syzbot sent a hung task report and Eric explains that adversarial
receiver may keep RWIN at 0 for a long time, so we are not guaranteed
to make forward progress. Thread which took tx_lock and went to sleep
may not release tx_lock for hours. Use interruptible sleep where
possible and reschedule the work if it can't take the lock.
Testing: existing selftest passes
Reported-by: syzbot+9c0268252b8ef967c62e@xxxxxxxxxxxxxxxxxxxxxxxxx
Fixes: 79ffe6087e91 ("net/tls: add a TX lock")
Link: https://lore.kernel.org/all/000000000000e412e905f5b46201@xxxxxxxxxx/
Cc: stable@xxxxxxxxxxxxxxx # wait 4 weeks
Signed-off-by: Jakub Kicinski <kuba@xxxxxxxxxx>
---
CC: borisp@xxxxxxxxxx
CC: john.fastabend@xxxxxxxxx
CC: simon.horman@xxxxxxxxxxxxx
---
net/tls/tls_sw.c | 26 +++++++++++++++++++-------
1 file changed, 19 insertions(+), 7 deletions(-)
diff --git a/net/tls/tls_sw.c b/net/tls/tls_sw.c
index 021d760f9133..635b8bf6b937 100644
--- a/net/tls/tls_sw.c
+++ b/net/tls/tls_sw.c
@@ -956,7 +956,9 @@ int tls_sw_sendmsg(struct sock *sk, struct msghdr *msg, size_t size)
MSG_CMSG_COMPAT))
return -EOPNOTSUPP;
- mutex_lock(&tls_ctx->tx_lock);
+ ret = mutex_lock_interruptible(&tls_ctx->tx_lock);
+ if (ret)
+ return ret;
lock_sock(sk);
if (unlikely(msg->msg_controllen)) {
@@ -1290,7 +1292,9 @@ int tls_sw_sendpage(struct sock *sk, struct page *page,
MSG_SENDPAGE_NOTLAST | MSG_SENDPAGE_NOPOLICY))
return -EOPNOTSUPP;
- mutex_lock(&tls_ctx->tx_lock);
+ ret = mutex_lock_interruptible(&tls_ctx->tx_lock);
+ if (ret)
+ return ret;
lock_sock(sk);
ret = tls_sw_do_sendpage(sk, page, offset, size, flags);
release_sock(sk);
@@ -2435,11 +2439,19 @@ static void tx_work_handler(struct work_struct *work)
if (!test_and_clear_bit(BIT_TX_SCHEDULED, &ctx->tx_bitmask))
return;
- mutex_lock(&tls_ctx->tx_lock);
- lock_sock(sk);
- tls_tx_records(sk, -1);
- release_sock(sk);
- mutex_unlock(&tls_ctx->tx_lock);
+
+ if (mutex_trylock(&tls_ctx->tx_lock)) {
+ lock_sock(sk);
+ tls_tx_records(sk, -1);
+ release_sock(sk);
+ mutex_unlock(&tls_ctx->tx_lock);
+ } else if (!test_and_set_bit(BIT_TX_SCHEDULED, &ctx->tx_bitmask)) {
+ /* Someone is holding the tx_lock, they will likely run Tx
+ * and cancel the work on their way out of the lock section.
+ * Schedule a long delay just in case.
+ */
+ schedule_delayed_work(&ctx->tx_work.work, msecs_to_jiffies(10));
+ }
}
static bool tls_is_tx_ready(struct tls_sw_context_tx *ctx)
--
2.39.2
