[PATCH v3] sunrpc: fix lost wakeups when stopping service threads

[Kal Cutter Conley <kalcutterc@nvidia.com>]

Stopping NFS can leave rpc.nfsd 0 in an uninterruptible wait with
this stack:

  svc_stop_kthreads [sunrpc]
  svc_set_num_threads [sunrpc]
  nfsd_svc [nfsd]
  write_threads [nfsd]
  nfsctl_transaction_write [nfsd]

SP_NEED_VICTIM and SP_VICTIM_REMAINS remain set while all remaining
nfsd threads sleep in svc_recv() and none has RQ_VICTIM set.

The stop request is published with unordered set_bit() operations.
svc_pool_wake_idle_thread() checks the idle list and task state before
calling wake_up_process(). If it sees an empty list or TASK_RUNNING,
it skips the wakeup and its memory barrier. The requester can therefore
miss a sleeping worker while that worker misses the stop request.

Use smp_mb__after_atomic() after the flag updates in
svc_stop_kthreads() and svc_wake_up() to order work publication before
both lockless checks, even when no wakeup occurs. Pair with
set_current_state() and llist_add() in svc_thread_wait_for_work(), so
the worker observes the pending work or the requester observes the
idle worker and wakes it.

Fixes: 60749cbe3d8a ("sunrpc: change sp_nrthreads from atomic_t to unsigned int.")
Signed-off-by: Kal Cutter Conley <kalcutterc@nvidia.com>
---
Changes in v3:
- Drop the redundant <asm/barrier.h> include from svc.c.

Changes in v2:
- Move the barrier into svc_stop_kthreads() and svc_wake_up(), using
  smp_mb__after_atomic(). This avoids adding a redundant barrier on the
  transport enqueue path.

v1: https://lore.kernel.org/r/20261030130629.103716-1-kalcutterc@nvidia.com

 net/sunrpc/svc.c      | 5 +++++
 net/sunrpc/svc_xprt.c | 5 +++++
 2 files changed, 10 insertions(+)

diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
index af8d87c1de83..6100603e0a34 100644
--- a/net/sunrpc/svc.c
+++ b/net/sunrpc/svc.c
@@ -784,6 +784,11 @@ svc_stop_kthreads(struct svc_serv *serv, struct svc_pool *pool, int nrservs)
 	do {
 		set_bit(SP_VICTIM_REMAINS, &pool->sp_flags);
 		set_bit(SP_NEED_VICTIM, &pool->sp_flags);
+		/*
+		 * Order flag updates before the idle list and task state checks.
+		 * Pairs with the barriers in svc_thread_wait_for_work().
+		 */
+		smp_mb__after_atomic();
 		svc_pool_wake_idle_thread(pool);
 		wait_on_bit(&pool->sp_flags, SP_VICTIM_REMAINS, TASK_IDLE);
 		nrservs++;
diff --git a/net/sunrpc/svc_xprt.c b/net/sunrpc/svc_xprt.c
index 076c624fe675..868e3da57b6c 100644
--- a/net/sunrpc/svc_xprt.c
+++ b/net/sunrpc/svc_xprt.c
@@ -632,6 +632,11 @@ void svc_wake_up(struct svc_serv *serv)
 	struct svc_pool *pool = &serv->sv_pools[0];
 
 	set_bit(SP_TASK_PENDING, &pool->sp_flags);
+	/*
+	 * Order flag updates before the idle list and task state checks.
+	 * Pairs with the barriers in svc_thread_wait_for_work().
+	 */
+	smp_mb__after_atomic();
 	svc_pool_wake_idle_thread(pool);
 }
 EXPORT_SYMBOL_GPL(svc_wake_up);

base-commit: 2d3259f652bb0f1d137ccabecac6763da05064ec
-- 
2.53.0