* [PATCH v1 3/7] gianfar: Add Multiple Queue Support
From: Sandeep Gopalpet @ 2009-10-26 16:27 UTC (permalink / raw)
To: netdev; +Cc: Sandeep Gopalpet
This patch introduces multiple Tx and Rx queues.
The incoming packets can be classified into different queues
based on filer rules (out of scope of this patch). The number
of queues enabled will be based on a DTS entries fsl,num_tx_queues
and fsl,num_rx_queues.
Although we are enabling multiple queues, the interrupt coalescing
is on per device level (etsec-1.7 doesn't support multiple rxics
and txics).
Signed-off-by: Sandeep Gopalpet <sandeep.kumar@freescale.com>
---
drivers/net/gianfar.c | 682 +++++++++++++++++++++++++++--------------
drivers/net/gianfar.h | 95 ++++++-
drivers/net/gianfar_ethtool.c | 70 +++--
drivers/net/gianfar_sysfs.c | 50 ++--
4 files changed, 606 insertions(+), 291 deletions(-)
diff --git a/drivers/net/gianfar.c b/drivers/net/gianfar.c
index fa0188e..82da5d9 100644
--- a/drivers/net/gianfar.c
+++ b/drivers/net/gianfar.c
@@ -143,6 +143,7 @@ void gfar_start(struct net_device *dev);
static void gfar_clear_exact_match(struct net_device *dev);
static void gfar_set_mac_for_addr(struct net_device *dev, int num, u8 *addr);
static int gfar_ioctl(struct net_device *dev, struct ifreq *rq, int cmd);
+u16 gfar_select_queue(struct net_device *dev, struct sk_buff *skb);
MODULE_AUTHOR("Freescale Semiconductor, Inc");
MODULE_DESCRIPTION("Gianfar Ethernet Driver");
@@ -171,71 +172,91 @@ static int gfar_init_bds(struct net_device *ndev)
struct gfar_priv_rx_q *rx_queue = NULL;
struct txbd8 *txbdp;
struct rxbd8 *rxbdp;
- int i;
-
- tx_queue = priv->tx_queue;
- rx_queue = priv->rx_queue;
+ int i, j;
- /* Initialize some variables in our dev structure */
- tx_queue->num_txbdfree = tx_queue->tx_ring_size;
- tx_queue->dirty_tx = tx_queue->cur_tx = tx_queue->tx_bd_base;
- rx_queue->cur_rx = rx_queue->rx_bd_base;
- tx_queue->skb_curtx = tx_queue->skb_dirtytx = 0;
- rx_queue->skb_currx = 0;
+ for (i = 0; i < priv->num_tx_queues; i++) {
+ tx_queue = priv->tx_queue[i];
+ /* Initialize some variables in our dev structure */
+ tx_queue->num_txbdfree = tx_queue->tx_ring_size;
+ tx_queue->dirty_tx = tx_queue->tx_bd_base;
+ tx_queue->cur_tx = tx_queue->tx_bd_base;
+ tx_queue->skb_curtx = 0;
+ tx_queue->skb_dirtytx = 0;
+
+ /* Initialize Transmit Descriptor Ring */
+ txbdp = tx_queue->tx_bd_base;
+ for (j = 0; j < tx_queue->tx_ring_size; j++) {
+ txbdp->lstatus = 0;
+ txbdp->bufPtr = 0;
+ txbdp++;
+ }
- /* Initialize Transmit Descriptor Ring */
- txbdp = tx_queue->tx_bd_base;
- for (i = 0; i < tx_queue->tx_ring_size; i++) {
- txbdp->lstatus = 0;
- txbdp->bufPtr = 0;
- txbdp++;
+ /* Set the last descriptor in the ring to indicate wrap */
+ txbdp--;
+ txbdp->status |= TXBD_WRAP;
}
- /* Set the last descriptor in the ring to indicate wrap */
- txbdp--;
- txbdp->status |= TXBD_WRAP;
+ for (i = 0; i < priv->num_rx_queues; i++) {
+ rx_queue = priv->rx_queue[i];
+ rx_queue->cur_rx = rx_queue->rx_bd_base;
+ rx_queue->skb_currx = 0;
+ rxbdp = rx_queue->rx_bd_base;
- rxbdp = rx_queue->rx_bd_base;
- for (i = 0; i < rx_queue->rx_ring_size; i++) {
- struct sk_buff *skb = rx_queue->rx_skbuff[i];
+ for (j = 0; j < rx_queue->rx_ring_size; j++) {
+ struct sk_buff *skb = rx_queue->rx_skbuff[j];
- if (skb) {
- gfar_init_rxbdp(rx_queue, rxbdp, rxbdp->bufPtr);
- } else {
- skb = gfar_new_skb(ndev);
- if (!skb) {
- pr_err("%s: Can't allocate RX buffers\n",
- ndev->name);
- return -ENOMEM;
+ if (skb) {
+ gfar_init_rxbdp(rx_queue, rxbdp,
+ rxbdp->bufPtr);
+ } else {
+ skb = gfar_new_skb(ndev);
+ if (!skb) {
+ pr_err("%s: Can't allocate RX buffers\n",
+ ndev->name);
+ goto err_rxalloc_fail;
+ }
+ rx_queue->rx_skbuff[j] = skb;
+
+ gfar_new_rxbdp(rx_queue, rxbdp, skb);
}
- rx_queue->rx_skbuff[i] = skb;
