Commit 0cada332 authored by Vinay Kumar Yadav's avatar Vinay Kumar Yadav Committed by David S. Miller

net/tls: fix race condition causing kernel panic

tls_sw_recvmsg() and tls_decrypt_done() can be run concurrently.
// tls_sw_recvmsg()
	if (atomic_read(&ctx->decrypt_pending))
		crypto_wait_req(-EINPROGRESS, &ctx->async_wait);
	else
		reinit_completion(&ctx->async_wait.completion);

//tls_decrypt_done()
  	pending = atomic_dec_return(&ctx->decrypt_pending);

  	if (!pending && READ_ONCE(ctx->async_notify))
  		complete(&ctx->async_wait.completion);

Consider the scenario tls_decrypt_done() is about to run complete()

	if (!pending && READ_ONCE(ctx->async_notify))

and tls_sw_recvmsg() reads decrypt_pending == 0, does reinit_completion(),
then tls_decrypt_done() runs complete(). This sequence of execution
results in wrong completion. Consequently, for next decrypt request,
it will not wait for completion, eventually on connection close, crypto
resources freed, there is no way to handle pending decrypt response.

This race condition can be avoided by having atomic_read() mutually
exclusive with atomic_dec_return(),complete().Intoduced spin lock to
ensure the mutual exclution.

Addressed similar problem in tx direction.

v1->v2:
- More readable commit message.
- Corrected the lock to fix new race scenario.
- Removed barrier which is not needed now.

