page_pool: fix inconsistency for page_pool_ring_[un]lock()
authorYunsheng Lin <linyunsheng@huawei.com>
Mon, 22 May 2023 03:17:14 +0000 (11:17 +0800)
committerJakub Kicinski <kuba@kernel.org>
Wed, 24 May 2023 03:25:13 +0000 (20:25 -0700)
page_pool_ring_[un]lock() use in_softirq() to decide which
spin lock variant to use, and when they are called in the
context with in_softirq() being false, spin_lock_bh() is
called in page_pool_ring_lock() while spin_unlock() is
called in page_pool_ring_unlock(), because spin_lock_bh()
has disabled the softirq in page_pool_ring_lock(), which
causes inconsistency for spin lock pair calling.

This patch fixes it by returning in_softirq state from
page_pool_producer_lock(), and use it to decide which
spin lock variant to use in page_pool_producer_unlock().

As pool->ring has both producer and consumer lock, so
rename it to page_pool_producer_[un]lock() to reflect
the actual usage. Also move them to page_pool.c as they
are only used there, and remove the 'inline' as the
compiler may have better idea to do inlining or not.

Fixes: 7886244736a4 ("net: page_pool: Add bulk support for ptr_ring")
Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Acked-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Link: https://lore.kernel.org/r/20230522031714.5089-1-linyunsheng@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
include/net/page_pool.h
net/core/page_pool.c

index c8ec2f34722be5484ddaa25bc01d359bdd715d89..126f9e294389ae1100b2e760e962cfb7445580f5 100644 (file)
@@ -399,22 +399,4 @@ static inline void page_pool_nid_changed(struct page_pool *pool, int new_nid)
                page_pool_update_nid(pool, new_nid);
 }
 
-static inline void page_pool_ring_lock(struct page_pool *pool)
-       __acquires(&pool->ring.producer_lock)
-{
-       if (in_softirq())
-               spin_lock(&pool->ring.producer_lock);
-       else
-               spin_lock_bh(&pool->ring.producer_lock);
-}
-
-static inline void page_pool_ring_unlock(struct page_pool *pool)
-       __releases(&pool->ring.producer_lock)
-{
-       if (in_softirq())
-               spin_unlock(&pool->ring.producer_lock);
-       else
-               spin_unlock_bh(&pool->ring.producer_lock);
-}
-
 #endif /* _NET_PAGE_POOL_H */
index e212e9d7edcb816f1c302ab2af76a5005a86137c..a3e12a61d456c4f5ded8b6cb4506e64dda46198c 100644 (file)
@@ -134,6 +134,29 @@ EXPORT_SYMBOL(page_pool_ethtool_stats_get);
 #define recycle_stat_add(pool, __stat, val)
 #endif
 
+static bool page_pool_producer_lock(struct page_pool *pool)
+       __acquires(&pool->ring.producer_lock)
+{
+       bool in_softirq = in_softirq();
+
+       if (in_softirq)
+               spin_lock(&pool->ring.producer_lock);
+       else
+               spin_lock_bh(&pool->ring.producer_lock);
+
+       return in_softirq;
+}
+
+static void page_pool_producer_unlock(struct page_pool *pool,
+                                     bool in_softirq)
+       __releases(&pool->ring.producer_lock)
+{
+       if (in_softirq)
+               spin_unlock(&pool->ring.producer_lock);
+       else
+               spin_unlock_bh(&pool->ring.producer_lock);
+}
+
 static int page_pool_init(struct page_pool *pool,
                          const struct page_pool_params *params)
 {
@@ -617,6 +640,7 @@ void page_pool_put_page_bulk(struct page_pool *pool, void **data,
                             int count)
 {
        int i, bulk_len = 0;
+       bool in_softirq;
 
        for (i = 0; i < count; i++) {
                struct page *page = virt_to_head_page(data[i]);
@@ -635,7 +659,7 @@ void page_pool_put_page_bulk(struct page_pool *pool, void **data,
                return;
 
        /* Bulk producer into ptr_ring page_pool cache */
-       page_pool_ring_lock(pool);
+       in_softirq = page_pool_producer_lock(pool);
        for (i = 0; i < bulk_len; i++) {
                if (__ptr_ring_produce(&pool->ring, data[i])) {
                        /* ring full */
@@ -644,7 +668,7 @@ void page_pool_put_page_bulk(struct page_pool *pool, void **data,
                }
        }
        recycle_stat_add(pool, ring, i);
-       page_pool_ring_unlock(pool);
+       page_pool_producer_unlock(pool, in_softirq);
 
        /* Hopefully all pages was return into ptr_ring */
        if (likely(i == bulk_len))