[PATCH net] napi: fix race inside napi_enable

The process will cause napi.state to contain NAPI_STATE_SCHED and
not in the poll_list, which will cause napi_disable() to get stuck.

The prefix "NAPI_STATE_" is removed in the figure below, and
NAPI_STATE_HASHED is ignored in napi.state.

                      CPU0       |                   CPU1       | napi.state
===============================================================================
napi_disable()                   |                              | SCHED | NPSVC
napi_enable()                    |                              |
{                                |                              |
    smp_mb__before_atomic();     |                              |
    clear_bit(SCHED, &n->state); |                              | NPSVC
                                 | napi_schedule_prep()         | SCHED | NPSVC
                                 | napi_poll()                  |
                                 |   napi_complete_done()       |
                                 |   {                          |
                                 |      if (n->state & (NPSVC | | (1)
                                 |               _BUSY_POLL)))  |
                                 |           return false;      |
                                 |     ................         |
                                 |   }                          | SCHED | NPSVC
                                 |                              |
    clear_bit(NPSVC, &n->state); |                              | SCHED
}                                |                              |
                                 |                              |
napi_schedule_prep()             |                              | SCHED | MISSED (2)

(1) Here return direct. Because of NAPI_STATE_NPSVC exists.
(2) NAPI_STATE_SCHED exists. So not add napi.poll_list to sd->poll_list

Since NAPI_STATE_SCHED already exists and napi is not in the
sd->poll_list queue, NAPI_STATE_SCHED cannot be cleared and will always
exist.

1. This will cause this queue to no longer receive packets.
2. If you encounter napi_disable under the protection of rtnl_lock, it
   will cause the entire rtnl_lock to be locked, affecting the overall
   system.

This patch uses cmpxchg to implement napi_enable(), which ensures that
there will be no race due to the separation of clear two bits.

Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Reviewed-by: Dust Li <dust.li@linux.alibaba.com>
---
 net/core/dev.c | 17 ++++++++++++-----
 1 file changed, 12 insertions(+), 5 deletions(-)

diff --git a/net/core/dev.c b/net/core/dev.c
index 74fd402d26dd..3457ae964b8c 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c

@@ -6923,12 +6923,19 @@ EXPORT_SYMBOL(napi_disable);
  */
 void napi_enable(struct napi_struct *n)
 {
+	unsigned long val, new;
+
 	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
-	smp_mb__before_atomic();
-	clear_bit(NAPI_STATE_SCHED, &n->state);
-	clear_bit(NAPI_STATE_NPSVC, &n->state);
-	if (n->dev->threaded && n->thread)
-		set_bit(NAPI_STATE_THREADED, &n->state);
+
+	do {
+		val = READ_ONCE(n->state);
+
+		new = val & ~(NAPIF_STATE_SCHED | NAPIF_STATE_NPSVC);
+
+		if (n->dev->threaded && n->thread)
+			new |= NAPIF_STATE_THREADED;
+
+	} while (cmpxchg(&n->state, val, new) != val);
 }
 EXPORT_SYMBOL(napi_enable);
 

-- 
2.31.0

On 2021/9/18 14:48, Xuan Zhuo wrote:

The process will cause napi.state to contain NAPI_STATE_SCHED and
not in the poll_list, which will cause napi_disable() to get stuck.

The prefix "NAPI_STATE_" is removed in the figure below, and
NAPI_STATE_HASHED is ignored in napi.state.

                      CPU0       |                   CPU1       | napi.state
===============================================================================
napi_disable()                   |                              | SCHED | NPSVC
napi_enable()                    |                              |
{                                |                              |
    smp_mb__before_atomic();     |                              |
    clear_bit(SCHED, &n->state); |                              | NPSVC
                                 | napi_schedule_prep()         | SCHED | NPSVC
                                 | napi_poll()                  |
                                 |   napi_complete_done()       |
                                 |   {                          |
                                 |      if (n->state & (NPSVC | | (1)
                                 |               _BUSY_POLL)))  |
                                 |           return false;      |
                                 |     ................         |
                                 |   }                          | SCHED | NPSVC
                                 |                              |
    clear_bit(NPSVC, &n->state); |                              | SCHED
}                                |                              |
                                 |                              |
napi_schedule_prep()             |                              | SCHED | MISSED (2)

(1) Here return direct. Because of NAPI_STATE_NPSVC exists.
(2) NAPI_STATE_SCHED exists. So not add napi.poll_list to sd->poll_list

Since NAPI_STATE_SCHED already exists and napi is not in the
sd->poll_list queue, NAPI_STATE_SCHED cannot be cleared and will always
exist.

1. This will cause this queue to no longer receive packets.
2. If you encounter napi_disable under the protection of rtnl_lock, it
   will cause the entire rtnl_lock to be locked, affecting the overall
   system.

This patch uses cmpxchg to implement napi_enable(), which ensures that
there will be no race due to the separation of clear two bits.

a Fixes tag is needed here.

It seems we might need to move the above BUG_ON here to preserve
the orginial BUG_ON behavior?

+
+		new = val & ~(NAPIF_STATE_SCHED | NAPIF_STATE_NPSVC);
+
+		if (n->dev->threaded && n->thread)
+			new |= NAPIF_STATE_THREADED;
+
+	} while (cmpxchg(&n->state, val, new) != val);
 }
 EXPORT_SYMBOL(napi_enable);