

Hello,

This series contains misc RCU fixes:

1. Restrict access to RCU CPU stall notifiers.

2. Remove unused macros from rcupdate.h, courtesy of Pedro Falcato.

3. Force quiescent states only for ongoing grace period, courtesy
of Zqiang.


Thanks
Neeraj

------------------------------------------------------------------------

Documentation/admin-guide/kernel-parameters.txt | 6 ++++++
include/linux/rcu_notifier.h | 6 +++---
include/linux/rcupdate.h | 3 ---
kernel/rcu/Kconfig.debug | 25 +++++++++++++++++++++++++
kernel/rcu/rcu.h | 8 +++++---
kernel/rcu/rcutorture.c | 12 +++++++-----
kernel/rcu/tree.c | 2 ++
kernel/rcu/tree_stall.h | 11 ++++++++++-
kernel/rcu/update.c | 6 ++++++
9 files changed, 64 insertions(+), 15 deletions(-)


2023-12-12 17:40:58

by Neeraj upadhyay

[permalink] [raw]
Subject: [PATCH rcu 1/3] rcu: Restrict access to RCU CPU stall notifiers

From: "Paul E. McKenney" <[email protected]>

Although the RCU CPU stall notifiers can be useful for dumping state when
tracking down delicate forward-progress bugs where NUMA effects cause
cache lines to be delivered to a given CPU regularly, but always in a
state that prevents that CPU from making forward progress. These bugs can
be detected by the RCU CPU stall-warning mechanism, but in some cases,
the stall-warnings printk()s disrupt the forward-progress bug before
any useful state can be obtained.

Unfortunately, the notifier mechanism added by commit 5b404fdabacf ("rcu:
Add RCU CPU stall notifier") can make matters worse if used at all
carelessly. For example, if the stall warning was caused by a lock not
being released, then any attempt to acquire that lock in the notifier
will hang. This will prevent not only the notifier from producing any
useful output, but it will also prevent the stall-warning message from
ever appearing.

This commit therefore hides this new RCU CPU stall notifier
mechanism under a new RCU_CPU_STALL_NOTIFIER Kconfig option that
depends on both DEBUG_KERNEL and RCU_EXPERT. In addition, the
rcupdate.rcu_cpu_stall_notifiers=1 kernel boot parameter must also
be specified. The RCU_CPU_STALL_NOTIFIER Kconfig option's help text
contains a warning and explains the dangers of careless use, recommending
lockless notifier code. In addition, a WARN() is triggered each time
that an attempt is made to register a stall-warning notifier in kernels
built with CONFIG_RCU_CPU_STALL_NOTIFIER=y.

This combination of measures will keep use of this mechanism confined to
debug kernels and away from routine deployments.

[ paulmck: Apply Dan Carpenter feedback. ]

Fixes: 5b404fdabacf ("rcu: Add RCU CPU stall notifier")
Reported-by: Linus Torvalds <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>
Reviewed-by: Joel Fernandes (Google) <[email protected]>
Signed-off-by: Neeraj Upadhyay (AMD) <[email protected]>
---
.../admin-guide/kernel-parameters.txt | 6 +++++
include/linux/rcu_notifier.h | 6 ++---
kernel/rcu/Kconfig.debug | 25 +++++++++++++++++++
kernel/rcu/rcu.h | 8 +++---
kernel/rcu/rcutorture.c | 12 +++++----
kernel/rcu/tree_stall.h | 11 +++++++-
kernel/rcu/update.c | 6 +++++
7 files changed, 62 insertions(+), 12 deletions(-)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 65731b060e3f..b72e2049c487 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -5302,6 +5302,12 @@
Dump ftrace buffer after reporting RCU CPU
stall warning.

+ rcupdate.rcu_cpu_stall_notifiers= [KNL]
+ Provide RCU CPU stall notifiers, but see the
+ warnings in the RCU_CPU_STALL_NOTIFIER Kconfig
+ option's help text. TL;DR: You almost certainly
+ do not want rcupdate.rcu_cpu_stall_notifiers.
+
rcupdate.rcu_cpu_stall_suppress= [KNL]
Suppress RCU CPU stall warning messages.

diff --git a/include/linux/rcu_notifier.h b/include/linux/rcu_notifier.h
index ebf371364581..5640f024773b 100644
--- a/include/linux/rcu_notifier.h
+++ b/include/linux/rcu_notifier.h
@@ -13,7 +13,7 @@
#define RCU_STALL_NOTIFY_NORM 1
#define RCU_STALL_NOTIFY_EXP 2

-#ifdef CONFIG_RCU_STALL_COMMON
+#if defined(CONFIG_RCU_STALL_COMMON) && defined(CONFIG_RCU_CPU_STALL_NOTIFIER)

