[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <9929d2390901140353i7158e43cv7991dc039b9b0a83@mail.gmail.com>
Date: Wed, 14 Jan 2009 03:53:01 -0800
From: "Jeff Kirsher" <jeffrey.t.kirsher@...el.com>
To: "Herbert Xu" <herbert@...dor.apana.org.au>
Cc: "David S. Miller" <davem@...emloft.net>, netdev@...r.kernel.org
Subject: Re: ixgbe: Replace LRO with GRO
On Wed, Jan 14, 2009 at 12:44 AM, Herbert Xu
<herbert@...dor.apana.org.au> wrote:
> Hi:
>
> Another day, another driver.
>
> This one's pretty simple, there is some netpoll bogosity around
> with the switch to netif_rx but at least it's harmless.
>
> ixgbe: Replace LRO with GRO
>
> This patch makes igb invoke the GRO hooks instead of LRO. As
> GRO has a compatible external interface to LRO this is a very
> straightforward replacement.
>
> As GRO uses the napi structure to track the held packets, I've
> modified the code paths involved to pass that along.
>
> Signed-off-by: Herbert Xu <herbert@...dor.apana.org.au>
>
> diff --git a/drivers/net/ixgbe/ixgbe.h b/drivers/net/ixgbe/ixgbe.h
> index e112008..6ac361a 100644
> --- a/drivers/net/ixgbe/ixgbe.h
> +++ b/drivers/net/ixgbe/ixgbe.h
> @@ -31,7 +31,6 @@
> #include <linux/types.h>
> #include <linux/pci.h>
> #include <linux/netdevice.h>
> -#include <linux/inet_lro.h>
> #include <linux/aer.h>
>
> #include "ixgbe_type.h"
> @@ -88,9 +87,6 @@
> #define IXGBE_TX_FLAGS_VLAN_PRIO_MASK 0x0000e000
> #define IXGBE_TX_FLAGS_VLAN_SHIFT 16
>
> -#define IXGBE_MAX_LRO_DESCRIPTORS 8
> -#define IXGBE_MAX_LRO_AGGREGATE 32
> -
> /* wrapper around a pointer to a socket buffer,
> * so a DMA handle can be stored along with the buffer */
> struct ixgbe_tx_buffer {
> @@ -142,8 +138,6 @@ struct ixgbe_ring {
> /* cpu for tx queue */
> int cpu;
> #endif
> - struct net_lro_mgr lro_mgr;
> - bool lro_used;
> struct ixgbe_queue_stats stats;
> u16 v_idx; /* maps directly to the index for this ring in the hardware
> * vector array, can also be used for finding the bit in EICR
> @@ -301,9 +295,6 @@ struct ixgbe_adapter {
>
> unsigned long state;
> u64 tx_busy;
> - u64 lro_aggregated;
> - u64 lro_flushed;
> - u64 lro_no_desc;
> unsigned int tx_ring_count;
> unsigned int rx_ring_count;
>
> diff --git a/drivers/net/ixgbe/ixgbe_ethtool.c b/drivers/net/ixgbe/ixgbe_ethtool.c
> index 67f87a7..4f6b5df 100644
> --- a/drivers/net/ixgbe/ixgbe_ethtool.c
> +++ b/drivers/net/ixgbe/ixgbe_ethtool.c
> @@ -89,8 +89,6 @@ static struct ixgbe_stats ixgbe_gstrings_stats[] = {
> {"rx_header_split", IXGBE_STAT(rx_hdr_split)},
> {"alloc_rx_page_failed", IXGBE_STAT(alloc_rx_page_failed)},
> {"alloc_rx_buff_failed", IXGBE_STAT(alloc_rx_buff_failed)},
> - {"lro_aggregated", IXGBE_STAT(lro_aggregated)},
> - {"lro_flushed", IXGBE_STAT(lro_flushed)},
> };
>
> #define IXGBE_QUEUE_STATS_LEN \
> @@ -808,15 +806,6 @@ static void ixgbe_get_ethtool_stats(struct net_device *netdev,
> int stat_count = sizeof(struct ixgbe_queue_stats) / sizeof(u64);
> int j, k;
> int i;
> - u64 aggregated = 0, flushed = 0, no_desc = 0;
> - for (i = 0; i < adapter->num_rx_queues; i++) {
> - aggregated += adapter->rx_ring[i].lro_mgr.stats.aggregated;
> - flushed += adapter->rx_ring[i].lro_mgr.stats.flushed;
> - no_desc += adapter->rx_ring[i].lro_mgr.stats.no_desc;
> - }
> - adapter->lro_aggregated = aggregated;
> - adapter->lro_flushed = flushed;
> - adapter->lro_no_desc = no_desc;
>
> ixgbe_update_stats(adapter);
> for (i = 0; i < IXGBE_GLOBAL_STATS_LEN; i++) {
> diff --git a/drivers/net/ixgbe/ixgbe_main.c b/drivers/net/ixgbe/ixgbe_main.c
> index acef3c6..23f25fe 100644
> --- a/drivers/net/ixgbe/ixgbe_main.c
> +++ b/drivers/net/ixgbe/ixgbe_main.c
