Allocating memory with GFP_KERNEL may cause the thread to be put into
sleep, and doing so within RCU read-side critical section is problematic
as it might cause the grace period to end early and trigger unexpected
objects destruction. Involve rcu_read_lock() into locks/call_kern.cocci,
as this requirement is analogous to other kinds of lock.

Signed-off-by: GONG, Ruiqi <gongruiqi1@huawei.com>
---
 scripts/coccinelle/locks/call_kern.cocci | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/scripts/coccinelle/locks/call_kern.cocci b/scripts/coccinelle/locks/call_kern.cocci
index 3720f86d67c5..89665ea61642 100644
--- a/scripts/coccinelle/locks/call_kern.cocci
+++ b/scripts/coccinelle/locks/call_kern.cocci
@@ -31,6 +31,7 @@ fn(...) {
      when != spin_unlock_irq(...)
      when != spin_unlock_irqrestore(...)
      when != local_irq_enable(...)
+     when != rcu_read_unlock(...)
      when any
  GFP_KERNEL@p
  ... when any
@@ -59,6 +60,8 @@ spin_lock_irq@p1
 spin_lock_irqsave@p1
 |
 local_irq_disable@p1
+|
+rcu_read_lock@p1
 )
  (...)
 ...  when != read_unlock_irq(...)
@@ -69,6 +72,7 @@ local_irq_disable@p1
      when != spin_unlock_irq(...)
      when != spin_unlock_irqrestore(...)
      when != local_irq_enable(...)
+     when != rcu_read_unlock(...)
 fn@p2(...)
 
 @depends on locked && patch@
-- 
2.25.1