int mlx4_ib_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc)
{
struct mlx4_ib_cq *cq = to_mcq(ibcq);
struct mlx4_ib_qp *cur_qp = NULL;
unsigned long flags;
int npolled;
int err = 0;
spin_lock_irqsave(&cq->lock, flags);
for (npolled = 0; npolled < num_entries; ++npolled) {
err = mlx4_ib_poll_one(cq, &cur_qp, wc + npolled);
if (err)
break;
}
if (npolled)
mlx4_cq_set_ci(&cq->mcq);
spin_unlock_irqrestore(&cq->lock, flags);
if (err == 0 || err == -EAGAIN)
return npolled;
else
return err;
}
void __mlx4_ib_cq_clean(struct mlx4_ib_cq *cq, u32 qpn, struct mlx4_ib_srq *srq)
{
u32 prod_index;
int nfreed = 0;
struct mlx4_cqe *cqe, *dest;
u8 owner_bit;
/*
* First we need to find the current producer index, so we
* know where to start cleaning from. It doesn't matter if HW
* adds new entries after this loop -- the QP we're worried
* about is already in RESET, so the new entries won't come
* from our QP and therefore don't need to be checked.
*/
for (prod_index = cq->mcq.cons_index; get_sw_cqe(cq, prod_index); ++prod_index)
if (prod_index == cq->mcq.cons_index + cq->ibcq.cqe)
break;
/*
* Now sweep backwards through the CQ, removing CQ entries
* that match our QP by copying older entries on top of them.
*/
while ((int) --prod_index - (int) cq->mcq.cons_index >= 0) {
cqe = get_cqe(cq, prod_index & cq->ibcq.cqe);
if ((be32_to_cpu(cqe->vlan_my_qpn) & MLX4_CQE_QPN_MASK) == qpn) {
if (srq && !(cqe->owner_sr_opcode & MLX4_CQE_IS_SEND_MASK))
mlx4_ib_free_srq_wqe(srq, be16_to_cpu(cqe->wqe_index));
++nfreed;
} else if (nfreed) {
dest = get_cqe(cq, (prod_index + nfreed) & cq->ibcq.cqe);
owner_bit = dest->owner_sr_opcode & MLX4_CQE_OWNER_MASK;
memcpy(dest, cqe, sizeof *cqe);
dest->owner_sr_opcode = owner_bit |
(dest->owner_sr_opcode & ~MLX4_CQE_OWNER_MASK);
}
}
if (nfreed) {
cq->mcq.cons_index += nfreed;
/*
* Make sure update of buffer contents is done before
* updating consumer index.
*/
wmb();
mlx4_cq_set_ci(&cq->mcq);
}
}
static int mlx4_en_process_tx_cq(struct net_device *dev,
struct mlx4_en_cq *cq)
{
struct mlx4_en_priv *priv = netdev_priv(dev);
struct mlx4_cq *mcq = &cq->mcq;
struct mlx4_en_tx_ring *ring = priv->tx_ring[cq->ring];
struct mlx4_cqe *cqe;
u16 index;
u16 new_index, ring_index, stamp_index;
u32 txbbs_skipped = 0;
#ifndef CONFIG_WQE_FORMAT_1
u32 txbbs_stamp = 0;
#endif
u32 cons_index = mcq->cons_index;
int size = cq->size;
u32 size_mask = ring->size_mask;
struct mlx4_cqe *buf = cq->buf;
u32 packets = 0;
u32 bytes = 0;
int factor = priv->cqe_factor;
u64 timestamp = 0;
int done = 0;
if (!priv->port_up)
return 0;
index = cons_index & size_mask;
cqe = &buf[(index << factor) + factor];
ring_index = ring->cons & size_mask;
stamp_index = ring_index;
/* Process all completed CQEs */
while (XNOR(cqe->owner_sr_opcode & MLX4_CQE_OWNER_MASK,
cons_index & size)) {
/*
* make sure we read the CQE after we read the
* ownership bit
*/
rmb();
if (unlikely((cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) ==
MLX4_CQE_OPCODE_ERROR)) {
en_err(priv, "CQE completed in error - vendor syndrom: 0x%x syndrom: 0x%x\n",
((struct mlx4_err_cqe *)cqe)->
vendor_err_syndrome,
((struct mlx4_err_cqe *)cqe)->syndrome);
}
/* Skip over last polled CQE */
new_index = be16_to_cpu(cqe->wqe_index) & size_mask;
do {
txbbs_skipped += ring->last_nr_txbb;
ring_index = (ring_index + ring->last_nr_txbb) & size_mask;
/* free next descriptor */
ring->last_nr_txbb = mlx4_en_free_tx_desc(
priv, ring, ring_index,
!!((ring->cons + txbbs_skipped) &
ring->size), timestamp);
#ifndef CONFIG_WQE_FORMAT_1
mlx4_en_stamp_wqe(priv, ring, stamp_index,
!!((ring->cons + txbbs_stamp) &
ring->size));
stamp_index = ring_index;
txbbs_stamp = txbbs_skipped;
#endif
packets++;
bytes += ring->tx_info[ring_index].nr_bytes;
} while (ring_index != new_index);
++cons_index;
index = cons_index & size_mask;
cqe = &buf[(index << factor) + factor];
}
/*
* To prevent CQ overflow we first update CQ consumer and only then
* the ring consumer.
