Merge branch 'report-rcu-qs-for-busy-network-kthreads'

Yan Zhai says:

====================
Report RCU QS for busy network kthreads

This changeset fixes a common problem for busy networking kthreads.
These threads, e.g. NAPI threads, typically will do:

* polling a batch of packets
* if there are more work, call cond_resched() to allow scheduling
* continue to poll more packets when rx queue is not empty

We observed this being a problem in production, since it can block RCU
tasks from making progress under heavy load. Investigation indicates
that just calling cond_resched() is insufficient for RCU tasks to reach
quiescent states. This also has the side effect of frequently clearing
the TIF_NEED_RESCHED flag on voluntary preempt kernels. As a result,
schedule() will not be called in these circumstances, despite schedule()
in fact provides required quiescent states. This at least affects NAPI
threads, napi_busy_loop, and also cpumap kthread.

By reporting RCU QSes in these kthreads periodically before cond_resched, the
blocked RCU waiters can correctly progress. Instead of just reporting QS for
RCU tasks, these code share the same concern as noted in the commit
d28139c4e967 ("rcu: Apply RCU-bh QSes to RCU-sched and RCU-preempt when safe").
So report a consolidated QS for safety.

It is worth noting that, although this problem is reproducible in
napi_busy_loop, it only shows up when setting the polling interval to as high
as 2ms, which is far larger than recommended 50us-100us in the documentation.
So napi_busy_loop is left untouched.

Lastly, this does not affect RT kernels, which does not enter the scheduler
through cond_resched(). Without the mentioned side effect, schedule() will
be called time by time, and clear the RCU task holdouts.

V4: https://lore.kernel.org/bpf/cover.1710525524.git.yan@cloudflare.com/
V3: https://lore.kernel.org/lkml/20240314145459.7b3aedf1@kernel.org/t/
V2: https://lore.kernel.org/bpf/ZeFPz4D121TgvCje@debian.debian/
V1: https://lore.kernel.org/lkml/Zd4DXTyCf17lcTfq@debian.debian/#t
====================

Link: https://lore.kernel.org/r/cover.1710877680.git.yan@cloudflare.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

3 files changed, 37 insertions, 0 deletions

diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index 16f519914415..17d7ed5f3ae6 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -247,6 +247,37 @@ do { \
 	cond_resched(); \
 } while (0)
 
+/**
+ * rcu_softirq_qs_periodic - Report RCU and RCU-Tasks quiescent states
+ * @old_ts: jiffies at start of processing.
+ *
+ * This helper is for long-running softirq handlers, such as NAPI threads in
+ * networking. The caller should initialize the variable passed in as @old_ts
+ * at the beginning of the softirq handler. When invoked frequently, this macro
+ * will invoke rcu_softirq_qs() every 100 milliseconds thereafter, which will
+ * provide both RCU and RCU-Tasks quiescent states. Note that this macro
+ * modifies its old_ts argument.
+ *
+ * Because regions of code that have disabled softirq act as RCU read-side
+ * critical sections, this macro should be invoked with softirq (and
+ * preemption) enabled.
+ *
+ * The macro is not needed when CONFIG_PREEMPT_RT is defined. RT kernels would
+ * have more chance to invoke schedule() calls and provide necessary quiescent
+ * states. As a contrast, calling cond_resched() only won't achieve the same
+ * effect because cond_resched() does not provide RCU-Tasks quiescent states.
+ */
+#define rcu_softirq_qs_periodic(old_ts) \
+do { \
+	if (!IS_ENABLED(CONFIG_PREEMPT_RT) && \
+	    time_after(jiffies, (old_ts) + HZ / 10)) { \
+		preempt_disable(); \
+		rcu_softirq_qs(); \
+		preempt_enable(); \
+		(old_ts) = jiffies; \
+	} \
+} while (0)
+
 /*
  * Infrastructure to implement the synchronize_() primitives in
  * TREE_RCU and rcu_barrier_() primitives in TINY_RCU.
diff --git a/kernel/bpf/cpumap.c b/kernel/bpf/cpumap.c
index 9ee8da477465..a8e34416e960 100644
--- a/kernel/bpf/cpumap.c
+++ b/kernel/bpf/cpumap.c
@@ -263,6 +263,7 @@ static int cpu_map_bpf_prog_run(struct bpf_cpu_map_entry *rcpu, void **frames,
 static int cpu_map_kthread_run(void *data)
 {
 	struct bpf_cpu_map_entry *rcpu = data;
+	unsigned long last_qs = jiffies;
 
 	complete(&rcpu->kthread_running);
 	set_current_state(TASK_INTERRUPTIBLE);
@@ -288,10 +289,12 @@ static int cpu_map_kthread_run(void *data)
 			if (__ptr_ring_empty(rcpu->queue)) {
 				schedule();
 				sched = 1;
+				last_qs = jiffies;
 			} else {
 				__set_current_state(TASK_RUNNING);
 			}
 		} else {
+			rcu_softirq_qs_periodic(last_qs);
 			sched = cond_resched();
 		}
 
diff --git a/net/core/dev.c b/net/core/dev.c
index 303a6ff46e4e..9a67003e49db 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -6743,6 +6743,8 @@ static int napi_threaded_poll(void *data)
 	void *have;
 
 	while (!napi_thread_wait(napi)) {
+		unsigned long last_qs = jiffies;
+
 		for (;;) {
 			bool repoll = false;
 
@@ -6767,6 +6769,7 @@ static int napi_threaded_poll(void *data)
 			if (!repoll)
 				break;
 
+			rcu_softirq_qs_periodic(last_qs);
 			cond_resched();
 		}
 	}