#include <linux/notifier.h>
#include <linux/types.h>
@@ -21,12 +21,12 @@
int rcu_stall_chain_notifier_register(struct notifier_block *n);
int rcu_stall_chain_notifier_unregister(struct notifier_block *n);

-#else // #ifdef CONFIG_RCU_STALL_COMMON
+#else // #if defined(CONFIG_RCU_STALL_COMMON) && defined(CONFIG_RCU_CPU_STALL_NOTIFIER)

// No RCU CPU stall warnings in Tiny RCU.
static inline int rcu_stall_chain_notifier_register(struct notifier_block *n) { return -EEXIST; }
static inline int rcu_stall_chain_notifier_unregister(struct notifier_block *n) { return -ENOENT; }

-#endif // #else // #ifdef CONFIG_RCU_STALL_COMMON
+#endif // #else // #if defined(CONFIG_RCU_STALL_COMMON) && defined(CONFIG_RCU_CPU_STALL_NOTIFIER)

#endif /* __LINUX_RCU_NOTIFIER_H */
diff --git a/kernel/rcu/Kconfig.debug b/kernel/rcu/Kconfig.debug
index 2984de629f74..9b0b52e1836f 100644
--- a/kernel/rcu/Kconfig.debug
+++ b/kernel/rcu/Kconfig.debug
@@ -105,6 +105,31 @@ config RCU_CPU_STALL_CPUTIME
The boot option rcupdate.rcu_cpu_stall_cputime has the same function
as this one, but will override this if it exists.

+config RCU_CPU_STALL_NOTIFIER
+ bool "Provide RCU CPU-stall notifiers"
+ depends on RCU_STALL_COMMON
+ depends on DEBUG_KERNEL
+ depends on RCU_EXPERT
+ default n
+ help
+ WARNING: You almost certainly do not want this!!!
+
+ Enable RCU CPU-stall notifiers, which are invoked just before
+ printing the RCU CPU stall warning. As such, bugs in notifier
+ callbacks can prevent stall warnings from being printed.
+ And the whole reason that a stall warning is being printed is
+ that something is hung up somewhere. Therefore, the notifier
+ callbacks must be written extremely carefully, preferably
+ containing only lockless code. After all, it is quite possible
+ that the whole reason that the RCU CPU stall is happening in
+ the first place is that someone forgot to release whatever lock
+ that you are thinking of acquiring. In which case, having your
+ notifier callback acquire that lock will hang, preventing the
+ RCU CPU stall warning from appearing.
+
+ Say Y here if you want RCU CPU stall notifiers (you don't want them)
+ Say N if you are unsure.
+
config RCU_TRACE
bool "Enable tracing for RCU"
depends on DEBUG_KERNEL
diff --git a/kernel/rcu/rcu.h b/kernel/rcu/rcu.h
index b531c33e9545..f94f65877f2b 100644
--- a/kernel/rcu/rcu.h
+++ b/kernel/rcu/rcu.h
@@ -262,6 +262,8 @@ static inline bool rcu_stall_is_suppressed_at_boot(void)
return rcu_cpu_stall_suppress_at_boot && !rcu_inkernel_boot_has_ended();
}

+extern int rcu_cpu_stall_notifiers;
+
#ifdef CONFIG_RCU_STALL_COMMON

extern int rcu_cpu_stall_ftrace_dump;
@@ -659,10 +661,10 @@ static inline bool rcu_cpu_beenfullyonline(int cpu) { return true; }
bool rcu_cpu_beenfullyonline(int cpu);
#endif

-#ifdef CONFIG_RCU_STALL_COMMON
+#if defined(CONFIG_RCU_STALL_COMMON) && defined(CONFIG_RCU_CPU_STALL_NOTIFIER)
int rcu_stall_notifier_call_chain(unsigned long val, void *v);
-#else // #ifdef CONFIG_RCU_STALL_COMMON
+#else // #if defined(CONFIG_RCU_STALL_COMMON) && defined(CONFIG_RCU_CPU_STALL_NOTIFIER)
static inline int rcu_stall_notifier_call_chain(unsigned long val, void *v) { return NOTIFY_DONE; }
-#endif // #else // #ifdef CONFIG_RCU_STALL_COMMON
+#endif // #else // #if defined(CONFIG_RCU_STALL_COMMON) && defined(CONFIG_RCU_CPU_STALL_NOTIFIER)