- gfar_new_rxbdp(rx_queue, rxbdp, skb);
+ rxbdp++;
}
- rxbdp++;
}
return 0;
+
+err_rxalloc_fail:
+ free_skb_resources(priv);
+ return -ENOMEM;
}
static int gfar_alloc_skb_resources(struct net_device *ndev)
{
void *vaddr;
- int i;
+ dma_addr_t addr;
+ int i, j, k;
struct gfar_private *priv = netdev_priv(ndev);
struct device *dev = &priv->ofdev->dev;
struct gfar_priv_tx_q *tx_queue = NULL;
struct gfar_priv_rx_q *rx_queue = NULL;
+ struct gfar __iomem *regs = priv->gfargrp.regs;
+ u32 *baddr;
+
+ priv->total_tx_ring_size = 0;
+ for (i = 0; i < priv->num_tx_queues; i++)
+ priv->total_tx_ring_size += priv->tx_queue[i]->tx_ring_size;
- tx_queue = priv->tx_queue;
- rx_queue = priv->rx_queue;
+ priv->total_rx_ring_size = 0;
+ for (i = 0; i < priv->num_rx_queues; i++)
+ priv->total_rx_ring_size += priv->rx_queue[i]->rx_ring_size;
/* Allocate memory for the buffer descriptors */
vaddr = dma_alloc_coherent(dev,
- sizeof(*tx_queue->tx_bd_base) * tx_queue->tx_ring_size +
- sizeof(*rx_queue->rx_bd_base) * rx_queue->rx_ring_size,
- &tx_queue->tx_bd_dma_base, GFP_KERNEL);
+ sizeof(struct txbd8) * priv->total_tx_ring_size +
+ sizeof(struct rxbd8) * priv->total_rx_ring_size,
+ &addr, GFP_KERNEL);
if (!vaddr) {
if (netif_msg_ifup(priv))
pr_err("%s: Could not allocate buffer descriptors!\n",
@@ -243,38 +264,62 @@ static int gfar_alloc_skb_resources(struct net_device *ndev)
return -ENOMEM;
}
- tx_queue->tx_bd_base = vaddr;
- tx_queue->dev = ndev;
+ baddr = ®s->tbase0;
+ for (i = 0; i < priv->num_tx_queues; i++) {
+ tx_queue = priv->tx_queue[i];
+ tx_queue->tx_bd_base = (struct txbd8 *) vaddr;
+ tx_queue->tx_bd_dma_base = addr;
+ tx_queue->dev = ndev;
+ /* enet DMA only understands physical addresses */
+ gfar_write(baddr, addr);
+ addr += sizeof(struct txbd8) *tx_queue->tx_ring_size;
+ vaddr += sizeof(struct txbd8) *tx_queue->tx_ring_size;
+ baddr += 2;
+ }
+ baddr = ®s->rbase0;
/* Start the rx descriptor ring where the tx ring leaves off */
- vaddr = vaddr + sizeof(*tx_queue->tx_bd_base) * tx_queue->tx_ring_size;
- rx_queue->rx_bd_base = vaddr;
- rx_queue->dev = ndev;
+ for (i = 0; i < priv->num_rx_queues; i++) {
+ rx_queue = priv->rx_queue[i];
+ rx_queue->rx_bd_base = (struct rxbd8 *) vaddr;
+ rx_queue->dev = ndev;
+ gfar_write(baddr, addr);
+ addr += sizeof (struct rxbd8) * rx_queue->rx_ring_size;
+ vaddr += sizeof (struct rxbd8) * rx_queue->rx_ring_size;
+ baddr += 2;
+ }
/* Setup the skbuff rings */
- tx_queue->tx_skbuff = kmalloc(sizeof(*tx_queue->tx_skbuff) *
+ for (i = 0; i < priv->num_tx_queues; i++) {
+ tx_queue = priv->tx_queue[i];
+ tx_queue->tx_skbuff = kmalloc(sizeof(*tx_queue->tx_skbuff) *
tx_queue->tx_ring_size, GFP_KERNEL);
- if (!tx_queue->tx_skbuff) {
- if (netif_msg_ifup(priv))
- pr_err("%s: Could not allocate tx_skbuff\n",
- ndev->name);
- goto cleanup;
- }
+ if (!tx_queue->tx_skbuff) {
+ if (netif_msg_ifup(priv))
+ pr_err("%s: Could not allocate tx_skbuff\n",
+ ndev->name);
+ goto cleanup;
+ }
- for (i = 0; i < tx_queue->tx_ring_size; i++)
- tx_queue->tx_skbuff[i] = NULL;
+ for (k = 0; k < tx_queue->tx_ring_size; k++)
+ tx_queue->tx_skbuff[k] = NULL;
+ }
- rx_queue->rx_skbuff = kmalloc(sizeof(*rx_queue->rx_skbuff) *
+ for (i = 0; i < priv->num_rx_queues; i++) {
+ rx_queue = priv->rx_queue[i];
+ rx_queue->rx_skbuff = kmalloc(sizeof(*rx_queue->rx_skbuff) *
rx_queue->rx_ring_size, GFP_KERNEL);
- if (!rx_queue->rx_skbuff) {
- if (netif_msg_ifup(priv))
- pr_err("%s: Could not allocate rx_skbuff\n",
- ndev->name);
- goto cleanup;
- }
- for (i = 0; i < rx_queue->rx_ring_size; i++)
- rx_queue->rx_skbuff[i] = NULL;
+ if (!rx_queue->rx_skbuff) {
+ if (netif_msg_ifup(priv))
+ pr_err("%s: Could not allocate rx_skbuff\n",
+ ndev->name);
+ goto cleanup;
+ }
+
+ for (j = 0; j < rx_queue->rx_ring_size; j++)
