net/bnxt: update ring group after ring stop start
authorAjit Khaparde <ajit.khaparde@broadcom.com>
Mon, 2 Aug 2021 04:44:30 +0000 (21:44 -0700)
committerAjit Khaparde <ajit.khaparde@broadcom.com>
Thu, 19 Aug 2021 04:39:23 +0000 (06:39 +0200)
A Rx ring stop start sequence may result in the FW returning
a different set of Rx ring and AGG ring IDs. If the ring group
is not updated with the new IDs, the HW sees the host driver using
incorrect BD types for the Rx ring and AGG ring. This can cause
the chip to go into a bad state or encounter RE_flush issue
or leak mbufs in the HW.

Fix this by issuing a bnxt_hwrm_ring_grp_free() and an
bnxt_hwrm_ring_grp_alloc() to refresh the ring group information.

Fixes: 9b63c6fd70e3 ("net/bnxt: support Rx/Tx queue start/stop")
Cc: stable@dpdk.org
Signed-off-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
drivers/net/bnxt/bnxt_hwrm.c
drivers/net/bnxt/bnxt_ring.c

index 73deb41..eb9de45 100644 (file)
@@ -2717,6 +2717,9 @@ void bnxt_free_hwrm_rx_ring(struct bnxt *bp, int queue_index)
        struct bnxt_ring *ring = rxr->rx_ring_struct;
        struct bnxt_cp_ring_info *cpr = rxq->cp_ring;
 
+       if (BNXT_HAS_RING_GRPS(bp))
+               bnxt_hwrm_ring_grp_free(bp, queue_index);
+
        bnxt_hwrm_ring_free(bp, ring,
                            HWRM_RING_FREE_INPUT_RING_TYPE_RX,
                            cpr->cp_ring_struct->fw_ring_id);
index b05c470..957b175 100644 (file)
@@ -631,6 +631,12 @@ int bnxt_alloc_hwrm_rx_ring(struct bnxt *bp, int queue_index)
        if (rc)
                goto err_out;
 
+       if (BNXT_HAS_RING_GRPS(bp)) {
+               rc = bnxt_hwrm_ring_grp_alloc(bp, queue_index);
+               if (rc)
+                       goto err_out;
+       }
+
        if (rxq->rx_started) {
                if (bnxt_init_one_rx_ring(rxq)) {
                        PMD_DRV_LOG(ERR, "bnxt_init_one_rx_ring failed!\n");