#endif /* __LINUX_RCU_H */
diff --git a/kernel/rcu/rcutorture.c b/kernel/rcu/rcutorture.c
index 30fc9d34e329..07a6a183c555 100644
--- a/kernel/rcu/rcutorture.c
+++ b/kernel/rcu/rcutorture.c
@@ -2450,10 +2450,12 @@ static int rcu_torture_stall(void *args)
unsigned long stop_at;

VERBOSE_TOROUT_STRING("rcu_torture_stall task started");
- ret = rcu_stall_chain_notifier_register(&rcu_torture_stall_block);
- if (ret)
- pr_info("%s: rcu_stall_chain_notifier_register() returned %d, %sexpected.\n",
- __func__, ret, !IS_ENABLED(CONFIG_RCU_STALL_COMMON) ? "un" : "");
+ if (rcu_cpu_stall_notifiers) {
+ ret = rcu_stall_chain_notifier_register(&rcu_torture_stall_block);
+ if (ret)
+ pr_info("%s: rcu_stall_chain_notifier_register() returned %d, %sexpected.\n",
+ __func__, ret, !IS_ENABLED(CONFIG_RCU_STALL_COMMON) ? "un" : "");
+ }
if (stall_cpu_holdoff > 0) {
VERBOSE_TOROUT_STRING("rcu_torture_stall begin holdoff");
schedule_timeout_interruptible(stall_cpu_holdoff * HZ);
@@ -2497,7 +2499,7 @@ static int rcu_torture_stall(void *args)
cur_ops->readunlock(idx);
}
pr_alert("%s end.\n", __func__);
- if (!ret) {
+ if (rcu_cpu_stall_notifiers && !ret) {
ret = rcu_stall_chain_notifier_unregister(&rcu_torture_stall_block);
if (ret)
pr_info("%s: rcu_stall_chain_notifier_unregister() returned %d.\n", __func__, ret);
diff --git a/kernel/rcu/tree_stall.h b/kernel/rcu/tree_stall.h
index ac8e86babe44..5d666428546b 100644
--- a/kernel/rcu/tree_stall.h
+++ b/kernel/rcu/tree_stall.h
@@ -1061,6 +1061,7 @@ static int __init rcu_sysrq_init(void)
}
early_initcall(rcu_sysrq_init);

+#ifdef CONFIG_RCU_CPU_STALL_NOTIFIER

//////////////////////////////////////////////////////////////////////////////
//
@@ -1081,7 +1082,13 @@ static ATOMIC_NOTIFIER_HEAD(rcu_cpu_stall_notifier_list);
*/
int rcu_stall_chain_notifier_register(struct notifier_block *n)
{
- return atomic_notifier_chain_register(&rcu_cpu_stall_notifier_list, n);
+ int rcsn = rcu_cpu_stall_notifiers;
+
+ WARN(1, "Adding %pS() to RCU stall notifier list (%s).\n", n->notifier_call,
+ rcsn ? "possibly suppressing RCU CPU stall warnings" : "failed, so all is well");
+ if (rcsn)
+ return atomic_notifier_chain_register(&rcu_cpu_stall_notifier_list, n);
+ return -EEXIST;
}
EXPORT_SYMBOL_GPL(rcu_stall_chain_notifier_register);

@@ -1115,3 +1122,5 @@ int rcu_stall_notifier_call_chain(unsigned long val, void *v)
{
return atomic_notifier_call_chain(&rcu_cpu_stall_notifier_list, val, v);
}
+
+#endif // #ifdef CONFIG_RCU_CPU_STALL_NOTIFIER
diff --git a/kernel/rcu/update.c b/kernel/rcu/update.c
index c534d6806d3d..46aaaa9fe339 100644
--- a/kernel/rcu/update.c
+++ b/kernel/rcu/update.c
@@ -538,9 +538,15 @@ long torture_sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
EXPORT_SYMBOL_GPL(torture_sched_setaffinity);
#endif

+int rcu_cpu_stall_notifiers __read_mostly; // !0 = provide stall notifiers (rarely useful)
+EXPORT_SYMBOL_GPL(rcu_cpu_stall_notifiers);
+
#ifdef CONFIG_RCU_STALL_COMMON
int rcu_cpu_stall_ftrace_dump __read_mostly;
module_param(rcu_cpu_stall_ftrace_dump, int, 0644);
+#ifdef CONFIG_RCU_CPU_STALL_NOTIFIER
+module_param(rcu_cpu_stall_notifiers, int, 0444);
+#endif // #ifdef CONFIG_RCU_CPU_STALL_NOTIFIER
int rcu_cpu_stall_suppress __read_mostly; // !0 = suppress stall warnings.
EXPORT_SYMBOL_GPL(rcu_cpu_stall_suppress);
module_param(rcu_cpu_stall_suppress, int, 0644);
--
2.40.1

2023-12-12 17:41:11

by Neeraj upadhyay

[permalink] [raw]
Subject: [PATCH rcu 3/3] rcu: Force quiescent states only for ongoing grace period

From: Zqiang <[email protected]>

If an rcutorture test scenario creates an fqs_task kthread, it will
periodically invoke rcu_force_quiescent_state() in order to start
force-quiescent-state (FQS) operations. However, an FQS operation
will be started even if there is no RCU grace period in progress.
Although testing FQS operations startup when there is no grace period in
progress is necessary, it need not happen all that often. This commit
therefore causes rcu_force_quiescent_state() to take an early exit
if there is no grace period in progress.