+ rx_queue->rx_skbuff[j] = NULL;
+ }
if (gfar_init_bds(ndev))
goto cleanup;
@@ -289,30 +334,22 @@ cleanup:
static void gfar_init_mac(struct net_device *ndev)
{
struct gfar_private *priv = netdev_priv(ndev);
- struct gfar_priv_tx_q *tx_queue = NULL;
- struct gfar_priv_rx_q *rx_queue = NULL;
struct gfar __iomem *regs = priv->gfargrp.regs;
u32 rctrl = 0;
u32 tctrl = 0;
u32 attrs = 0;
- tx_queue = priv->tx_queue;
- rx_queue = priv->rx_queue;
-
- /* enet DMA only understands physical addresses */
- gfar_write(®s->tbase0, tx_queue->tx_bd_dma_base);
- gfar_write(®s->rbase0, tx_queue->tx_bd_dma_base +
- sizeof(*tx_queue->tx_bd_base) *
- tx_queue->tx_ring_size);
-
/* Configure the coalescing support */
gfar_write(®s->txic, 0);
- if (tx_queue->txcoalescing)
- gfar_write(®s->txic, tx_queue->txic);
+ if (priv->tx_queue[0]->txcoalescing)
+ gfar_write(®s->txic, priv->tx_queue[0]->txic);
gfar_write(®s->rxic, 0);
- if (rx_queue->rxcoalescing)
- gfar_write(®s->rxic, rx_queue->rxic);
+ if (priv->rx_queue[0]->rxcoalescing)
+ gfar_write(®s->rxic, priv->rx_queue[0]->rxic);
+
+ if (priv->rx_filer_enable)
+ rctrl |= RCTRL_FILREN;
if (priv->rx_csum_enable)
rctrl |= RCTRL_CHECKSUMMING;
@@ -341,6 +378,8 @@ static void gfar_init_mac(struct net_device *ndev)
if (ndev->features & NETIF_F_IP_CSUM)
tctrl |= TCTRL_INIT_CSUM;
+ tctrl |= TCTRL_TXSCHED_PRIO;
+
gfar_write(®s->tctrl, tctrl);
/* Set the extraction length and index */
@@ -374,6 +413,7 @@ static const struct net_device_ops gfar_netdev_ops = {
.ndo_set_multicast_list = gfar_set_multi,
.ndo_tx_timeout = gfar_timeout,
.ndo_do_ioctl = gfar_ioctl,
+ .ndo_select_queue = gfar_select_queue,
.ndo_vlan_rx_register = gfar_vlan_rx_register,
.ndo_set_mac_address = eth_mac_addr,
.ndo_validate_addr = eth_validate_addr,
@@ -382,36 +422,131 @@ static const struct net_device_ops gfar_netdev_ops = {
#endif
};
+inline void lock_rx_qs(struct gfar_private *priv)
+{
+ int i = 0x0;
+
+ for (i = 0; i < priv->num_rx_queues; i++)
+ spin_lock(&priv->rx_queue[i]->rxlock);
+}
+
+inline void lock_tx_qs(struct gfar_private *priv)
+{
+ int i = 0x0;
+
+ for (i = 0; i < priv->num_tx_queues; i++)
+ spin_lock(&priv->tx_queue[i]->txlock);
+}
+
+inline void unlock_rx_qs(struct gfar_private *priv)
+{
+ int i = 0x0;
+
+ for (i = 0; i < priv->num_rx_queues; i++)
+ spin_unlock(&priv->rx_queue[i]->rxlock);
+}
+
+inline void unlock_tx_qs(struct gfar_private *priv)
+{
+ int i = 0x0;
+
+ for (i = 0; i < priv->num_tx_queues; i++)
+ spin_unlock(&priv->tx_queue[i]->txlock);
+}
+
/* Returns 1 if incoming frames use an FCB */
static inline int gfar_uses_fcb(struct gfar_private *priv)
{
return priv->vlgrp || priv->rx_csum_enable;
}
-static int gfar_of_init(struct net_device *dev)
+u16 gfar_select_queue(struct net_device *dev, struct sk_buff *skb)
+{
+ return skb_get_queue_mapping(skb);
+}
+static void free_tx_pointers(struct gfar_private *priv)
+{
+ int i = 0;
+
+ for (i = 0; i < priv->num_tx_queues; i++)
+ kfree(priv->tx_queue[i]);
+}
+
+static void free_rx_pointers(struct gfar_private *priv)
+{
+ int i = 0;
+
+ for (i = 0; i < priv->num_rx_queues; i++)
+ kfree(priv->rx_queue[i]);
+}
+
+static int gfar_of_init(struct of_device *ofdev, struct net_device **pdev)
{
const char *model;
const char *ctype;
const void *mac_addr;
u64 addr, size;
- int err = 0;
- struct gfar_private *priv = netdev_priv(dev);
- struct device_node *np = priv->node;
+ int err = 0, i;
+ struct net_device *dev = NULL;
+ struct gfar_private *priv = NULL;
+ struct device_node *np = ofdev->node;
const u32 *stash;
const u32 *stash_len;
const u32 *stash_idx;
+ unsigned int num_tx_qs, num_rx_qs;
+ u32 *tx_queues, *rx_queues;
if (!np || !of_device_is_available(np))
return -ENODEV;
+ /* parse the num of tx and rx queues */
