[RFC v2 16/18] rcu: Check actual RCU_GP_FLAG_FQS when handling quiescent state
From: Petr Mladek
Date: Mon Sep 21 2015 - 09:06:58 EST
This change will help a lot if we want to split RCU kthreads into
few kthread works.
The variable "qf" was added by the commit 88d6df612cc3c99f56 ("rcu: Prevent
spurious-wakeup DoS attack on rcu_gp_kthread()"). IMHO, the primary fix
is the "ret" handling.
If I read the code correctly, RCU_GP_FLAG_FQS should not get lost
at this stage. rsp->gp_flags are written in these functions:
+ rcu_gp_init()
+ rcu_gp_cleanup()
+ Both are safe. They are called from other part of the kthread
that can't get accessed when "gf" is used.
+ rcu_gp_fqs()
+ Safe. It is the function that we call when the flag is set.
+ rcu_report_qs_rsp()
+ force_quiescent_state()
+ Both should be safe. They set RCU_GP_FLAG_FQS and therefore they
could not cause loss of rcu_gp_fqs() call.
+ rcu_start_gp_advanced()
+ This should be safe. It is called only when the last grace period
was completed and we are opening a new one. Therefore it should
not happen when "gf" is set and tested because the grace period
is closed by the kthread.
Signed-off-by: Petr Mladek <pmladek@xxxxxxxx>
---
kernel/rcu/tree.c | 10 ++++------
1 file changed, 4 insertions(+), 6 deletions(-)
diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 286e300794f0..08d1d3e63b9b 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -1908,13 +1908,12 @@ static int rcu_gp_init(struct rcu_state *rsp)
* Helper function for wait_event_interruptible_timeout() wakeup
* at force-quiescent-state time.
*/
-static bool rcu_gp_fqs_check_wake(struct rcu_state *rsp, int *gfp)
+static bool rcu_gp_fqs_check_wake(struct rcu_state *rsp)
{
struct rcu_node *rnp = rcu_get_root(rsp);
/* Someone like call_rcu() requested a force-quiescent-state scan. */
- *gfp = READ_ONCE(rsp->gp_flags);
- if (*gfp & RCU_GP_FLAG_FQS)
+ if (READ_ONCE(rsp->gp_flags) & RCU_GP_FLAG_FQS)
return true;
/* The current grace period has completed. */
@@ -2072,7 +2071,6 @@ static unsigned long normalize_jiffies_till_next_fqs(void)
*/
static int __noreturn rcu_gp_kthread(void *arg)
{
- int gf;
unsigned long timeout, j;
struct rcu_state *rsp = arg;
struct rcu_node *rnp = rcu_get_root(rsp);
@@ -2111,7 +2109,7 @@ static int __noreturn rcu_gp_kthread(void *arg)
TPS("fqswait"));
rsp->gp_state = RCU_GP_WAIT_FQS;
wait_event_interruptible_timeout(rsp->gp_wq,
- rcu_gp_fqs_check_wake(rsp, &gf),
+ rcu_gp_fqs_check_wake(rsp),
timeout);
rsp->gp_state = RCU_GP_DOING_FQS;
try_again:
@@ -2122,7 +2120,7 @@ try_again:
break;
/* If time for quiescent-state forcing, do it. */
if (ULONG_CMP_GE(jiffies, rsp->jiffies_force_qs) ||
- (gf & RCU_GP_FLAG_FQS)) {
+ (READ_ONCE(rsp->gp_flags) & RCU_GP_FLAG_FQS)) {
trace_rcu_grace_period(rsp->name,
READ_ONCE(rsp->gpnum),
TPS("fqsstart"));
--
1.8.5.6
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/