*/
mcq->cons_index = cons_index;
mlx4_cq_set_ci(mcq);
wmb();
ring->cons += txbbs_skipped;
/* Wakeup Tx queue if it was stopped and ring is not full */
if (unlikely(ring->blocked) &&
(ring->prod - ring->cons) <= ring->full_size) {
ring->blocked = 0;
#ifdef CONFIG_RATELIMIT
if (cq->ring < priv->native_tx_ring_num) {
if (atomic_fetchadd_int(&priv->blocked, -1) == 1)
atomic_clear_int(&dev->if_drv_flags ,IFF_DRV_OACTIVE);
priv->port_stats.wake_queue++;
}
#else
if (atomic_fetchadd_int(&priv->blocked, -1) == 1)
atomic_clear_int(&dev->if_drv_flags ,IFF_DRV_OACTIVE);
priv->port_stats.wake_queue++;
#endif
ring->wake_queue++;
}
return done;
}
int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int budget)
{
struct mlx4_en_priv *priv = netdev_priv(dev);
struct mlx4_cqe *cqe;
struct mlx4_en_rx_ring *ring = &priv->rx_ring[cq->ring];
struct skb_frag_struct *skb_frags;
struct mlx4_en_rx_desc *rx_desc;
struct sk_buff *skb;
int index;
int nr;
unsigned int length;
int polled = 0;
int ip_summed;
if (!priv->port_up)
return 0;
/* We assume a 1:1 mapping between CQEs and Rx descriptors, so Rx
* descriptor offset can be deduced from the CQE index instead of
* reading 'cqe->index' */
index = cq->mcq.cons_index & ring->size_mask;
cqe = &cq->buf[index];
/* Process all completed CQEs */
while (XNOR(cqe->owner_sr_opcode & MLX4_CQE_OWNER_MASK,
cq->mcq.cons_index & cq->size)) {
skb_frags = ring->rx_info + (index << priv->log_rx_info);
rx_desc = ring->buf + (index << ring->log_stride);
/*
* make sure we read the CQE after we read the ownership bit
*/
rmb();
/* Drop packet on bad receive or bad checksum */
if (unlikely((cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) ==
MLX4_CQE_OPCODE_ERROR)) {
en_err(priv, "CQE completed in error - vendor "
"syndrom:%d syndrom:%d\n",
((struct mlx4_err_cqe *) cqe)->vendor_err_syndrome,
((struct mlx4_err_cqe *) cqe)->syndrome);
goto next;
}
if (unlikely(cqe->badfcs_enc & MLX4_CQE_BAD_FCS)) {
en_dbg(RX_ERR, priv, "Accepted frame with bad FCS\n");
goto next;
}
/*
* Packet is OK - process it.
*/
length = be32_to_cpu(cqe->byte_cnt);
ring->bytes += length;
ring->packets++;
if (likely(dev->features & NETIF_F_RXCSUM)) {
if ((cqe->status & cpu_to_be16(MLX4_CQE_STATUS_IPOK)) &&
(cqe->checksum == cpu_to_be16(0xffff))) {
priv->port_stats.rx_chksum_good++;
/* This packet is eligible for LRO if it is:
* - DIX Ethernet (type interpretation)
* - TCP/IP (v4)
* - without IP options
* - not an IP fragment */
if (dev->features & NETIF_F_GRO) {
struct sk_buff *gro_skb = napi_get_frags(&cq->napi);
if (!gro_skb)
goto next;
nr = mlx4_en_complete_rx_desc(
priv, rx_desc,
skb_frags, skb_shinfo(gro_skb)->frags,
ring->page_alloc, length);
if (!nr)
goto next;
skb_shinfo(gro_skb)->nr_frags = nr;
gro_skb->len = length;
gro_skb->data_len = length;
gro_skb->truesize += length;
gro_skb->ip_summed = CHECKSUM_UNNECESSARY;
if (priv->vlgrp && (cqe->vlan_my_qpn &
cpu_to_be32(MLX4_CQE_VLAN_PRESENT_MASK)))
vlan_gro_frags(&cq->napi, priv->vlgrp, be16_to_cpu(cqe->sl_vid));
else
napi_gro_frags(&cq->napi);
goto next;
}
/* LRO not possible, complete processing here */
ip_summed = CHECKSUM_UNNECESSARY;
} else {
ip_summed = CHECKSUM_NONE;
priv->port_stats.rx_chksum_none++;
}
} else {
ip_summed = CHECKSUM_NONE;
priv->port_stats.rx_chksum_none++;
}
skb = mlx4_en_rx_skb(priv, rx_desc, skb_frags,
ring->page_alloc, length);
if (!skb) {
priv->stats.rx_dropped++;
goto next;
}
if (unlikely(priv->validate_loopback)) {