+ tx_queues = (u32 *)of_get_property(np, "fsl,num_tx_queues", NULL);
+ num_tx_qs = tx_queues ? *tx_queues : 1;
+
+ if (num_tx_qs > MAX_TX_QS) {
+ printk(KERN_ERR "num_tx_qs(=%d) greater than MAX_TX_QS(=%d)\n",
+ num_tx_qs, MAX_TX_QS);
+ printk(KERN_ERR "Cannot do alloc_etherdev, aborting\n");
+ return -EINVAL;
+ }
+
+ rx_queues = (u32 *)of_get_property(np, "fsl,num_rx_queues", NULL);
+ num_rx_qs = rx_queues ? *rx_queues : 1;
+
+ if (num_rx_qs > MAX_RX_QS) {
+ printk(KERN_ERR "num_rx_qs(=%d) greater than MAX_RX_QS(=%d)\n",
+ num_tx_qs, MAX_TX_QS);
+ printk(KERN_ERR "Cannot do alloc_etherdev, aborting\n");
+ return -EINVAL;
+ }
+
+ *pdev = alloc_etherdev_mq(sizeof(*priv), num_tx_qs);
+ dev = *pdev;
+ if (NULL == dev)
+ return -ENOMEM;
+
+ priv = netdev_priv(dev);
+ priv->node = ofdev->node;
+ priv->ndev = dev;
+
+ dev->num_tx_queues = num_tx_qs;
+ dev->real_num_tx_queues = num_tx_qs;
+ priv->num_tx_queues = num_tx_qs;
+ priv->num_rx_queues = num_rx_qs;
+
/* get a pointer to the register memory */
addr = of_translate_address(np, of_get_address(np, 0, &size, NULL));
priv->gfargrp.regs = ioremap(addr, size);
- if (priv->gfargrp.regs == NULL)
- return -ENOMEM;
+ if (priv->gfargrp.regs == NULL) {
+ err = -ENOMEM;
+ goto err_out;
+ }
priv->gfargrp.priv = priv; /* back pointer from group to priv */
+ priv->gfargrp.rx_bit_map = DEFAULT_MAPPING;
+ priv->gfargrp.tx_bit_map = DEFAULT_MAPPING;
+
priv->gfargrp.interruptTransmit = irq_of_parse_and_map(np, 0);
model = of_get_property(np, "model", NULL);
@@ -430,6 +565,38 @@ static int gfar_of_init(struct net_device *dev)
}
}
+ for (i = 0; i < priv->num_tx_queues; i++)
+ priv->tx_queue[i] = NULL;
+ for (i = 0; i < priv->num_rx_queues; i++)
+ priv->rx_queue[i] = NULL;
+
+ for (i = 0; i < priv->num_tx_queues; i++) {
+ priv->tx_queue[i] = (struct gfar_priv_tx_q *)kmalloc(
+ sizeof (struct gfar_priv_tx_q), GFP_KERNEL);
+ if (!priv->tx_queue[i]) {
+ err = -ENOMEM;
+ goto tx_alloc_failed;
+ }
+ priv->tx_queue[i]->tx_skbuff = NULL;
+ priv->tx_queue[i]->qindex = i;
+ priv->tx_queue[i]->dev = dev;
+ spin_lock_init(&(priv->tx_queue[i]->txlock));
+ }
+
+ for (i = 0; i < priv->num_rx_queues; i++) {
+ priv->rx_queue[i] = (struct gfar_priv_rx_q *)kmalloc(
+ sizeof (struct gfar_priv_rx_q), GFP_KERNEL);
+ if (!priv->rx_queue[i]) {
+ err = -ENOMEM;
+ goto rx_alloc_failed;
+ }
+ priv->rx_queue[i]->rx_skbuff = NULL;
+ priv->rx_queue[i]->qindex = i;
+ priv->rx_queue[i]->dev = dev;
+ spin_lock_init(&(priv->rx_queue[i]->rxlock));
+ }
+
+
stash = of_get_property(np, "bd-stash", NULL);
if (stash) {
@@ -490,8 +657,13 @@ static int gfar_of_init(struct net_device *dev)
return 0;
+rx_alloc_failed:
+ free_rx_pointers(priv);
+tx_alloc_failed:
+ free_tx_pointers(priv);
err_out:
iounmap(priv->gfargrp.regs);
+ free_netdev(dev);
return err;
}
@@ -509,6 +681,17 @@ static int gfar_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
return phy_mii_ioctl(priv->phydev, if_mii(rq), cmd);
}
+static unsigned int reverse_bitmap(unsigned int bit_map, unsigned int max_qs)
+{
+ unsigned int new_bit_map = 0x0;
+ int mask = 0x1 << (max_qs - 1), i;
+ for (i = 0; i < max_qs; i++) {
+ if (bit_map & mask)
+ new_bit_map = new_bit_map + (1 << i);
+ mask = mask >> 0x1;
+ }
+ return new_bit_map;
+}
/* Set up the ethernet device structure, private data,
* and anything else we need before we start */
static int gfar_probe(struct of_device *ofdev,
@@ -518,14 +701,14 @@ static int gfar_probe(struct of_device *ofdev,
struct net_device *dev = NULL;
struct gfar_private *priv = NULL;
struct gfar __iomem *regs = NULL;
- int err = 0;
+ int err = 0, i;
int len_devname;
+ u32 rstat = 0, tstat = 0, rqueue = 0, tqueue = 0;
- /* Create an ethernet device instance */
- dev = alloc_etherdev(sizeof (*priv));
+ err = gfar_of_init(ofdev, &dev);
- if (NULL == dev)