Note that there will still be attempts to start an FQS scan in the
absence of a grace period because the grace period might end right
after the rcu_force_quiescent_state() function's check. In actual
testing, this happens about once every ten minutes, which should
provide adequate testing.

Signed-off-by: Zqiang <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>
Signed-off-by: Neeraj Upadhyay (AMD) <[email protected]>
---
kernel/rcu/tree.c | 2 ++
1 file changed, 2 insertions(+)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 3ac3c846105f..1ae851777806 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -2338,6 +2338,8 @@ void rcu_force_quiescent_state(void)
struct rcu_node *rnp;
struct rcu_node *rnp_old = NULL;

+ if (!rcu_gp_in_progress())
+ return;
/* Funnel through hierarchy to reduce memory contention. */
rnp = raw_cpu_read(rcu_data.mynode);
for (; rnp != NULL; rnp = rnp->parent) {
--
2.40.1

2023-12-12 17:41:17

by Neeraj upadhyay

[permalink] [raw]
Subject: [PATCH rcu 2/3] rcu: Remove unused macros from rcupdate.h

From: Pedro Falcato <[email protected]>

ulong2long, USHORT_CMP_GE and USHORT_CMP_LT are redundant and have been
unused for quite a few releases.

Signed-off-by: Pedro Falcato <[email protected]>
Signed-off-by: Paul E. McKenney <[email protected]>
Signed-off-by: Neeraj Upadhyay (AMD) <[email protected]>
---
include/linux/rcupdate.h | 3 ---
1 file changed, 3 deletions(-)

diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index f7206b2623c9..aa87c82236dd 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -34,9 +34,6 @@

#define ULONG_CMP_GE(a, b) (ULONG_MAX / 2 >= (a) - (b))
#define ULONG_CMP_LT(a, b) (ULONG_MAX / 2 < (a) - (b))
-#define ulong2long(a) (*(long *)(&(a)))
-#define USHORT_CMP_GE(a, b) (USHRT_MAX / 2 >= (unsigned short)((a) - (b)))
-#define USHORT_CMP_LT(a, b) (USHRT_MAX / 2 < (unsigned short)((a) - (b)))

/* Exported common interfaces */
void call_rcu(struct rcu_head *head, rcu_callback_t func);
--
2.40.1

2023-12-12 23:22:58

by Joel Fernandes

[permalink] [raw]
Subject: Re: [PATCH rcu 3/3] rcu: Force quiescent states only for ongoing grace period

On Tue, Dec 12, 2023 at 12:40 PM Neeraj Upadhyay (AMD)
<[email protected]> wrote:
>
> From: Zqiang <[email protected]>
>
> If an rcutorture test scenario creates an fqs_task kthread, it will
> periodically invoke rcu_force_quiescent_state() in order to start
> force-quiescent-state (FQS) operations. However, an FQS operation
> will be started even if there is no RCU grace period in progress.
> Although testing FQS operations startup when there is no grace period in
> progress is necessary, it need not happen all that often. This commit
> therefore causes rcu_force_quiescent_state() to take an early exit
> if there is no grace period in progress.
>
> Note that there will still be attempts to start an FQS scan in the
> absence of a grace period because the grace period might end right
> after the rcu_force_quiescent_state() function's check. In actual
> testing, this happens about once every ten minutes, which should
> provide adequate testing.
>
> Signed-off-by: Zqiang <[email protected]>
> Signed-off-by: Paul E. McKenney <[email protected]>
> Signed-off-by: Neeraj Upadhyay (AMD) <[email protected]>
> ---
> kernel/rcu/tree.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index 3ac3c846105f..1ae851777806 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -2338,6 +2338,8 @@ void rcu_force_quiescent_state(void)
> struct rcu_node *rnp;
> struct rcu_node *rnp_old = NULL;
>
> + if (!rcu_gp_in_progress())
> + return;

Reviewed-by: Joel Fernandes (Google) <[email protected]>

(Probably no need to resend with the tag, just update the patch in the PR).

Thanks.