ixgbe: fix data access on big endian cpu
authorXuelin Shi <xuelin.shi@freescale.com>
Wed, 29 Jul 2015 06:38:20 +0000 (14:38 +0800)
committerThomas Monjalon <thomas.monjalon@6wind.com>
Thu, 30 Jul 2015 00:15:32 +0000 (02:15 +0200)
1. cpu use data owned by ixgbe must use rte_le_to_cpu_xx(...)
2. cpu fill data to ixgbe must use rte_cpu_to_le_xx(...)
3. checking pci status with converted constant

Signed-off-by: Xuelin Shi <xuelin.shi@freescale.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
drivers/net/ixgbe/ixgbe_rxtx.c

index a0c8847..a046bec 100644 (file)
@@ -130,7 +130,7 @@ ixgbe_tx_free_bufs(struct ixgbe_tx_queue *txq)
 
        /* check DD bit on threshold descriptor */
        status = txq->tx_ring[txq->tx_next_dd].wb.status;
-       if (! (status & IXGBE_ADVTXD_STAT_DD))
+       if (!(status & rte_cpu_to_le_32(IXGBE_ADVTXD_STAT_DD)))
                return 0;
 
        /*
@@ -175,11 +175,14 @@ tx4(volatile union ixgbe_adv_tx_desc *txdp, struct rte_mbuf **pkts)
                pkt_len = (*pkts)->data_len;
 
                /* write data to descriptor */
-               txdp->read.buffer_addr = buf_dma_addr;
+               txdp->read.buffer_addr = rte_cpu_to_le_64(buf_dma_addr);
+
                txdp->read.cmd_type_len =
-                               ((uint32_t)DCMD_DTYP_FLAGS | pkt_len);
+                       rte_cpu_to_le_32((uint32_t)DCMD_DTYP_FLAGS | pkt_len);
+
                txdp->read.olinfo_status =
-                               (pkt_len << IXGBE_ADVTXD_PAYLEN_SHIFT);
+                       rte_cpu_to_le_32(pkt_len << IXGBE_ADVTXD_PAYLEN_SHIFT);
+
                rte_prefetch0(&(*pkts)->pool);
        }
 }
@@ -195,11 +198,11 @@ tx1(volatile union ixgbe_adv_tx_desc *txdp, struct rte_mbuf **pkts)
        pkt_len = (*pkts)->data_len;
 
        /* write data to descriptor */
-       txdp->read.buffer_addr = buf_dma_addr;
+       txdp->read.buffer_addr = rte_cpu_to_le_64(buf_dma_addr);
        txdp->read.cmd_type_len =
-                       ((uint32_t)DCMD_DTYP_FLAGS | pkt_len);
+                       rte_cpu_to_le_32((uint32_t)DCMD_DTYP_FLAGS | pkt_len);
        txdp->read.olinfo_status =
-                       (pkt_len << IXGBE_ADVTXD_PAYLEN_SHIFT);
+                       rte_cpu_to_le_32(pkt_len << IXGBE_ADVTXD_PAYLEN_SHIFT);
        rte_prefetch0(&(*pkts)->pool);
 }
 
@@ -511,6 +514,7 @@ ixgbe_xmit_cleanup(struct ixgbe_tx_queue *txq)
        uint16_t nb_tx_desc = txq->nb_tx_desc;
        uint16_t desc_to_clean_to;
        uint16_t nb_tx_to_clean;
+       uint32_t status;
 
        /* Determine the last descriptor needing to be cleaned */
        desc_to_clean_to = (uint16_t)(last_desc_cleaned + txq->tx_rs_thresh);
@@ -519,7 +523,8 @@ ixgbe_xmit_cleanup(struct ixgbe_tx_queue *txq)
 