- return -ENOMEM;
+ if (err)
+ return err;
priv = netdev_priv(dev);
priv->ndev = dev;
@@ -533,23 +716,6 @@ static int gfar_probe(struct of_device *ofdev,
priv->node = ofdev->node;
SET_NETDEV_DEV(dev, &ofdev->dev);
- err = gfar_of_init(dev);
-
- if (err)
- goto regs_fail;
-
- priv->tx_queue = (struct gfar_priv_tx_q *)kmalloc(
- sizeof (struct gfar_priv_tx_q), GFP_KERNEL);
- if (!priv->tx_queue)
- goto regs_fail;
-
- priv->rx_queue = (struct gfar_priv_rx_q *)kmalloc(
- sizeof (struct gfar_priv_rx_q), GFP_KERNEL);
- if (!priv->rx_queue)
- goto rx_queue_fail;
-
- spin_lock_init(&priv->tx_queue->txlock);
- spin_lock_init(&priv->rx_queue->rxlock);
spin_lock_init(&priv->gfargrp.grplock);
spin_lock_init(&priv->bflock);
INIT_WORK(&priv->reset_task, gfar_reset_task);
@@ -587,8 +753,8 @@ static int gfar_probe(struct of_device *ofdev,
dev->netdev_ops = &gfar_netdev_ops;
dev->ethtool_ops = &gfar_ethtool_ops;
- /* Register for napi ...NAPI is for each rx_queue */
- netif_napi_add(dev, &priv->rx_queue->napi, gfar_poll, GFAR_DEV_WEIGHT);
+ /* Register for napi ...We are registering NAPI for each grp */
+ netif_napi_add(dev, &priv->gfargrp.napi, gfar_poll, GFAR_DEV_WEIGHT);
if (priv->device_flags & FSL_GIANFAR_DEV_HAS_CSUM) {
priv->rx_csum_enable = 1;
@@ -644,17 +810,44 @@ static int gfar_probe(struct of_device *ofdev,
if (dev->features & NETIF_F_IP_CSUM)
dev->hard_header_len += GMAC_FCB_LEN;
+ /* Need to reverse the bit maps as bit_map's MSB is q0
+ * but, for_each_bit parses from right to left, which
+ * basically reverses the queue numbers */
+ priv->gfargrp.tx_bit_map = reverse_bitmap(priv->gfargrp.tx_bit_map, MAX_TX_QS);
+ priv->gfargrp.rx_bit_map = reverse_bitmap(priv->gfargrp.rx_bit_map, MAX_RX_QS);
+
+ /* Calculate RSTAT, TSTAT, RQUEUE and TQUEUE values */
+ for_each_bit(i, &priv->gfargrp.rx_bit_map, priv->num_rx_queues) {
+ priv->gfargrp.num_rx_queues++;
+ rstat = rstat | (RSTAT_CLEAR_RHALT >> i);
+ rqueue = rqueue | ((RQUEUE_EN0 | RQUEUE_EX0) >> i);
+ }
+ for_each_bit (i, &priv->gfargrp.tx_bit_map, priv->num_tx_queues) {
+ priv->gfargrp.num_tx_queues++;
+ tstat = tstat | (TSTAT_CLEAR_THALT >> i);
+ tqueue = tqueue | (TQUEUE_EN0 >> i);
+ }
+ priv->gfargrp.rstat = rstat;
+ priv->gfargrp.tstat = tstat;
+
+ gfar_write(®s->rqueue, rqueue);
+ gfar_write(®s->tqueue, tqueue);
+
priv->rx_buffer_size = DEFAULT_RX_BUFFER_SIZE;
/* Initializing some of the rx/tx queue level parameters */
- priv->tx_queue->tx_ring_size = DEFAULT_TX_RING_SIZE;
- priv->tx_queue->num_txbdfree = DEFAULT_TX_RING_SIZE;
- priv->tx_queue->txcoalescing = DEFAULT_TX_COALESCE;
- priv->tx_queue->txic = DEFAULT_TXIC;
+ for (i = 0; i < priv->num_tx_queues; i++) {
+ priv->tx_queue[i]->tx_ring_size = DEFAULT_TX_RING_SIZE;
+ priv->tx_queue[i]->num_txbdfree = DEFAULT_TX_RING_SIZE;
+ priv->tx_queue[i]->txcoalescing = DEFAULT_TX_COALESCE;
+ priv->tx_queue[i]->txic = DEFAULT_TXIC;
+ }
- priv->rx_queue->rx_ring_size = DEFAULT_RX_RING_SIZE;
- priv->rx_queue->rxcoalescing = DEFAULT_RX_COALESCE;
- priv->rx_queue->rxic = DEFAULT_RXIC;
+ for (i = 0; i < priv->num_rx_queues; i++) {
+ priv->rx_queue[i]->rx_ring_size = DEFAULT_RX_RING_SIZE;
+ priv->rx_queue[i]->rxcoalescing = DEFAULT_RX_COALESCE;
+ priv->rx_queue[i]->rxic = DEFAULT_RXIC;
+ }
/* Enable most messages by default */
priv->msg_enable = (NETIF_MSG_IFUP << 1 ) - 1;
@@ -699,17 +892,19 @@ static int gfar_probe(struct of_device *ofdev,
/* Even more device info helps when determining which kernel */
/* provided which set of benchmarks. */
printk(KERN_INFO "%s: Running with NAPI enabled\n", dev->name);
- printk(KERN_INFO "%s: %d/%d RX/TX BD ring size\n",
- dev->name, priv->rx_queue->rx_ring_size, priv->tx_queue->tx_ring_size);
+ for (i = 0; i < priv->num_rx_queues; i++)
+ printk(KERN_INFO "%s: :RX BD ring size for Q[%d]: %d\n",
