[dpdk-dev] [PATCH v2 2/3] ethdev: fail if Tx queue offload is not supported

[Andrew Rybchenko]

Do not allow to request unsupported Tx offload since all checks
are removed from PMDs because of consistency check in ethdev.
Otherwise application may rely on offload which is not actually
supported and send traffic with, for example, wrong checksums,
truncated packets or packets with garbage.

If application is using the new offload API, queue offloads only
may be added on Tx queue setup.

If application is using the old offload API, it knows nothing
about device level Tx offloads and requests everything required
on Tx queue setup using txq_flags. So, both device level (from PMD
point of view) and queue level offloads may be requested.
It is assumed that no PMD yet strictly separate device and queue
level offloads. If any PMD does it, it was broken for applications
which use the old offload API at the moment of PMD convertion anyway.

Fixes: 0330605295cf ("ethdev: new Rx/Tx offloads API")

Signed-off-by: Andrew Rybchenko <arybchenko at solarflare.com>
---
 lib/librte_ethdev/rte_ethdev.c | 51 +++++++++++++++++++++-------------
 1 file changed, 32 insertions(+), 19 deletions(-)

diff --git a/lib/librte_ethdev/rte_ethdev.c b/lib/librte_ethdev/rte_ethdev.c
index 54e1ee771..2b673013a 100644
--- a/lib/librte_ethdev/rte_ethdev.c
+++ b/lib/librte_ethdev/rte_ethdev.c
@@ -1727,25 +1727,38 @@ rte_eth_tx_queue_setup(uint16_t port_id, uint16_t tx_queue_id,
 	 */
 	local_conf.offloads &= ~dev->data->dev_conf.txmode.offloads;
 
-	/*
-	 * New added offloadings for this queue are those not enabled in
-	 * rte_eth_dev_configure( ) and they must be per-queue type.
-	 * A pure per-port offloading can't be enabled on a queue while
-	 * disabled on another queue. A pure per-port offloading can't
-	 * be enabled for any queue as new added one if it hasn't been
-	 * enabled in rte_eth_dev_configure( ).
-	 */
-	if ((local_conf.offloads & dev_info.tx_queue_offload_capa) !=
-	     local_conf.offloads) {
-		ethdev_log(ERR, "Ethdev port_id=%d tx_queue_id=%d, new "
-				"added offloads 0x%" PRIx64 " must be "
-				"within pre-queue offload capabilities 0x%"
-				PRIx64 " in %s( )\n",
-				port_id,
-				tx_queue_id,
-				local_conf.offloads,
-				dev_info.tx_queue_offload_capa,
-				__func__);
+	if (tx_conf->txq_flags & ETH_TXQ_FLAGS_IGNORE) {
+		/*
+		 * New added offloadings for this queue are those not enabled in
+		 * rte_eth_dev_configure( ) and they must be per-queue type.
+		 * A pure per-port offloading can't be enabled on a queue while
+		 * disabled on another queue. A pure per-port offloading can't
+		 * be enabled for any queue as new added one if it hasn't been
+		 * enabled in rte_eth_dev_configure( ).
+		 */
+		if ((local_conf.offloads & dev_info.tx_queue_offload_capa) !=
+		    local_conf.offloads) {
+			ethdev_log(ERR, "Ethdev port_id=%d tx_queue_id=%d, new "
+					"added offloads 0x%" PRIx64 " must be "
+					"within pre-queue offload capabilities 0x%"
+					PRIx64 " in %s( )\n",
+					port_id,
+					tx_queue_id,
+					local_conf.offloads,
+					dev_info.tx_queue_offload_capa,
+					__func__);
+		}
+	} else {
+		/*
+		 * Applications which are not converted yet to the new
+		 * Tx offload API may request device level offloads on
+		 * queue level (and nothing is requested on device level).
+		 * However, if the offload is not supported at all,
+		 * Tx queue setup must fail.
+		 */
+		if ((local_conf.offloads & dev_info.tx_offload_capa) !=
+		    local_conf.offloads)
+			return -EINVAL;
 	}
 
 	return eth_err(port_id, (*dev->dev_ops->tx_queue_setup)(dev,
-- 
2.17.0