net/tls: Add asynchronous resync

This patch adds support for asynchronous resynchronization in tls_device.
Async resync follows two distinct stages:

1. The NIC driver indicates that it would like to resync on some TLS
record within the received packet (P), but the driver does not
know (yet) which of the TLS records within the packet.
At this stage, the NIC driver will query the device to find the exact
TCP sequence for resync (tcpsn), however, the driver does not wait
for the device to provide the response.

2. Eventually, the device responds, and the driver provides the tcpsn
within the resync packet to KTLS. Now, KTLS can check the tcpsn against
any processed TLS records within packet P, and also against any record
that is processed in the future within packet P.

The asynchronous resync path simplifies the device driver, as it can
save bits on the packet completion (32-bit TCP sequence), and pass this
information on an asynchronous command instead.

Signed-off-by: Boris Pismenny <[email protected]>
Signed-off-by: Tariq Toukan <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>

Author: Boris Pismenny

include/net/tls.h

@@ -291,11 +291,19 @@ struct tlsdev_ops {
 enum tls_offload_sync_type {
 	TLS_OFFLOAD_SYNC_TYPE_DRIVER_REQ = 0,
 	TLS_OFFLOAD_SYNC_TYPE_CORE_NEXT_HINT = 1,
+	TLS_OFFLOAD_SYNC_TYPE_DRIVER_REQ_ASYNC = 2,
 };
 
 #define TLS_DEVICE_RESYNC_NH_START_IVAL		2
 #define TLS_DEVICE_RESYNC_NH_MAX_IVAL		128
 
+#define TLS_DEVICE_RESYNC_ASYNC_LOGMAX		13
+struct tls_offload_resync_async {
+	atomic64_t req;
+	u32 loglen;
+	u32 log[TLS_DEVICE_RESYNC_ASYNC_LOGMAX];
+};
+
 struct tls_offload_context_rx {
 	/* sw must be the first member of tls_offload_context_rx */
 	struct tls_sw_context_rx sw;
@@ -314,6 +322,10 @@ struct tls_offload_context_rx {
 			u32 decrypted_failed;
 			u32 decrypted_tgt;
 		} resync_nh;
+		/* TLS_OFFLOAD_SYNC_TYPE_DRIVER_REQ_ASYNC */
+		struct {
+			struct tls_offload_resync_async *resync_async;
+		};
 	};
 	u8 driver_state[] __aligned(8);
 	/* The TLS layer reserves room for driver specific state
@@ -606,13 +618,37 @@ tls_driver_ctx(const struct sock *sk, enum tls_offload_ctx_dir direction)
 }
 #endif
 
+#define RESYNC_REQ BIT(0)
+#define RESYNC_REQ_ASYNC BIT(1)
 /* The TLS context is valid until sk_destruct is called */
 static inline void tls_offload_rx_resync_request(struct sock *sk, __be32 seq)
 {
 	struct tls_context *tls_ctx = tls_get_ctx(sk);
 	struct tls_offload_context_rx *rx_ctx = tls_offload_ctx_rx(tls_ctx);
 
-	atomic64_set(&rx_ctx->resync_req, ((u64)ntohl(seq) << 32) | 1);
+	atomic64_set(&rx_ctx->resync_req, ((u64)ntohl(seq) << 32) | RESYNC_REQ);
+}
+
+/* Log all TLS record header TCP sequences in [seq, seq+len] */
+static inline void
+tls_offload_rx_resync_async_request_start(struct sock *sk, __be32 seq, u16 len)
+{
+	struct tls_context *tls_ctx = tls_get_ctx(sk);
+	struct tls_offload_context_rx *rx_ctx = tls_offload_ctx_rx(tls_ctx);
+
+	atomic64_set(&rx_ctx->resync_async->req, ((u64)ntohl(seq) << 32) |
+		     (len << 16) | RESYNC_REQ | RESYNC_REQ_ASYNC);
+	rx_ctx->resync_async->loglen = 0;
+}
+
+static inline void
+tls_offload_rx_resync_async_request_end(struct sock *sk, __be32 seq)
+{
+	struct tls_context *tls_ctx = tls_get_ctx(sk);
+	struct tls_offload_context_rx *rx_ctx = tls_offload_ctx_rx(tls_ctx);
+
+	atomic64_set(&rx_ctx->resync_async->req,
+		     ((u64)ntohl(seq) << 32) | RESYNC_REQ);
 }
 
 static inline void

net/tls/tls_device.c

@@ -690,6 +690,47 @@ static void tls_device_resync_rx(struct tls_context *tls_ctx,
 	TLS_INC_STATS(sock_net(sk), LINUX_MIB_TLSRXDEVICERESYNC);
 }
 
+static bool
+tls_device_rx_resync_async(struct tls_offload_resync_async *resync_async,
+			   s64 resync_req, u32 *seq)
+{
+	u32 is_async = resync_req & RESYNC_REQ_ASYNC;
+	u32 req_seq = resync_req >> 32;
+	u32 req_end = req_seq + ((resync_req >> 16) & 0xffff);
+
+	if (is_async) {
+		/* asynchronous stage: log all headers seq such that
+		 * req_seq <= seq <= end_seq, and wait for real resync request
+		 */
+		if (between(*seq, req_seq, req_end) &&
+		    resync_async->loglen < TLS_DEVICE_RESYNC_ASYNC_LOGMAX)
+			resync_async->log[resync_async->loglen++] = *seq;
+
+		return false;
+	}
+
+	/* synchronous stage: check against the logged entries and
+	 * proceed to check the next entries if no match was found
+	 */
+	while (resync_async->loglen) {
+		if (req_seq == resync_async->log[resync_async->loglen - 1] &&
+		    atomic64_try_cmpxchg(&resync_async->req,
+					 &resync_req, 0)) {
+			resync_async->loglen = 0;
+			*seq = req_seq;
+			return true;
+		}
+		resync_async->loglen--;
+	}
+
+	if (req_seq == *seq &&
+	    atomic64_try_cmpxchg(&resync_async->req,
+				 &resync_req, 0))
+		return true;
+
+	return false;
+}
+
 void tls_device_rx_resync_new_rec(struct sock *sk, u32 rcd_len, u32 seq)
 {
 	struct tls_context *tls_ctx = tls_get_ctx(sk);
@@ -736,6 +777,16 @@ void tls_device_rx_resync_new_rec(struct sock *sk, u32 rcd_len, u32 seq)
 		seq += rcd_len;
 		tls_bigint_increment(rcd_sn, prot->rec_seq_size);
 		break;
+	case TLS_OFFLOAD_SYNC_TYPE_DRIVER_REQ_ASYNC:
+		resync_req = atomic64_read(&rx_ctx->resync_async->req);
+		is_req_pending = resync_req;
+		if (likely(!is_req_pending))
+			return;
+
+		if (!tls_device_rx_resync_async(rx_ctx->resync_async,
+						resync_req, &seq))
+			return;
+		break;
 	}
 
 	tls_device_resync_rx(tls_ctx, sk, seq, rcd_sn);