+ dev->name, i, priv->rx_queue[i]->rx_ring_size);
+ for(i = 0; i < priv->num_tx_queues; i++)
+ printk(KERN_INFO "%s:TX BD ring size for Q[%d]: %d\n",
+ dev->name, i, priv->tx_queue[i]->tx_ring_size);
return 0;
register_fail:
iounmap(priv->gfargrp.regs);
- kfree(priv->rx_queue);
-rx_queue_fail:
- kfree(priv->tx_queue);
-regs_fail:
+ free_tx_pointers(priv);
+ free_rx_pointers(priv);
if (priv->phy_node)
of_node_put(priv->phy_node);
if (priv->tbi_node)
@@ -742,8 +937,6 @@ static int gfar_suspend(struct device *dev)
{
struct gfar_private *priv = dev_get_drvdata(dev);
struct net_device *ndev = priv->ndev;
- struct gfar_priv_tx_q *tx_queue = NULL;
- struct gfar_priv_rx_q *rx_queue = NULL;
struct gfar __iomem *regs = NULL;
unsigned long flags;
u32 tempval;
@@ -752,13 +945,13 @@ static int gfar_suspend(struct device *dev)
(priv->device_flags & FSL_GIANFAR_DEV_HAS_MAGIC_PACKET);
netif_device_detach(ndev);
- tx_queue = priv->tx_queue;
- rx_queue = priv->rx_queue;
regs = priv->gfargrp.regs;
if (netif_running(ndev)) {
- spin_lock_irqsave(&tx_queue->txlock, flags);
- spin_lock(&rx_queue->rxlock);
+
+ local_irq_save(flags);
+ lock_tx_qs(priv);
+ lock_rx_qs(priv);
gfar_halt_nodisable(ndev);
@@ -772,10 +965,11 @@ static int gfar_suspend(struct device *dev)
gfar_write(®s->maccfg1, tempval);
- spin_unlock(&rx_queue->rxlock);
- spin_unlock_irqrestore(&tx_queue->txlock, flags);
+ unlock_rx_qs(priv);
+ unlock_tx_qs(priv);
+ local_irq_restore(flags);
- napi_disable(&rx_queue->napi);
+ napi_disable(&priv->gfargrp.napi);
if (magic_packet) {
/* Enable interrupt on Magic Packet */
@@ -797,8 +991,6 @@ static int gfar_resume(struct device *dev)
{
struct gfar_private *priv = dev_get_drvdata(dev);
struct net_device *ndev = priv->ndev;
- struct gfar_priv_tx_q *tx_queue = NULL;
- struct gfar_priv_rx_q *rx_queue = NULL;
struct gfar __iomem *regs = NULL;
unsigned long flags;
u32 tempval;
@@ -816,12 +1008,11 @@ static int gfar_resume(struct device *dev)
/* Disable Magic Packet mode, in case something
* else woke us up.
*/
- rx_queue = priv->rx_queue;
- tx_queue = priv->tx_queue;
regs = priv->gfargrp.regs;
- spin_lock_irqsave(&tx_queue->txlock, flags);
- spin_lock(&rx_queue->rxlock);
+ local_irq_save(flags);
+ lock_tx_qs(priv);
+ lock_rx_qs(priv);
tempval = gfar_read(®s->maccfg2);
tempval &= ~MACCFG2_MPEN;
@@ -829,12 +1020,13 @@ static int gfar_resume(struct device *dev)
gfar_start(ndev);
- spin_unlock(&rx_queue->rxlock);
- spin_unlock_irqrestore(&tx_queue->txlock, flags);
+ unlock_rx_qs(priv);
+ unlock_tx_qs(priv);
+ local_irq_restore(flags);
netif_device_attach(ndev);
- napi_enable(&rx_queue->napi);
+ napi_enable(&priv->gfargrp.napi);
return 0;
}
@@ -861,7 +1053,7 @@ static int gfar_restore(struct device *dev)
phy_start(priv->phydev);
netif_device_attach(ndev);
- napi_enable(&priv->napi);
+ napi_enable(&priv->gfargrp.napi);
return 0;
}
@@ -1115,23 +1307,21 @@ void gfar_halt(struct net_device *dev)
void stop_gfar(struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
- struct gfar_priv_tx_q *tx_queue = NULL;
- struct gfar_priv_rx_q *rx_queue = NULL;
unsigned long flags;
phy_stop(priv->phydev);
- tx_queue = priv->tx_queue;
- rx_queue = priv->rx_queue;
/* Lock it down */
- spin_lock_irqsave(&tx_queue->txlock, flags);
- spin_lock(&rx_queue->rxlock);
+ local_irq_save(flags);
+ lock_tx_qs(priv);
+ lock_rx_qs(priv);
gfar_halt(dev);
- spin_unlock(&rx_queue->rxlock);
- spin_unlock_irqrestore(&tx_queue->txlock, flags);
+ unlock_rx_qs(priv);
+ unlock_tx_qs(priv);
+ local_irq_restore(flags);
/* Free the IRQs */
if (priv->device_flags & FSL_GIANFAR_DEV_HAS_MULTI_INTR) {
@@ -1145,24 +1335,14 @@ void stop_gfar(struct net_device *dev)
free_skb_resources(priv);
}
-/* If there are any tx skbs or rx skbs still around, free them.