> @@ -400,23 +400,20 @@ static int __ixgbe_notify_dca(struct device *dev, void *data)
> * @rx_ring: rx descriptor ring (for a specific queue) to setup
> * @rx_desc: rx descriptor
> **/
> -static void ixgbe_receive_skb(struct ixgbe_adapter *adapter,
> +static void ixgbe_receive_skb(struct ixgbe_q_vector *q_vector,
> struct sk_buff *skb, u8 status,
> - struct ixgbe_ring *ring,
> union ixgbe_adv_rx_desc *rx_desc)
> {
> + struct ixgbe_adapter *adapter = q_vector->adapter;
> + struct napi_struct *napi = &q_vector->napi;
> bool is_vlan = (status & IXGBE_RXD_STAT_VP);
> u16 tag = le16_to_cpu(rx_desc->wb.upper.vlan);
>
> - if (adapter->netdev->features & NETIF_F_LRO &&
> - skb->ip_summed == CHECKSUM_UNNECESSARY) {
> + if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
> if (adapter->vlgrp && is_vlan && (tag != 0))
> - lro_vlan_hwaccel_receive_skb(&ring->lro_mgr, skb,
> - adapter->vlgrp, tag,
> - rx_desc);
> + vlan_gro_receive(napi, adapter->vlgrp, tag, skb);
> else
> - lro_receive_skb(&ring->lro_mgr, skb, rx_desc);
> - ring->lro_used = true;
> + napi_gro_receive(napi, skb);
> } else {
> if (!(adapter->flags & IXGBE_FLAG_IN_NETPOLL)) {
> if (adapter->vlgrp && is_vlan && (tag != 0))
> @@ -571,10 +568,11 @@ static inline u16 ixgbe_get_pkt_info(union ixgbe_adv_rx_desc *rx_desc)
> return rx_desc->wb.lower.lo_dword.hs_rss.pkt_info;
> }
>
> -static bool ixgbe_clean_rx_irq(struct ixgbe_adapter *adapter,
> +static bool ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
> struct ixgbe_ring *rx_ring,
> int *work_done, int work_to_do)
> {
> + struct ixgbe_adapter *adapter = q_vector->adapter;
> struct pci_dev *pdev = adapter->pdev;
> union ixgbe_adv_rx_desc *rx_desc, *next_rxd;
> struct ixgbe_rx_buffer *rx_buffer_info, *next_buffer;
> @@ -675,7 +673,7 @@ static bool ixgbe_clean_rx_irq(struct ixgbe_adapter *adapter,
> total_rx_packets++;
>
> skb->protocol = eth_type_trans(skb, adapter->netdev);
> - ixgbe_receive_skb(adapter, skb, staterr, rx_ring, rx_desc);
> + ixgbe_receive_skb(q_vector, skb, staterr, rx_desc);
>
> next_desc:
> rx_desc->wb.upper.status_error = 0;
> @@ -693,11 +691,6 @@ next_desc:
> staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
> }
>
> - if (rx_ring->lro_used) {
> - lro_flush_all(&rx_ring->lro_mgr);
> - rx_ring->lro_used = false;
> - }
> -
> rx_ring->next_to_clean = i;
> cleaned_count = IXGBE_DESC_UNUSED(rx_ring);
>
> @@ -1049,7 +1042,7 @@ static int ixgbe_clean_rxonly(struct napi_struct *napi, int budget)
> ixgbe_update_rx_dca(adapter, rx_ring);
> #endif
>
> - ixgbe_clean_rx_irq(adapter, rx_ring, &work_done, budget);
> + ixgbe_clean_rx_irq(q_vector, rx_ring, &work_done, budget);
>
> /* If all Rx work done, exit the polling mode */
> if (work_done < budget) {
> @@ -1092,7 +1085,7 @@ static int ixgbe_clean_rxonly_many(struct napi_struct *napi, int budget)
> if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
> ixgbe_update_rx_dca(adapter, rx_ring);
> #endif
> - ixgbe_clean_rx_irq(adapter, rx_ring, &work_done, budget);
> + ixgbe_clean_rx_irq(q_vector, rx_ring, &work_done, budget);
> enable_mask |= rx_ring->v_idx;
> r_idx = find_next_bit(q_vector->rxr_idx, adapter->num_rx_queues,
> r_idx + 1);
> @@ -1565,33 +1558,6 @@ static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter, int index)
> IXGBE_WRITE_REG(&adapter->hw, IXGBE_SRRCTL(index), srrctl);
> }
>
> -/**
> - * ixgbe_get_skb_hdr - helper function for LRO header processing
> - * @skb: pointer to sk_buff to be added to LRO packet
> - * @iphdr: pointer to ip header structure
> - * @tcph: pointer to tcp header structure
> - * @hdr_flags: pointer to header flags
> - * @priv: private data
> - **/
> -static int ixgbe_get_skb_hdr(struct sk_buff *skb, void **iphdr, void **tcph,
> - u64 *hdr_flags, void *priv)
> -{
> - union ixgbe_adv_rx_desc *rx_desc = priv;
> -
> - /* Verify that this is a valid IPv4 TCP packet */
> - if (!((ixgbe_get_pkt_info(rx_desc) & IXGBE_RXDADV_PKTTYPE_IPV4) &&
> - (ixgbe_get_pkt_info(rx_desc) & IXGBE_RXDADV_PKTTYPE_TCP)))
> - return -1;
> -
> - /* Set network headers */
> - skb_reset_network_header(skb);
> - skb_set_transport_header(skb, ip_hdrlen(skb));
> - *iphdr = ip_hdr(skb);
> - *tcph = tcp_hdr(skb);
> - *hdr_flags = LRO_IPV4 | LRO_TCP;
> - return 0;
> -}
> -
> #define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
> (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
>
> @@ -1663,16 +1629,6 @@ static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
> adapter->rx_ring[i].head = IXGBE_RDH(j);
> adapter->rx_ring[i].tail = IXGBE_RDT(j);
> adapter->rx_ring[i].rx_buf_len = rx_buf_len;
> - /* Intitial LRO Settings */
> - adapter->rx_ring[i].lro_mgr.max_aggr = IXGBE_MAX_LRO_AGGREGATE;
> - adapter->rx_ring[i].lro_mgr.max_desc = IXGBE_MAX_LRO_DESCRIPTORS;
> - adapter->rx_ring[i].lro_mgr.get_skb_header = ixgbe_get_skb_hdr;
> - adapter->rx_ring[i].lro_mgr.features = LRO_F_EXTRACT_VLAN_ID;
> - if (!(adapter->flags & IXGBE_FLAG_IN_NETPOLL))
> - adapter->rx_ring[i].lro_mgr.features |= LRO_F_NAPI;
> - adapter->rx_ring[i].lro_mgr.dev = adapter->netdev;
> - adapter->rx_ring[i].lro_mgr.ip_summed = CHECKSUM_UNNECESSARY;
> - adapter->rx_ring[i].lro_mgr.ip_summed_aggr = CHECKSUM_UNNECESSARY;
>
> ixgbe_configure_srrctl(adapter, j);
> }
> @@ -2303,7 +2259,7 @@ static int ixgbe_poll(struct napi_struct *napi, int budget)
> #endif
>
> tx_cleaned = ixgbe_clean_tx_irq(adapter, adapter->tx_ring);
> - ixgbe_clean_rx_irq(adapter, adapter->rx_ring, &work_done, budget);
> + ixgbe_clean_rx_irq(q_vector, adapter->rx_ring, &work_done, budget);
>
> if (tx_cleaned)
> work_done = budget;
> @@ -2919,12 +2875,6 @@ int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
> struct pci_dev *pdev = adapter->pdev;
> int size;
>
> - size = sizeof(struct net_lro_desc) * IXGBE_MAX_LRO_DESCRIPTORS;
> - rx_ring->lro_mgr.lro_arr = vmalloc(size);
> - if (!rx_ring->lro_mgr.lro_arr)
> - return -ENOMEM;
> - memset(rx_ring->lro_mgr.lro_arr, 0, size);
> -
> size = sizeof(struct ixgbe_rx_buffer) * rx_ring->count;
> rx_ring->rx_buffer_info = vmalloc(size);
> if (!rx_ring->rx_buffer_info) {
> @@ -2953,8 +2903,6 @@ int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
> return 0;
>
> alloc_failed:
> - vfree(rx_ring->lro_mgr.lro_arr);
> - rx_ring->lro_mgr.lro_arr = NULL;
> return -ENOMEM;
> }
>
> @@ -3032,9 +2980,6 @@ void ixgbe_free_rx_resources(struct ixgbe_adapter *adapter,
> {
> struct pci_dev *pdev = adapter->pdev;
>
> - vfree(rx_ring->lro_mgr.lro_arr);
> - rx_ring->lro_mgr.lro_arr = NULL;
> -
> ixgbe_clean_rx_ring(adapter, rx_ring);
>
> vfree(rx_ring->rx_buffer_info);
> @@ -4136,7 +4081,7 @@ static int __devinit ixgbe_probe(struct pci_dev *pdev,
> netdev->features |= NETIF_F_IPV6_CSUM;
> netdev->features |= NETIF_F_TSO;
> netdev->features |= NETIF_F_TSO6;
> - netdev->features |= NETIF_F_LRO;
> + netdev->features |= NETIF_F_GRO;
>
> netdev->vlan_features |= NETIF_F_TSO;
> netdev->vlan_features |= NETIF_F_TSO6;
>
> Cheers,
> --
> Visit Openswan at http://www.openswan.org/
> Email: Herbert Xu ~{PmV>HI~} <herbert@...dor.apana.org.au>
> Home Page: http://gondor.apana.org.au/~herbert/
> PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
Thanks Herbert, we will get this in testing later today. Initially
the patch looks fine.
--
Cheers,
Jeff
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists