On Fri, Jan 19, 2024 at 11:27 AM Zhengchao ShaoHi Eric:
<shaozhengchao@xxxxxxxxxx> wrote:
I analyze the potential sleeping issue of the following processes:
Thread A Thread B
... netlink_create //ref = 1
do_mq_notify ...
sock = netlink_getsockbyfilp ... //ref = 2
info->notify_sock = sock; ...
... netlink_sendmsg
... skb = netlink_alloc_large_skb //skb->head is vmalloced
... netlink_unicast
... sk = netlink_getsockbyportid //ref = 3
... netlink_sendskb
... __netlink_sendskb
... skb_queue_tail //put skb to sk_receive_queue
... sock_put //ref = 2
... ...
... netlink_release
... deferred_put_nlk_sk //ref = 1
mqueue_flush_file
spin_lock
remove_notification
netlink_sendskb
sock_put //ref = 0
sk_free
...
__sk_destruct
netlink_sock_destruct
skb_queue_purge //get skb from sk_receive_queue
...
__skb_queue_purge_reason
kfree_skb_reason
__kfree_skb
...
skb_release_all
skb_release_head_state
netlink_skb_destructor
vfree(skb->head) //sleeping while holding spinlock
In netlink_sendmsg, if the memory pointed to by skb->head is allocated by
vmalloc, and is put to sk_receive_queue queue, also the skb is not freed.
When the mqueue executes flush, the sleeping bug will occur. Put sock
after releasing the spinlock.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
I think netlink started to use vmalloc() from commit c05cdb1b864fThank you for your review. Yes, you are right. Sorry for my mistake.
("netlink: allow large data transfers from user-space")
OK, I will send v4 after verification.Signed-off-by: Zhengchao Shao <shaozhengchao@xxxxxxxxxx>
---
v3: Put sock after releasing the spinlock.
v2: CCed some networking maintainer & netdev list
---
ipc/mqueue.c | 15 +++++++++++++--
1 file changed, 13 insertions(+), 2 deletions(-)
diff --git a/ipc/mqueue.c b/ipc/mqueue.c
index 5eea4dc0509e..4832343b7049 100644
--- a/ipc/mqueue.c
+++ b/ipc/mqueue.c
@@ -664,12 +664,23 @@ static ssize_t mqueue_read_file(struct file *filp, char __user *u_data,
static int mqueue_flush_file(struct file *filp, fl_owner_t id)
{
struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
+ struct sock *sk = NULL;
spin_lock(&info->lock);
- if (task_tgid(current) == info->notify_owner)
- remove_notification(info);
+ if (task_tgid(current) == info->notify_owner) {
+ if (info->notify_owner != NULL &&
+ info->notify.sigev_notify == SIGEV_THREAD) {
+ sk = info->notify_sock;
+ sock_hold(sk);
+ }
+ remove_notification(info);
+ }
spin_unlock(&info->lock);
+
+ if (sk)
+ sock_put(sk);
+
return 0;
}
Note that we could instead call vfree_atomic() from netlink_skb_destructor()
diff --git a/net/netlink/af_netlink.c b/net/netlink/af_netlink.c
index 4ed8ffd58ff375f3fa9f262e6f3b4d1a1aaf2731..9c962347cf859f16fc76e4d8a2fd22cdb3d142d6
100644
--- a/net/netlink/af_netlink.c
+++ b/net/netlink/af_netlink.c
@@ -374,7 +374,7 @@ static void netlink_skb_destructor(struct sk_buff *skb)
if (is_vmalloc_addr(skb->head)) {
if (!skb->cloned ||
!atomic_dec_return(&(skb_shinfo(skb)->dataref)))
- vfree(skb->head);
+ vfree_atomic(skb->head);
skb->head = NULL;
}
These big skbs are quite rare IMO, and we also could attemptIt looks good to me. I would like to do it if you want...
high-order allocations
in netlink_alloc_large_skb(), using kvmalloc() instead of vmalloc()
(next week when net-next opens)