- * Then free tx_skbuff and rx_skbuff */
-static void free_skb_resources(struct gfar_private *priv)
+static void free_skb_tx_queue(struct gfar_priv_tx_q *tx_queue)
{
- struct device *dev = &priv->ofdev->dev;
- struct rxbd8 *rxbdp;
struct txbd8 *txbdp;
- struct gfar_priv_tx_q *tx_queue = NULL;
- struct gfar_priv_rx_q *rx_queue = NULL;
+ struct gfar_private *priv = netdev_priv(tx_queue->dev);
int i, j;
- /* Go through all the buffer descriptors and free their data buffers */
- tx_queue = priv->tx_queue;
txbdp = tx_queue->tx_bd_base;
- if (!tx_queue->tx_skbuff)
- goto skip_tx_skbuff;
-
for (i = 0; i < tx_queue->tx_ring_size; i++) {
if (!tx_queue->tx_skbuff[i])
continue;
@@ -1170,7 +1350,8 @@ static void free_skb_resources(struct gfar_private *priv)
dma_unmap_single(&priv->ofdev->dev, txbdp->bufPtr,
txbdp->length, DMA_TO_DEVICE);
txbdp->lstatus = 0;
- for (j = 0; j < skb_shinfo(tx_queue->tx_skbuff[i])->nr_frags; j++) {
+ for (j = 0; j < skb_shinfo(tx_queue->tx_skbuff[i])->nr_frags;
+ j++) {
txbdp++;
dma_unmap_page(&priv->ofdev->dev, txbdp->bufPtr,
txbdp->length, DMA_TO_DEVICE);
@@ -1179,36 +1360,58 @@ static void free_skb_resources(struct gfar_private *priv)
dev_kfree_skb_any(tx_queue->tx_skbuff[i]);
tx_queue->tx_skbuff[i] = NULL;
}
-
kfree(tx_queue->tx_skbuff);
-skip_tx_skbuff:
+}
- rx_queue = priv->rx_queue;
- rxbdp = rx_queue->rx_bd_base;
+static void free_skb_rx_queue(struct gfar_priv_rx_q *rx_queue)
+{
+ struct rxbd8 *rxbdp;
+ struct gfar_private *priv = netdev_priv(rx_queue->dev);
+ int i;
- if (!rx_queue->rx_skbuff)
- goto skip_rx_skbuff;
+ rxbdp = rx_queue->rx_bd_base;
for (i = 0; i < rx_queue->rx_ring_size; i++) {
if (rx_queue->rx_skbuff[i]) {
- dma_unmap_single(&priv->ofdev->dev, rxbdp->bufPtr,
- priv->rx_buffer_size,
+ dma_unmap_single(&priv->ofdev->dev,
+ rxbdp->bufPtr, priv->rx_buffer_size,
DMA_FROM_DEVICE);
dev_kfree_skb_any(rx_queue->rx_skbuff[i]);
rx_queue->rx_skbuff[i] = NULL;
}
-
rxbdp->lstatus = 0;
rxbdp->bufPtr = 0;
rxbdp++;
}
-
kfree(rx_queue->rx_skbuff);
-skip_rx_skbuff:
+}
+
+/* If there are any tx skbs or rx skbs still around, free them.
+ * Then free tx_skbuff and rx_skbuff */
+static void free_skb_resources(struct gfar_private *priv)
+{
+ struct gfar_priv_tx_q *tx_queue = NULL;
+ struct gfar_priv_rx_q *rx_queue = NULL;
+ int i;
+
+ /* Go through all the buffer descriptors and free their data buffers */
+ for (i = 0; i < priv->num_tx_queues; i++) {
+ tx_queue = priv->tx_queue[i];
+ if(!tx_queue->tx_skbuff)
+ free_skb_tx_queue(tx_queue);
+ }
- dma_free_coherent(dev, sizeof(*txbdp) * tx_queue->tx_ring_size +
- sizeof(*rxbdp) * rx_queue->rx_ring_size,
- tx_queue->tx_bd_base, tx_queue->tx_bd_dma_base);
+ for (i = 0; i < priv->num_rx_queues; i++) {
+ rx_queue = priv->rx_queue[i];
+ if(!rx_queue->rx_skbuff)
+ free_skb_rx_queue(rx_queue);
+ }
+
+ dma_free_coherent(&priv->ofdev->dev,
+ sizeof(struct txbd8) * priv->total_tx_ring_size +
+ sizeof(struct rxbd8) * priv->total_rx_ring_size,
+ priv->tx_queue[0]->tx_bd_base,
+ priv->tx_queue[0]->tx_bd_dma_base);
}
void gfar_start(struct net_device *dev)
@@ -1233,8 +1436,8 @@ void gfar_start(struct net_device *dev)
gfar_write(®s->dmactrl, tempval);
/* Clear THLT/RHLT, so that the DMA starts polling now */
- gfar_write(®s->tstat, TSTAT_CLEAR_THALT);
- gfar_write(®s->rstat, RSTAT_CLEAR_RHALT);
+ gfar_write(®s->tstat, priv->gfargrp.tstat);
+ gfar_write(®s->rstat, priv->gfargrp.rstat);
/* Unmask the interrupts we look for */
gfar_write(®s->imask, IMASK_DEFAULT);
@@ -1329,7 +1532,7 @@ static int gfar_enet_open(struct net_device *dev)
struct gfar_private *priv = netdev_priv(dev);
int err;
- napi_enable(&priv->rx_queue->napi);
+ napi_enable(&priv->gfargrp.napi);
skb_queue_head_init(&priv->rx_recycle);
@@ -1341,17 +1544,17 @@ static int gfar_enet_open(struct net_device *dev)
err = init_phy(dev);
if (err) {
- napi_disable(&priv->rx_queue->napi);
+ napi_disable(&priv->gfargrp.napi);
return err;
}
err = startup_gfar(dev);
if (err) {
- napi_disable(&priv->rx_queue->napi);
+ napi_disable(&priv->gfargrp.napi);
return err;
}
- netif_start_queue(dev);
+ netif_tx_start_all_queues(dev);
device_set_wakeup_enable(&dev->dev, priv->wol_en);
@@ -1421,16 +1624,20 @@ static int gfar_start_xmit(struct sk_buff *skb, struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
struct gfar_priv_tx_q *tx_queue = NULL;
+ struct netdev_queue *txq;
struct gfar __iomem *regs = NULL;
struct txfcb *fcb = NULL;
struct txbd8 *txbdp, *txbdp_start, *base;