        /* Check to make sure the last descriptor to clean is done */
        desc_to_clean_to = sw_ring[desc_to_clean_to].last_id;
-       if (! (txr[desc_to_clean_to].wb.status & IXGBE_TXD_STAT_DD))
+       status = txr[desc_to_clean_to].wb.status;
+       if (!(status & rte_cpu_to_le_32(IXGBE_TXD_STAT_DD)))
        {
                PMD_TX_FREE_LOG(DEBUG,
                                "TX descriptor %4u is not done"
@@ -1061,14 +1066,15 @@ ixgbe_rx_scan_hw_ring(struct ixgbe_rx_queue *rxq)
        int s[LOOK_AHEAD], nb_dd;
 #endif /* RTE_NEXT_ABI */
        int i, j, nb_rx = 0;
-
+       uint32_t status;
 
        /* get references to current descriptor and S/W ring entry */
        rxdp = &rxq->rx_ring[rxq->rx_tail];
        rxep = &rxq->sw_ring[rxq->rx_tail];
 
+       status = rxdp->wb.upper.status_error;
        /* check to make sure there is at least 1 packet to receive */
-       if (! (rxdp->wb.upper.status_error & IXGBE_RXDADV_STAT_DD))
+       if (!(status & rte_cpu_to_le_32(IXGBE_RXDADV_STAT_DD)))
                return 0;
 
        /*
@@ -1080,7 +1086,7 @@ ixgbe_rx_scan_hw_ring(struct ixgbe_rx_queue *rxq)
        {
                /* Read desc statuses backwards to avoid race condition */
                for (j = LOOK_AHEAD-1; j >= 0; --j)
-                       s[j] = rxdp[j].wb.upper.status_error;
+                       s[j] = rte_le_to_cpu_32(rxdp[j].wb.upper.status_error);
 
 #ifdef RTE_NEXT_ABI
                for (j = LOOK_AHEAD - 1; j >= 0; --j)
@@ -1098,7 +1104,8 @@ ixgbe_rx_scan_hw_ring(struct ixgbe_rx_queue *rxq)
                /* Translate descriptor info to mbuf format */
                for (j = 0; j < nb_dd; ++j) {
                        mb = rxep[j].mbuf;
-                       pkt_len = (uint16_t)(rxdp[j].wb.upper.length - rxq->crc_len);
+                       pkt_len = rte_le_to_cpu_16(rxdp[j].wb.upper.length) -
+                                 rxq->crc_len;
                        mb->data_len = pkt_len;
                        mb->pkt_len = pkt_len;
                        mb->vlan_tci = rte_le_to_cpu_16(rxdp[j].wb.upper.vlan);
@@ -1114,7 +1121,8 @@ ixgbe_rx_scan_hw_ring(struct ixgbe_rx_queue *rxq)
                                ixgbe_rxd_pkt_info_to_pkt_type(pkt_info[j]);
 #else /* RTE_NEXT_ABI */
                        pkt_flags  = rx_desc_hlen_type_rss_to_pkt_flags(
-                                       rxdp[j].wb.lower.lo_dword.data);
+                                       rte_le_to_cpu_32(
+                                       rxdp[j].wb.lower.lo_dword.data));
                        /* reuse status field from scan list */
                        pkt_flags |= rx_desc_status_to_pkt_flags(s[j]);
                        pkt_flags |= rx_desc_error_to_pkt_flags(s[j]);
@@ -1122,12 +1130,14 @@ ixgbe_rx_scan_hw_ring(struct ixgbe_rx_queue *rxq)
 #endif /* RTE_NEXT_ABI */
 
                        if (likely(pkt_flags & PKT_RX_RSS_HASH))
-                               mb->hash.rss = rxdp[j].wb.lower.hi_dword.rss;
+                               mb->hash.rss = rte_le_to_cpu_32(
+                                   rxdp[j].wb.lower.hi_dword.rss);
                        else if (pkt_flags & PKT_RX_FDIR) {
-                               mb->hash.fdir.hash =
-                                       (uint16_t)((rxdp[j].wb.lower.hi_dword.csum_ip.csum)
-                                               & IXGBE_ATR_HASH_MASK);
-                               mb->hash.fdir.id = rxdp[j].wb.lower.hi_dword.csum_ip.ip_id;
+                               mb->hash.fdir.hash = rte_le_to_cpu_16(
+                                   rxdp[j].wb.lower.hi_dword.csum_ip.csum) &
+                                   IXGBE_ATR_HASH_MASK;
+                               mb->hash.fdir.id = rte_le_to_cpu_16(
+                                   rxdp[j].wb.lower.hi_dword.csum_ip.ip_id);
                        }
                }
 