validate_loopback(priv, skb);
goto next;
}
skb->ip_summed = ip_summed;
skb->protocol = eth_type_trans(skb, dev);
skb_record_rx_queue(skb, cq->ring);
/* Push it up the stack */
if (priv->vlgrp && (be32_to_cpu(cqe->vlan_my_qpn) &
MLX4_CQE_VLAN_PRESENT_MASK)) {
vlan_hwaccel_receive_skb(skb, priv->vlgrp,
be16_to_cpu(cqe->sl_vid));
} else
netif_receive_skb(skb);
next:
++cq->mcq.cons_index;
index = (cq->mcq.cons_index) & ring->size_mask;
cqe = &cq->buf[index];
if (++polled == budget) {
/* We are here because we reached the NAPI budget -
* flush only pending LRO sessions */
goto out;
}
}
out:
AVG_PERF_COUNTER(priv->pstats.rx_coal_avg, polled);
mlx4_cq_set_ci(&cq->mcq);
wmb(); /* ensure HW sees CQ consumer before we post new buffers */
ring->cons = cq->mcq.cons_index;
ring->prod += polled; /* Polled descriptors were realocated in place */
mlx4_en_update_rx_prod_db(ring);
return polled;
}
static void mlx4_en_process_tx_cq(struct net_device *dev, struct mlx4_en_cq *cq)
{
struct mlx4_en_priv *priv = netdev_priv(dev);
struct mlx4_cq *mcq = &cq->mcq;
struct mlx4_en_tx_ring *ring = &priv->tx_ring[cq->ring];
struct mlx4_cqe *cqe = cq->buf;
u16 index;
u16 new_index;
u32 txbbs_skipped = 0;
u32 cq_last_sav;
/* index always points to the first TXBB of the last polled descriptor */
index = ring->cons & ring->size_mask;
new_index = be16_to_cpu(cqe->wqe_index) & ring->size_mask;
if (index == new_index)
return;
if (!priv->port_up)
return;
/*
* We use a two-stage loop:
* - the first samples the HW-updated CQE
* - the second frees TXBBs until the last sample
* This lets us amortize CQE cache misses, while still polling the CQ
* until is quiescent.
*/
cq_last_sav = mcq->cons_index;
do {
do {
/* Skip over last polled CQE */
index = (index + ring->last_nr_txbb) & ring->size_mask;
txbbs_skipped += ring->last_nr_txbb;
/* Poll next CQE */
ring->last_nr_txbb = mlx4_en_free_tx_desc(
priv, ring, index,
!!((ring->cons + txbbs_skipped) &
ring->size));
++mcq->cons_index;
} while (index != new_index);
new_index = be16_to_cpu(cqe->wqe_index) & ring->size_mask;
} while (index != new_index);
AVG_PERF_COUNTER(priv->pstats.tx_coal_avg,
(u32) (mcq->cons_index - cq_last_sav));
/*
* To prevent CQ overflow we first update CQ consumer and only then
* the ring consumer.
*/
mlx4_cq_set_ci(mcq);
wmb();
ring->cons += txbbs_skipped;
/* Wakeup Tx queue if this ring stopped it */
if (unlikely(ring->blocked)) {
if ((u32) (ring->prod - ring->cons) <=
ring->size - HEADROOM - MAX_DESC_TXBBS) {
ring->blocked = 0;
netif_tx_wake_queue(netdev_get_tx_queue(dev, cq->ring));
priv->port_stats.wake_queue++;
}
}
}
static bool mlx4_en_process_tx_cq(struct ether *dev,
struct mlx4_en_cq *cq)
{
struct mlx4_en_priv *priv = netdev_priv(dev);
struct mlx4_cq *mcq = &cq->mcq;
struct mlx4_en_tx_ring *ring = priv->tx_ring[cq->ring];
struct mlx4_cqe *cqe;
uint16_t index;
uint16_t new_index, ring_index, stamp_index;
uint32_t txbbs_skipped = 0;
uint32_t txbbs_stamp = 0;
uint32_t cons_index = mcq->cons_index;
int size = cq->size;
uint32_t size_mask = ring->size_mask;
struct mlx4_cqe *buf = cq->buf;
uint32_t packets = 0;
uint32_t bytes = 0;
int factor = priv->cqe_factor;
uint64_t timestamp = 0;
int done = 0;
int budget = priv->tx_work_limit;
uint32_t last_nr_txbb;
uint32_t ring_cons;
if (!priv->port_up)
return true;
#if 0 // AKAROS_PORT
netdev_txq_bql_complete_prefetchw(ring->tx_queue);
#endif
index = cons_index & size_mask;