Fixes: a42055e8 ("net/tls: Add support for async encryption of records for performance")
Signed-off-by: default avatarVinay Kumar Yadav <vinay.yadav@chelsio.com>
Reviewed-by: default avatarJakub Kicinski <kuba@kernel.org>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent 98790bba
...@@ -135,6 +135,8 @@ struct tls_sw_context_tx { ...@@ -135,6 +135,8 @@ struct tls_sw_context_tx {
struct tls_rec *open_rec; struct tls_rec *open_rec;
struct list_head tx_list; struct list_head tx_list;
atomic_t encrypt_pending; atomic_t encrypt_pending;
/* protect crypto_wait with encrypt_pending */
spinlock_t encrypt_compl_lock;
int async_notify; int async_notify;
u8 async_capable:1; u8 async_capable:1;
...@@ -155,6 +157,8 @@ struct tls_sw_context_rx { ...@@ -155,6 +157,8 @@ struct tls_sw_context_rx {
u8 async_capable:1; u8 async_capable:1;
u8 decrypted:1; u8 decrypted:1;
atomic_t decrypt_pending; atomic_t decrypt_pending;
/* protect crypto_wait with decrypt_pending*/
spinlock_t decrypt_compl_lock;
bool async_notify; bool async_notify;
}; };
......
...@@ -206,10 +206,12 @@ static void tls_decrypt_done(struct crypto_async_request *req, int err) ...@@ -206,10 +206,12 @@ static void tls_decrypt_done(struct crypto_async_request *req, int err)
kfree(aead_req); kfree(aead_req);
spin_lock_bh(&ctx->decrypt_compl_lock);
pending = atomic_dec_return(&ctx->decrypt_pending); pending = atomic_dec_return(&ctx->decrypt_pending);
if (!pending && READ_ONCE(ctx->async_notify)) if (!pending && ctx->async_notify)
complete(&ctx->async_wait.completion); complete(&ctx->async_wait.completion);
spin_unlock_bh(&ctx->decrypt_compl_lock);
} }
static int tls_do_decryption(struct sock *sk, static int tls_do_decryption(struct sock *sk,
...@@ -467,10 +469,12 @@ static void tls_encrypt_done(struct crypto_async_request *req, int err) ...@@ -467,10 +469,12 @@ static void tls_encrypt_done(struct crypto_async_request *req, int err)
ready = true; ready = true;
} }
spin_lock_bh(&ctx->encrypt_compl_lock);
pending = atomic_dec_return(&ctx->encrypt_pending); pending = atomic_dec_return(&ctx->encrypt_pending);
if (!pending && READ_ONCE(ctx->async_notify)) if (!pending && ctx->async_notify)
complete(&ctx->async_wait.completion); complete(&ctx->async_wait.completion);
spin_unlock_bh(&ctx->encrypt_compl_lock);
if (!ready) if (!ready)
return; return;
...@@ -929,6 +933,7 @@ int tls_sw_sendmsg(struct sock *sk, struct msghdr *msg, size_t size) ...@@ -929,6 +933,7 @@ int tls_sw_sendmsg(struct sock *sk, struct msghdr *msg, size_t size)
int num_zc = 0; int num_zc = 0;
int orig_size; int orig_size;
int ret = 0; int ret = 0;
int pending;
if (msg->msg_flags & ~(MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL)) if (msg->msg_flags & ~(MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL))
return -EOPNOTSUPP; return -EOPNOTSUPP;
...@@ -1095,13 +1100,19 @@ int tls_sw_sendmsg(struct sock *sk, struct msghdr *msg, size_t size) ...@@ -1095,13 +1100,19 @@ int tls_sw_sendmsg(struct sock *sk, struct msghdr *msg, size_t size)
goto send_end; goto send_end;
} else if (num_zc) { } else if (num_zc) {
/* Wait for pending encryptions to get completed */ /* Wait for pending encryptions to get completed */
smp_store_mb(ctx->async_notify, true); spin_lock_bh(&ctx->encrypt_compl_lock);
ctx->async_notify = true;
if (atomic_read(&ctx->encrypt_pending)) pending = atomic_read(&ctx->encrypt_pending);
spin_unlock_bh(&ctx->encrypt_compl_lock);
if (pending)
crypto_wait_req(-EINPROGRESS, &ctx->async_wait); crypto_wait_req(-EINPROGRESS, &ctx->async_wait);
else else
reinit_completion(&ctx->async_wait.completion); reinit_completion(&ctx->async_wait.completion);
/* There can be no concurrent accesses, since we have no
* pending encrypt operations
*/
WRITE_ONCE(ctx->async_notify, false); WRITE_ONCE(ctx->async_notify, false);
if (ctx->async_wait.err) { if (ctx->async_wait.err) {
...@@ -1732,6 +1743,7 @@ int tls_sw_recvmsg(struct sock *sk, ...@@ -1732,6 +1743,7 @@ int tls_sw_recvmsg(struct sock *sk,
bool is_kvec = iov_iter_is_kvec(&msg->msg_iter); bool is_kvec = iov_iter_is_kvec(&msg->msg_iter);
bool is_peek = flags & MSG_PEEK; bool is_peek = flags & MSG_PEEK;
int num_async = 0; int num_async = 0;
int pending;
flags |= nonblock; flags |= nonblock;
...@@ -1894,8 +1906,11 @@ int tls_sw_recvmsg(struct sock *sk, ...@@ -1894,8 +1906,11 @@ int tls_sw_recvmsg(struct sock *sk,
recv_end: recv_end:
if (num_async) { if (num_async) {
/* Wait for all previously submitted records to be decrypted */ /* Wait for all previously submitted records to be decrypted */
smp_store_mb(ctx->async_notify, true); spin_lock_bh(&ctx->decrypt_compl_lock);
if (atomic_read(&ctx->decrypt_pending)) { ctx->async_notify = true;
pending = atomic_read(&ctx->decrypt_pending);
spin_unlock_bh(&ctx->decrypt_compl_lock);
if (pending) {
err = crypto_wait_req(-EINPROGRESS, &ctx->async_wait); err = crypto_wait_req(-EINPROGRESS, &ctx->async_wait);
if (err) { if (err) {
/* one of async decrypt failed */ /* one of async decrypt failed */
...@@ -1907,6 +1922,10 @@ int tls_sw_recvmsg(struct sock *sk, ...@@ -1907,6 +1922,10 @@ int tls_sw_recvmsg(struct sock *sk,
} else { } else {
reinit_completion(&ctx->async_wait.completion); reinit_completion(&ctx->async_wait.completion);
} }
/* There can be no concurrent accesses, since we have no
* pending decrypt operations
*/
WRITE_ONCE(ctx->async_notify, false); WRITE_ONCE(ctx->async_notify, false);
/* Drain records from the rx_list & copy if required */ /* Drain records from the rx_list & copy if required */
...@@ -2293,6 +2312,7 @@ int tls_set_sw_offload(struct sock *sk, struct tls_context *ctx, int tx) ...@@ -2293,6 +2312,7 @@ int tls_set_sw_offload(struct sock *sk, struct tls_context *ctx, int tx)
if (tx) { if (tx) {
crypto_init_wait(&sw_ctx_tx->async_wait); crypto_init_wait(&sw_ctx_tx->async_wait);
spin_lock_init(&sw_ctx_tx->encrypt_compl_lock);
crypto_info = &ctx->crypto_send.info; crypto_info = &ctx->crypto_send.info;
cctx = &ctx->tx; cctx = &ctx->tx;
aead = &sw_ctx_tx->aead_send; aead = &sw_ctx_tx->aead_send;
...@@ -2301,6 +2321,7 @@ int tls_set_sw_offload(struct sock *sk, struct tls_context *ctx, int tx) ...@@ -2301,6 +2321,7 @@ int tls_set_sw_offload(struct sock *sk, struct tls_context *ctx, int tx)
sw_ctx_tx->tx_work.sk = sk; sw_ctx_tx->tx_work.sk = sk;
} else { } else {
crypto_init_wait(&sw_ctx_rx->async_wait); crypto_init_wait(&sw_ctx_rx->async_wait);
spin_lock_init(&sw_ctx_rx->decrypt_compl_lock);
crypto_info = &ctx->crypto_recv.info; crypto_info = &ctx->crypto_recv.info;
cctx = &ctx->rx; cctx = &ctx->rx;
skb_queue_head_init(&sw_ctx_rx->rx_list); skb_queue_head_init(&sw_ctx_rx->rx_list);
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment