提交 86357b19 编写于 作者: A Andy Grover 提交者: David S. Miller

RDS: Fix potential race around rds_i[bw]_allocation

"At rds_ib_recv_refill_one(), it first executes atomic_read(&rds_ib_allocation)
for if-condition checking,

and then executes atomic_inc(&rds_ib_allocation) if the condition was
not satisfied.

However, if any other code which updates rds_ib_allocation executes
between these two atomic operation executions,
it seems that it may result race condition. (especially when
rds_ib_allocation + 1 == rds_ib_sysctl_max_recv_allocation)"

This patch fixes this by using atomic_inc_unless to eliminate the
possibility of allocating more than rds_ib_sysctl_max_recv_allocation
and then decrementing the count if the allocation fails. It also
makes an identical change to the iwarp transport.
Reported-by: NShin Hong <hongshin@gmail.com>
Signed-off-by: NAndy Grover <andy.grover@oracle.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
上级 244546f0
...@@ -143,15 +143,16 @@ static int rds_ib_recv_refill_one(struct rds_connection *conn, ...@@ -143,15 +143,16 @@ static int rds_ib_recv_refill_one(struct rds_connection *conn,
int ret = -ENOMEM; int ret = -ENOMEM;
if (recv->r_ibinc == NULL) { if (recv->r_ibinc == NULL) {
if (atomic_read(&rds_ib_allocation) >= rds_ib_sysctl_max_recv_allocation) { if (!atomic_add_unless(&rds_ib_allocation, 1, rds_ib_sysctl_max_recv_allocation)) {
rds_ib_stats_inc(s_ib_rx_alloc_limit); rds_ib_stats_inc(s_ib_rx_alloc_limit);
goto out; goto out;
} }
recv->r_ibinc = kmem_cache_alloc(rds_ib_incoming_slab, recv->r_ibinc = kmem_cache_alloc(rds_ib_incoming_slab,
kptr_gfp); kptr_gfp);
if (recv->r_ibinc == NULL) if (recv->r_ibinc == NULL) {
atomic_dec(&rds_ib_allocation);
goto out; goto out;
atomic_inc(&rds_ib_allocation); }
INIT_LIST_HEAD(&recv->r_ibinc->ii_frags); INIT_LIST_HEAD(&recv->r_ibinc->ii_frags);
rds_inc_init(&recv->r_ibinc->ii_inc, conn, conn->c_faddr); rds_inc_init(&recv->r_ibinc->ii_inc, conn, conn->c_faddr);
} }
......
...@@ -143,15 +143,16 @@ static int rds_iw_recv_refill_one(struct rds_connection *conn, ...@@ -143,15 +143,16 @@ static int rds_iw_recv_refill_one(struct rds_connection *conn,
int ret = -ENOMEM; int ret = -ENOMEM;
if (recv->r_iwinc == NULL) { if (recv->r_iwinc == NULL) {
if (atomic_read(&rds_iw_allocation) >= rds_iw_sysctl_max_recv_allocation) { if (!atomic_add_unless(&rds_iw_allocation, 1, rds_iw_sysctl_max_recv_allocation)) {
rds_iw_stats_inc(s_iw_rx_alloc_limit); rds_iw_stats_inc(s_iw_rx_alloc_limit);
goto out; goto out;
} }
recv->r_iwinc = kmem_cache_alloc(rds_iw_incoming_slab, recv->r_iwinc = kmem_cache_alloc(rds_iw_incoming_slab,
kptr_gfp); kptr_gfp);
if (recv->r_iwinc == NULL) if (recv->r_iwinc == NULL) {
atomic_dec(&rds_iw_allocation);
goto out; goto out;
atomic_inc(&rds_iw_allocation); }
INIT_LIST_HEAD(&recv->r_iwinc->ii_frags); INIT_LIST_HEAD(&recv->r_iwinc->ii_frags);
rds_inc_init(&recv->r_iwinc->ii_inc, conn, conn->c_faddr); rds_inc_init(&recv->r_iwinc->ii_inc, conn, conn->c_faddr);
} }
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册