cqe = mlx4_en_get_cqe(buf, index, priv->cqe_size) + factor;
last_nr_txbb = ACCESS_ONCE(ring->last_nr_txbb);
ring_cons = ACCESS_ONCE(ring->cons);
ring_index = ring_cons & size_mask;
stamp_index = ring_index;
/* Process all completed CQEs */
while (XNOR(cqe->owner_sr_opcode & MLX4_CQE_OWNER_MASK,
cons_index & size) && (done < budget)) {
/*
* make sure we read the CQE after we read the
* ownership bit
*/
bus_rmb();
if (unlikely((cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) ==
MLX4_CQE_OPCODE_ERROR)) {
struct mlx4_err_cqe *cqe_err = (struct mlx4_err_cqe *)cqe;
en_err(priv, "CQE error - vendor syndrome: 0x%x syndrome: 0x%x\n",
cqe_err->vendor_err_syndrome,
cqe_err->syndrome);
}
/* Skip over last polled CQE */
new_index = be16_to_cpu(cqe->wqe_index) & size_mask;
do {
txbbs_skipped += last_nr_txbb;
ring_index = (ring_index + last_nr_txbb) & size_mask;
if (ring->tx_info[ring_index].ts_requested)
timestamp = mlx4_en_get_cqe_ts(cqe);
/* free next descriptor */
last_nr_txbb = mlx4_en_free_tx_desc(
priv, ring, ring_index,
!!((ring_cons + txbbs_skipped) &
ring->size), timestamp);
mlx4_en_stamp_wqe(priv, ring, stamp_index,
!!((ring_cons + txbbs_stamp) &
ring->size));
stamp_index = ring_index;
txbbs_stamp = txbbs_skipped;
packets++;
bytes += ring->tx_info[ring_index].nr_bytes;
} while ((++done < budget) && (ring_index != new_index));
++cons_index;
index = cons_index & size_mask;
cqe = mlx4_en_get_cqe(buf, index, priv->cqe_size) + factor;
}
/*
* To prevent CQ overflow we first update CQ consumer and only then
* the ring consumer.
*/
mcq->cons_index = cons_index;
mlx4_cq_set_ci(mcq);
wmb();
/* we want to dirty this cache line once */
ACCESS_ONCE(ring->last_nr_txbb) = last_nr_txbb;
ACCESS_ONCE(ring->cons) = ring_cons + txbbs_skipped;
#if 0 // AKAROS_PORT
netdev_tx_completed_queue(ring->tx_queue, packets, bytes);
/*
* Wakeup Tx queue if this stopped, and at least 1 packet
* was completed
*/
if (netif_tx_queue_stopped(ring->tx_queue) && txbbs_skipped > 0) {
netif_tx_wake_queue(ring->tx_queue);
ring->wake_queue++;
}
#endif
return done < budget;
}
int mlx4_en_process_tx_cq(struct net_device *dev,
struct mlx4_en_cq *cq,
int budget)
{
struct mlx4_en_priv *priv = netdev_priv(dev);
struct mlx4_cq *mcq = &cq->mcq;
struct mlx4_en_tx_ring *ring = &priv->tx_ring[cq->ring];
struct mlx4_cqe *cqe;
u16 index;
u16 new_index, ring_index;
u32 txbbs_skipped = 0;
u32 cons_index = mcq->cons_index;
int size = cq->size;
u32 size_mask = ring->size_mask;
struct mlx4_cqe *buf = cq->buf;
int factor = priv->cqe_factor;
int done = 0;
index = cons_index & size_mask;
cqe = &buf[(index << factor) + factor];
ring_index = ring->cons & size_mask;
/* Process all completed CQEs */
while (XNOR(cqe->owner_sr_opcode & MLX4_CQE_OWNER_MASK,
cons_index & size) && done < budget) {
/*
* make sure we read the CQE after we read the
* ownership bit
*/
rmb();
/* Skip over last polled CQE */
new_index = be16_to_cpu(cqe->wqe_index) & size_mask;
do {
txbbs_skipped += ring->last_nr_txbb;
ring_index = (ring_index + ring->last_nr_txbb) & size_mask;
/* free next descriptor */
ring->last_nr_txbb = mlx4_en_free_tx_desc(
priv, ring, ring_index,
!!((ring->cons + txbbs_skipped) &
ring->size));
} while ((++done < budget) && ring_index != new_index);
++cons_index;
index = cons_index & size_mask;
cqe = &buf[(index << factor) + factor];
}
/*
* To prevent CQ overflow we first update CQ consumer and only then
* the ring consumer.
