tg3: Optimize rx double copy test

On a PCIX bus, the 5701 has a bug which requires the driver to double
copy all rx packets.  The rx code uses the rx_offset device member as a
flag to determine if this workaround should take effect.  The following
patch will modify the rx_offset member such that this test will become
less clear.

The patch starts by integrating the workaround check into the packet
length check.  It rounds out the implementation by relaxing the
workaround restrictions if the platform has efficient unaligned
accesses.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Reviewed-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/drivers/net/tg3.c b/drivers/net/tg3.c
index 3e893231fef32c2266ae7ec4c9b9982cbd9aef97..57914050b098310844677140cf50bf0743b97190 100644 (file)
--- a/drivers/net/tg3.c
+++ b/drivers/net/tg3.c
@@ -144,6 +144,24 @@
 
 #define TG3_RSS_MIN_NUM_MSIX_VECS      2
 
+/* Due to a hardware bug, the 5701 can only DMA to memory addresses
+ * that are at least dword aligned when used in PCIX mode.  The driver
+ * works around this bug by double copying the packet.  This workaround
+ * is built into the normal double copy length check for efficiency.
+ *
+ * However, the double copy is only necessary on those architectures
+ * where unaligned memory accesses are inefficient.  For those architectures
+ * where unaligned memory accesses incur little penalty, we can reintegrate
+ * the 5701 in the normal rx path.  Doing so saves a device structure
+ * dereference by hardcoding the double copy threshold in place.
+ */
+#define TG3_RX_COPY_THRESHOLD          256
+#if NET_IP_ALIGN == 0 || defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS)
+       #define TG3_RX_COPY_THRESH(tp)  TG3_RX_COPY_THRESHOLD
+#else
+       #define TG3_RX_COPY_THRESH(tp)  ((tp)->rx_copy_thresh)
+#endif
+
 /* minimum number of free TX descriptors required to wake up TX process */
 #define TG3_TX_WAKEUP_THRESH(tnapi)            ((tnapi)->tx_pending / 4)
 
@@ -4639,12 +4657,7 @@ static int tg3_rx(struct tg3_napi *tnapi, int budget)
                len = ((desc->idx_len & RXD_LEN_MASK) >> RXD_LEN_SHIFT) -
                      ETH_FCS_LEN;
 
-               if (len > RX_COPY_THRESHOLD &&
-                   tp->rx_offset == NET_IP_ALIGN) {
-                   /* rx_offset will likely not equal NET_IP_ALIGN
-                    * if this is a 5701 card running in PCI-X mode
-                    * [see tg3_get_invariants()]
-                    */
+               if (len > TG3_RX_COPY_THRESH(tp)) {
                        int skb_size;
 
                        skb_size = tg3_alloc_rx_skb(tp, tpr, opaque_key,
@@ -13469,9 +13482,14 @@ static int __devinit tg3_get_invariants(struct tg3 *tp)
                tp->tg3_flags &= ~TG3_FLAG_POLL_SERDES;
 
        tp->rx_offset = NET_IP_ALIGN;
+       tp->rx_copy_thresh = TG3_RX_COPY_THRESHOLD;
        if (GET_ASIC_REV(tp->pci_chip_rev_id) == ASIC_REV_5701 &&
-           (tp->tg3_flags & TG3_FLAG_PCIX_MODE) != 0)
+           (tp->tg3_flags & TG3_FLAG_PCIX_MODE) != 0) {
                tp->rx_offset = 0;
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+               tp->rx_copy_thresh = ~0;
+#endif
+       }
 
        tp->rx_std_max_post = TG3_RX_RING_SIZE;
 

diff --git a/drivers/net/tg3.h b/drivers/net/tg3.h
index 9e7fe0e7cdb865fda9ae6a6018b764390a738adb..43dd1d26243c838c957d2f0dd1b5bbf7fbf0473b 100644 (file)
--- a/drivers/net/tg3.h
+++ b/drivers/net/tg3.h
@@ -23,8 +23,6 @@
 #define TG3_BDINFO_NIC_ADDR            0xcUL /* 32-bit */
 #define TG3_BDINFO_SIZE                        0x10UL
 
-#define RX_COPY_THRESHOLD              256
-
 #define TG3_RX_INTERNAL_RING_SZ_5906   32
 
 #define RX_STD_MAX_SIZE                        1536
@@ -2754,9 +2752,11 @@ struct tg3 {
        struct tg3_napi                 napi[TG3_IRQ_MAX_VECS];
        void                            (*write32_rx_mbox) (struct tg3 *, u32,
                                                            u32);
+       u32                             rx_copy_thresh;
        u32                             rx_pending;
        u32                             rx_jumbo_pending;
        u32                             rx_std_max_post;
+       u32                             rx_offset;
        u32                             rx_pkt_map_sz;
 #if TG3_VLAN_TAG_USED
        struct vlan_group               *vlgrp;
@@ -2776,7 +2776,6 @@ struct tg3 {
        unsigned long                   last_event_jiffies;
        };
 
-       u32                             rx_offset;
        u32                             tg3_flags;
 #define TG3_FLAG_TAGGED_STATUS         0x00000001
 #define TG3_FLAG_TXD_MBOX_HWBUG                0x00000002