@@ -1346,7 +1356,7 @@ ixgbe_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
                 */
                rxdp = &rx_ring[rx_id];
                staterr = rxdp->wb.upper.status_error;
-               if (! (staterr & rte_cpu_to_le_32(IXGBE_RXDADV_STAT_DD)))
+               if (!(staterr & rte_cpu_to_le_32(IXGBE_RXDADV_STAT_DD)))
                        break;
                rxd = *rxdp;
 
@@ -1464,12 +1474,14 @@ ixgbe_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
 #endif /* RTE_NEXT_ABI */
 
                if (likely(pkt_flags & PKT_RX_RSS_HASH))
-                       rxm->hash.rss = rxd.wb.lower.hi_dword.rss;
+                       rxm->hash.rss = rte_le_to_cpu_32(
+                                               rxd.wb.lower.hi_dword.rss);
                else if (pkt_flags & PKT_RX_FDIR) {
-                       rxm->hash.fdir.hash =
-                               (uint16_t)((rxd.wb.lower.hi_dword.csum_ip.csum)
-                                          & IXGBE_ATR_HASH_MASK);
-                       rxm->hash.fdir.id = rxd.wb.lower.hi_dword.csum_ip.ip_id;
+                       rxm->hash.fdir.hash = rte_le_to_cpu_16(
+                                       rxd.wb.lower.hi_dword.csum_ip.csum) &
+                                       IXGBE_ATR_HASH_MASK;
+                       rxm->hash.fdir.id = rte_le_to_cpu_16(
+                                       rxd.wb.lower.hi_dword.csum_ip.ip_id);
                }
                /*
                 * Store the mbuf address into the next entry of the array
@@ -1980,7 +1992,7 @@ ixgbe_reset_tx_queue(struct ixgbe_tx_queue *txq)
        prev = (uint16_t) (txq->nb_tx_desc - 1);
        for (i = 0; i < txq->nb_tx_desc; i++) {
                volatile union ixgbe_adv_tx_desc *txd = &txq->tx_ring[i];
-               txd->wb.status = IXGBE_TXD_STAT_DD;
+               txd->wb.status = rte_cpu_to_le_32(IXGBE_TXD_STAT_DD);
                txe[i].mbuf = NULL;
                txe[i].last_id = i;
                txe[prev].next_id = i;
@@ -2586,7 +2598,8 @@ ixgbe_dev_rx_queue_count(struct rte_eth_dev *dev, uint16_t rx_queue_id)
        rxdp = &(rxq->rx_ring[rxq->rx_tail]);
 
        while ((desc < rxq->nb_rx_desc) &&
-               (rxdp->wb.upper.status_error & IXGBE_RXDADV_STAT_DD)) {
+               (rxdp->wb.upper.status_error &
+                       rte_cpu_to_le_32(IXGBE_RXDADV_STAT_DD))) {
                desc += IXGBE_RXQ_SCAN_INTERVAL;
                rxdp += IXGBE_RXQ_SCAN_INTERVAL;
                if (rxq->rx_tail + desc >= rxq->nb_rx_desc)
@@ -2611,7 +2624,8 @@ ixgbe_dev_rx_descriptor_done(void *rx_queue, uint16_t offset)
                desc -= rxq->nb_rx_desc;
 
        rxdp = &rxq->rx_ring[desc];
-       return !!(rxdp->wb.upper.status_error & IXGBE_RXDADV_STAT_DD);
+       return !!(rxdp->wb.upper.status_error &
+                       rte_cpu_to_le_32(IXGBE_RXDADV_STAT_DD));
 }
 
 void __attribute__((cold))