*/
mcq->cons_index = cons_index;
mlx4_cq_set_ci(mcq);
wmb();
ring->cons += txbbs_skipped;
atomic_sub(txbbs_skipped, &ring->inflight);
/* Wakeup Tx queue if this ring stopped it */
if (unlikely(ring->blocked && txbbs_skipped > 0)) {
ring->blocked = 0;
#ifndef __VMKERNEL_MLX4_EN_TX_HASH__
netif_tx_wake_queue(netdev_get_tx_queue(dev, cq->ring));
#else
netif_tx_wake_queue(netdev_get_tx_queue(dev, ring->reported_index));
#endif /* NOT __VMKERNEL_MLX4_EN_TX_HASH__ */
priv->port_stats.wake_queue++;
}
return done;
}
u32 mlx4_en_recycle_tx_desc(struct mlx4_en_priv *priv,
struct mlx4_en_tx_ring *ring,
int index, u64 timestamp,
int napi_mode)
{
struct mlx4_en_tx_info *tx_info = &ring->tx_info[index];
struct mlx4_en_rx_alloc frame = {
.page = tx_info->page,
.dma = tx_info->map0_dma,
};
if (!mlx4_en_rx_recycle(ring->recycle_ring, &frame)) {
dma_unmap_page(priv->ddev, tx_info->map0_dma,
PAGE_SIZE, priv->dma_dir);
put_page(tx_info->page);
}
return tx_info->nr_txbb;
}
int mlx4_en_free_tx_buf(struct net_device *dev, struct mlx4_en_tx_ring *ring)
{
struct mlx4_en_priv *priv = netdev_priv(dev);
int cnt = 0;
/* Skip last polled descriptor */
ring->cons += ring->last_nr_txbb;
en_dbg(DRV, priv, "Freeing Tx buf - cons:0x%x prod:0x%x\n",
ring->cons, ring->prod);
if ((u32) (ring->prod - ring->cons) > ring->size) {
if (netif_msg_tx_err(priv))
en_warn(priv, "Tx consumer passed producer!\n");
return 0;
}
while (ring->cons != ring->prod) {
ring->last_nr_txbb = ring->free_tx_desc(priv, ring,
ring->cons & ring->size_mask,
0, 0 /* Non-NAPI caller */);
ring->cons += ring->last_nr_txbb;
cnt++;
}
if (ring->tx_queue)
netdev_tx_reset_queue(ring->tx_queue);
if (cnt)
en_dbg(DRV, priv, "Freed %d uncompleted tx descriptors\n", cnt);
return cnt;
}
bool mlx4_en_process_tx_cq(struct net_device *dev,
struct mlx4_en_cq *cq, int napi_budget)
{
struct mlx4_en_priv *priv = netdev_priv(dev);
struct mlx4_cq *mcq = &cq->mcq;
struct mlx4_en_tx_ring *ring = priv->tx_ring[cq->type][cq->ring];
struct mlx4_cqe *cqe;
u16 index, ring_index, stamp_index;
u32 txbbs_skipped = 0;
u32 txbbs_stamp = 0;
u32 cons_index = mcq->cons_index;
int size = cq->size;
u32 size_mask = ring->size_mask;
struct mlx4_cqe *buf = cq->buf;
u32 packets = 0;
u32 bytes = 0;
int factor = priv->cqe_factor;
int done = 0;
int budget = priv->tx_work_limit;
u32 last_nr_txbb;
u32 ring_cons;
if (unlikely(!priv->port_up))
return true;
netdev_txq_bql_complete_prefetchw(ring->tx_queue);
index = cons_index & size_mask;
cqe = mlx4_en_get_cqe(buf, index, priv->cqe_size) + factor;
last_nr_txbb = READ_ONCE(ring->last_nr_txbb);
ring_cons = READ_ONCE(ring->cons);
ring_index = ring_cons & size_mask;
stamp_index = ring_index;
/* Process all completed CQEs */
while (XNOR(cqe->owner_sr_opcode & MLX4_CQE_OWNER_MASK,
cons_index & size) && (done < budget)) {
u16 new_index;
/*
* make sure we read the CQE after we read the
* ownership bit
*/
dma_rmb();
if (unlikely((cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) ==
MLX4_CQE_OPCODE_ERROR)) {
struct mlx4_err_cqe *cqe_err = (struct mlx4_err_cqe *)cqe;
en_err(priv, "CQE error - vendor syndrome: 0x%x syndrome: 0x%x\n",
cqe_err->vendor_err_syndrome,
cqe_err->syndrome);
}
/* Skip over last polled CQE */
new_index = be16_to_cpu(cqe->wqe_index) & size_mask;
do {
u64 timestamp = 0;
txbbs_skipped += last_nr_txbb;
ring_index = (ring_index + last_nr_txbb) & size_mask;
if (unlikely(ring->tx_info[ring_index].ts_requested))
timestamp = mlx4_en_get_cqe_ts(cqe);
/* free next descriptor */
last_nr_txbb = ring->free_tx_desc(
priv, ring, ring_index,
timestamp, napi_budget);
mlx4_en_stamp_wqe(priv, ring, stamp_index,
!!((ring_cons + txbbs_stamp) &
ring->size));
stamp_index = ring_index;
txbbs_stamp = txbbs_skipped;
packets++;
bytes += ring->tx_info[ring_index].nr_bytes;
} while ((++done < budget) && (ring_index != new_index));
++cons_index;
index = cons_index & size_mask;
cqe = mlx4_en_get_cqe(buf, index, priv->cqe_size) + factor;
}
/*
* To prevent CQ overflow we first update CQ consumer and only then
* the ring consumer.
*/
mcq->cons_index = cons_index;
mlx4_cq_set_ci(mcq);
wmb();
/* we want to dirty this cache line once */
WRITE_ONCE(ring->last_nr_txbb, last_nr_txbb);
WRITE_ONCE(ring->cons, ring_cons + txbbs_skipped);
if (cq->type == TX_XDP)
return done < budget;
netdev_tx_completed_queue(ring->tx_queue, packets, bytes);
/* Wakeup Tx queue if this stopped, and ring is not full.
*/
if (netif_tx_queue_stopped(ring->tx_queue) &&
!mlx4_en_is_tx_ring_full(ring)) {
netif_tx_wake_queue(ring->tx_queue);
ring->wake_queue++;
}
return done < budget;
}
void mlx4_en_tx_irq(struct mlx4_cq *mcq)
{
struct mlx4_en_cq *cq = container_of(mcq, struct mlx4_en_cq, mcq);
struct mlx4_en_priv *priv = netdev_priv(cq->dev);
if (likely(priv->port_up))
napi_schedule_irqoff(&cq->napi);
else
mlx4_en_arm_cq(priv, cq);
}
/* TX CQ polling - called by NAPI */
int mlx4_en_poll_tx_cq(struct napi_struct *napi, int budget)
{
struct mlx4_en_cq *cq = container_of(napi, struct mlx4_en_cq, napi);
struct net_device *dev = cq->dev;
struct mlx4_en_priv *priv = netdev_priv(dev);
bool clean_complete;
clean_complete = mlx4_en_process_tx_cq(dev, cq, budget);
if (!clean_complete)
return budget;
napi_complete(napi);
mlx4_en_arm_cq(priv, cq);
return 0;
}
static struct mlx4_en_tx_desc *mlx4_en_bounce_to_desc(struct mlx4_en_priv *priv,
struct mlx4_en_tx_ring *ring,
u32 index,
unsigned int desc_size)
{
u32 copy = (ring->size - index) << LOG_TXBB_SIZE;
int i;
for (i = desc_size - copy - 4; i >= 0; i -= 4) {
if ((i & (TXBB_SIZE - 1)) == 0)
wmb();
*((u32 *) (ring->buf + i)) =
*((u32 *) (ring->bounce_buf + copy + i));
}
for (i = copy - 4; i >= 4 ; i -= 4) {
if ((i & (TXBB_SIZE - 1)) == 0)
wmb();
*((u32 *)(ring->buf + (index << LOG_TXBB_SIZE) + i)) =
*((u32 *) (ring->bounce_buf + i));
}
/* Return real descriptor location */
return ring->buf + (index << LOG_TXBB_SIZE);
}
int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int budget)
{
struct mlx4_en_priv *priv = netdev_priv(dev);
struct mlx4_cqe *cqe;
struct mlx4_en_rx_ring *ring = &priv->rx_ring[cq->ring];
struct mbuf **mb_list;
struct mlx4_en_rx_desc *rx_desc;
struct mbuf *mb;
#ifdef INET
struct lro_entry *queued;
#endif
int index;
unsigned int length;
int polled = 0;
if (!priv->port_up)
return 0;
/* We assume a 1:1 mapping between CQEs and Rx descriptors, so Rx
* descriptor offset can be deduced from the CQE index instead of
* reading 'cqe->index' */
index = cq->mcq.cons_index & ring->size_mask;
cqe = &cq->buf[index];
/* Process all completed CQEs */
while (XNOR(cqe->owner_sr_opcode & MLX4_CQE_OWNER_MASK,
cq->mcq.cons_index & cq->size)) {
mb_list = ring->rx_info + (index << priv->log_rx_info);
rx_desc = ring->buf + (index << ring->log_stride);
/*
* make sure we read the CQE after we read the ownership bit
*/
rmb();
if (invalid_cqe(priv, cqe))
goto next;
/*
* Packet is OK - process it.
*/
length = be32_to_cpu(cqe->byte_cnt);
mb = mlx4_en_rx_mb(priv, rx_desc, mb_list, length);
if (!mb) {
ring->errors++;
goto next;
}
ring->bytes += length;
ring->packets++;
if (unlikely(priv->validate_loopback)) {
validate_loopback(priv, mb);
goto next;
}
mb->m_pkthdr.flowid = cq->ring;
mb->m_flags |= M_FLOWID;
mb->m_pkthdr.rcvif = dev;
if (be32_to_cpu(cqe->vlan_my_qpn) &
MLX4_CQE_VLAN_PRESENT_MASK) {
mb->m_pkthdr.ether_vtag = be16_to_cpu(cqe->sl_vid);
mb->m_flags |= M_VLANTAG;
}
if (likely(priv->rx_csum) &&
(cqe->status & cpu_to_be16(MLX4_CQE_STATUS_IPOK)) &&
(cqe->checksum == cpu_to_be16(0xffff))) {
priv->port_stats.rx_chksum_good++;
mb->m_pkthdr.csum_flags =
CSUM_IP_CHECKED | CSUM_IP_VALID |
CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
mb->m_pkthdr.csum_data = htons(0xffff);
/* This packet is eligible for LRO if it is:
* - DIX Ethernet (type interpretation)
* - TCP/IP (v4)
* - without IP options
* - not an IP fragment
*/
#ifdef INET
if (mlx4_en_can_lro(cqe->status) &&
(dev->if_capenable & IFCAP_LRO)) {
if (ring->lro.lro_cnt != 0 &&
tcp_lro_rx(&ring->lro, mb, 0) == 0)
goto next;
}
#endif
/* LRO not possible, complete processing here */
INC_PERF_COUNTER(priv->pstats.lro_misses);
} else {
mb->m_pkthdr.csum_flags = 0;
priv->port_stats.rx_chksum_none++;
#ifdef INET
if (priv->ip_reasm &&
cqe->status & cpu_to_be16(MLX4_CQE_STATUS_IPV4) &&
!mlx4_en_rx_frags(priv, ring, mb, cqe))
goto next;
#endif
}
/* Push it up the stack */
dev->if_input(dev, mb);
next:
++cq->mcq.cons_index;
index = (cq->mcq.cons_index) & ring->size_mask;
cqe = &cq->buf[index];
if (++polled == budget)
goto out;
}
/* Flush all pending IP reassembly sessions */
out:
#ifdef INET
mlx4_en_flush_frags(priv, ring);
while ((queued = SLIST_FIRST(&ring->lro.lro_active)) != NULL) {
SLIST_REMOVE_HEAD(&ring->lro.lro_active, next);
tcp_lro_flush(&ring->lro, queued);
}
#endif
AVG_PERF_COUNTER(priv->pstats.rx_coal_avg, polled);
mlx4_cq_set_ci(&cq->mcq);
wmb(); /* ensure HW sees CQ consumer before we post new buffers */
ring->cons = cq->mcq.cons_index;
ring->prod += polled; /* Polled descriptors were realocated in place */
mlx4_en_update_rx_prod_db(ring);
return polled;
}
static void mlx4_en_process_tx_cq(struct net_device *dev, struct mlx4_en_cq *cq)
{
struct mlx4_en_priv *priv = netdev_priv(dev);
struct mlx4_cq *mcq = &cq->mcq;
struct mlx4_en_tx_ring *ring = &priv->tx_ring[cq->ring];
struct mlx4_cqe *cqe;
u16 index;
u16 new_index, ring_index;
u32 txbbs_skipped = 0;
u32 cons_index = mcq->cons_index;
int size = cq->size;
u32 size_mask = ring->size_mask;
struct mlx4_cqe *buf = cq->buf;
if (!priv->port_up)
return;
index = cons_index & size_mask;
cqe = &buf[index];
ring_index = ring->cons & size_mask;
/* */
while (XNOR(cqe->owner_sr_opcode & MLX4_CQE_OWNER_MASK,
cons_index & size)) {
/*
*/
rmb();
/* */
new_index = be16_to_cpu(cqe->wqe_index) & size_mask;
do {
txbbs_skipped += ring->last_nr_txbb;
ring_index = (ring_index + ring->last_nr_txbb) & size_mask;
/* */
ring->last_nr_txbb = mlx4_en_free_tx_desc(
priv, ring, ring_index,
!!((ring->cons + txbbs_skipped) &
ring->size));
} while (ring_index != new_index);
++cons_index;
index = cons_index & size_mask;
cqe = &buf[index];
}
/*
*/
mcq->cons_index = cons_index;
mlx4_cq_set_ci(mcq);
wmb();
ring->cons += txbbs_skipped;
/* */
if (unlikely(ring->blocked)) {
if ((u32) (ring->prod - ring->cons) <=
ring->size - HEADROOM - MAX_DESC_TXBBS) {
ring->blocked = 0;
netif_tx_wake_queue(netdev_get_tx_queue(dev, cq->ring));
priv->port_stats.wake_queue++;
}
}
}
int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int budget)
{
struct mlx4_en_priv *priv = netdev_priv(dev);
struct mlx4_en_dev *mdev = priv->mdev;
struct mlx4_cqe *cqe;
struct mlx4_en_rx_ring *ring = &priv->rx_ring[cq->ring];
struct skb_frag_struct *skb_frags;
struct mlx4_en_rx_desc *rx_desc;
struct sk_buff *skb;
int index;
unsigned int length;
int polled = 0;
int ip_summed;
if (!priv->port_up)
return 0;
/* We assume a 1:1 mapping between CQEs and Rx descriptors, so Rx
* descriptor offset can be deduced from the CQE index instead of
* reading 'cqe->index' */
index = cq->mcq.cons_index & ring->size_mask;
cqe = &cq->buf[index];
/* Process all completed CQEs */
while (XNOR(cqe->owner_sr_opcode & MLX4_CQE_OWNER_MASK,
cq->mcq.cons_index & cq->size)) {
skb_frags = ring->rx_info + (index << priv->log_rx_info);
rx_desc = ring->buf + (index << ring->log_stride);
/*
* make sure we read the CQE after we read the ownership bit
*/
rmb();
/* Drop packet on bad receive or bad checksum */
if (unlikely((cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) ==
MLX4_CQE_OPCODE_ERROR)) {
mlx4_err(mdev, "CQE completed in error - vendor "
"syndrom:%d syndrom:%d\n",
((struct mlx4_err_cqe *) cqe)->vendor_err_syndrome,
((struct mlx4_err_cqe *) cqe)->syndrome);
goto next;
}
if (unlikely(cqe->badfcs_enc & MLX4_CQE_BAD_FCS)) {
mlx4_dbg(RX_ERR, priv, "Accepted frame with bad FCS\n");
goto next;
}
/*
* Packet is OK - process it.
*/
length = be32_to_cpu(cqe->byte_cnt);
ring->bytes += length;
ring->packets++;
if (likely(priv->rx_csum)) {
if ((cqe->status & MLX4_CQE_STATUS_IPOK) &&
(cqe->checksum == 0xffff)) {
priv->port_stats.rx_chksum_good++;
if (mdev->profile.num_lro &&
!mlx4_en_lro_rx(priv, ring, rx_desc,
skb_frags, length, cqe))
goto next;
/* LRO not possible, complete processing here */
ip_summed = CHECKSUM_UNNECESSARY;
INC_PERF_COUNTER(priv->pstats.lro_misses);
} else {
ip_summed = CHECKSUM_NONE;
priv->port_stats.rx_chksum_none++;
}
} else {
ip_summed = CHECKSUM_NONE;
priv->port_stats.rx_chksum_none++;
}
skb = mlx4_en_rx_skb(priv, rx_desc, skb_frags,
ring->page_alloc, length);
if (!skb) {
priv->stats.rx_dropped++;
goto next;
}
skb->ip_summed = ip_summed;
skb->protocol = eth_type_trans(skb, dev);
/* Push it up the stack */
if (priv->vlgrp && (be32_to_cpu(cqe->vlan_my_qpn) &
MLX4_CQE_VLAN_PRESENT_MASK)) {
vlan_hwaccel_receive_skb(skb, priv->vlgrp,
be16_to_cpu(cqe->sl_vid));
} else
netif_receive_skb(skb);
dev->last_rx = jiffies;
next:
++cq->mcq.cons_index;
index = (cq->mcq.cons_index) & ring->size_mask;
cqe = &cq->buf[index];
if (++polled == budget) {
/* We are here because we reached the NAPI budget -
* flush only pending LRO sessions */
if (mdev->profile.num_lro)
mlx4_en_lro_flush(priv, ring, 0);
goto out;
}
}
/* If CQ is empty flush all LRO sessions unconditionally */
if (mdev->profile.num_lro)
mlx4_en_lro_flush(priv, ring, 1);
out:
AVG_PERF_COUNTER(priv->pstats.rx_coal_avg, polled);
mlx4_cq_set_ci(&cq->mcq);
wmb(); /* ensure HW sees CQ consumer before we post new buffers */
ring->cons = cq->mcq.cons_index;
ring->prod += polled; /* Polled descriptors were realocated in place */
if (unlikely(!ring->full)) {
mlx4_en_copy_desc(priv, ring, ring->cons - polled,
ring->prod - polled, polled);
mlx4_en_fill_rx_buf(dev, ring);
}
mlx4_en_update_rx_prod_db(ring);
return polled;
}
static void mlx4_en_process_tx_cq(struct net_device *dev, struct mlx4_en_cq *cq)
{
struct mlx4_en_priv *priv = netdev_priv(dev);
struct mlx4_cq *mcq = &cq->mcq;
struct mlx4_en_tx_ring *ring = &priv->tx_ring[cq->ring];
struct mlx4_cqe *cqe;
u16 index;
u16 new_index, ring_index;
u32 txbbs_skipped = 0;
u32 cons_index = mcq->cons_index;
int size = cq->size;
u32 size_mask = ring->size_mask;
struct mlx4_cqe *buf = cq->buf;
if (!priv->port_up)
return;
index = cons_index & size_mask;
cqe = &buf[index];
ring_index = ring->cons & size_mask;
/* Process all completed CQEs */
while (XNOR(cqe->owner_sr_opcode & MLX4_CQE_OWNER_MASK,
cons_index & size)) {
/*
* make sure we read the CQE after we read the
* ownership bit
*/
rmb();
/* Skip over last polled CQE */
new_index = be16_to_cpu(cqe->wqe_index) & size_mask;
do {
txbbs_skipped += ring->last_nr_txbb;
ring_index = (ring_index + ring->last_nr_txbb) & size_mask;
/* free next descriptor */
ring->last_nr_txbb = mlx4_en_free_tx_desc(
priv, ring, ring_index,
!!((ring->cons + txbbs_skipped) &
ring->size));
} while (ring_index != new_index);
++cons_index;
index = cons_index & size_mask;
cqe = &buf[index];
}
/*
* To prevent CQ overflow we first update CQ consumer and only then
* the ring consumer.
*/
mcq->cons_index = cons_index;
mlx4_cq_set_ci(mcq);
wmb();
ring->cons += txbbs_skipped;
/* Wakeup Tx queue if this ring stopped it */
if (unlikely(ring->blocked)) {
if ((u32) (ring->prod - ring->cons) <=
ring->size - HEADROOM - MAX_DESC_TXBBS) {
ring->blocked = 0;
netif_tx_wake_queue(netdev_get_tx_queue(dev, cq->ring));
priv->port_stats.wake_queue++;
}
}
}