u32 lstatus;
- int i;
+ int i, rq = 0;
u32 bufaddr;
unsigned long flags;
unsigned int nr_frags, length;
- tx_queue = priv->tx_queue;
+
+ rq = skb->queue_mapping;
+ tx_queue = priv->tx_queue[rq];
+ txq = netdev_get_tx_queue(dev, rq);
base = tx_queue->tx_bd_base;
regs = priv->gfargrp.regs;
@@ -1458,7 +1665,7 @@ static int gfar_start_xmit(struct sk_buff *skb, struct net_device *dev)
/* check if there is space to queue this packet */
if ((nr_frags+1) > tx_queue->num_txbdfree) {
/* no space, stop the queue */
- netif_stop_queue(dev);
+ netif_tx_stop_queue(txq);
dev->stats.tx_fifo_errors++;
spin_unlock_irqrestore(&tx_queue->txlock, flags);
return NETDEV_TX_BUSY;
@@ -1550,13 +1757,13 @@ static int gfar_start_xmit(struct sk_buff *skb, struct net_device *dev)
/* If the next BD still needs to be cleaned up, then the bds
are full. We need to tell the kernel to stop sending us stuff. */
if (!tx_queue->num_txbdfree) {
- netif_stop_queue(dev);
+ netif_tx_stop_queue(txq);
dev->stats.tx_fifo_errors++;
}
/* Tell the DMA to go go go */
- gfar_write(®s->tstat, TSTAT_CLEAR_THALT);
+ gfar_write(®s->tstat, TSTAT_CLEAR_THALT >> tx_queue->qindex);
/* Unlock priv */
spin_unlock_irqrestore(&tx_queue->txlock, flags);
@@ -1569,7 +1776,7 @@ static int gfar_close(struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
- napi_disable(&priv->rx_queue->napi);
+ napi_disable(&priv->gfargrp.napi);
skb_queue_purge(&priv->rx_recycle);
cancel_work_sync(&priv->reset_task);
@@ -1579,7 +1786,7 @@ static int gfar_close(struct net_device *dev)
phy_disconnect(priv->phydev);
priv->phydev = NULL;
- netif_stop_queue(dev);
+ netif_tx_stop_all_queues(dev);
return 0;
}
@@ -1598,14 +1805,13 @@ static void gfar_vlan_rx_register(struct net_device *dev,
struct vlan_group *grp)
{
struct gfar_private *priv = netdev_priv(dev);
- struct gfar_priv_rx_q *rx_queue = NULL;
struct gfar __iomem *regs = NULL;
unsigned long flags;
u32 tempval;
- rx_queue = priv->rx_queue;
regs = priv->gfargrp.regs;
- spin_lock_irqsave(&rx_queue->rxlock, flags);
+ local_irq_save(flags);
+ lock_rx_qs(priv);
priv->vlgrp = grp;
@@ -1639,7 +1845,8 @@ static void gfar_vlan_rx_register(struct net_device *dev,
gfar_change_mtu(dev, dev->mtu);
- spin_unlock_irqrestore(&rx_queue->rxlock, flags);
+ unlock_rx_qs(priv);
+ local_irq_restore(flags);
}
static int gfar_change_mtu(struct net_device *dev, int new_mtu)
@@ -1711,10 +1918,10 @@ static void gfar_reset_task(struct work_struct *work)
struct net_device *dev = priv->ndev;
if (dev->flags & IFF_UP) {
- netif_stop_queue(dev);
+ netif_tx_stop_all_queues(dev);
stop_gfar(dev);
startup_gfar(dev);
- netif_start_queue(dev);
+ netif_tx_start_all_queues(dev);
}
netif_tx_schedule_all(dev);
@@ -1745,7 +1952,7 @@ static int gfar_clean_tx_ring(struct gfar_priv_tx_q *tx_queue)
int howmany = 0;
u32 lstatus;
- rx_queue = priv->rx_queue;
+ rx_queue = priv->rx_queue[tx_queue->qindex];
bdp = tx_queue->dirty_tx;
skb_dirtytx = tx_queue->skb_dirtytx;
@@ -1798,8 +2005,8 @@ static int gfar_clean_tx_ring(struct gfar_priv_tx_q *tx_queue)
}
/* If we freed a buffer, we can restart transmission, if necessary */
- if (netif_queue_stopped(dev) && tx_queue->num_txbdfree)
- netif_wake_queue(dev);
+ if (__netif_subqueue_stopped(dev, tx_queue->qindex) && tx_queue->num_txbdfree)
+ netif_wake_subqueue(dev, tx_queue->qindex);
/* Update dirty indicators */
tx_queue->skb_dirtytx = skb_dirtytx;
@@ -1812,19 +2019,12 @@ static int gfar_clean_tx_ring(struct gfar_priv_tx_q *tx_queue)
static void gfar_schedule_cleanup(struct gfar_priv_grp *gfargrp)
{
- struct gfar_private *priv = gfargrp->priv;
- struct gfar_priv_tx_q *tx_queue = NULL;
- struct gfar_priv_rx_q *rx_queue = NULL;
unsigned long flags;
- rx_queue = priv->rx_queue;
- tx_queue = priv->tx_queue;
- spin_lock_irqsave(&tx_queue->txlock, flags);
- spin_lock(&rx_queue->rxlock);
-
- if (napi_schedule_prep(&rx_queue->napi)) {
+ spin_lock_irqsave(&gfargrp->grplock, flags);
+ if (napi_schedule_prep(&gfargrp->napi)) {
gfar_write(&gfargrp->regs->imask, IMASK_RTX_DISABLED);
- __napi_schedule(&rx_queue->napi);
+ __napi_schedule(&gfargrp->napi);
} else {
/*
* Clear IEVENT, so interrupts aren't called again
@@ -1832,9 +2032,8 @@ static void gfar_schedule_cleanup(struct gfar_priv_grp *gfargrp)
*/
gfar_write(&gfargrp->regs->ievent, IEVENT_RTX_MASK);
}
+ spin_unlock_irqrestore(&gfargrp->grplock, flags);
- spin_unlock(&rx_queue->rxlock);
- spin_unlock_irqrestore(&tx_queue->txlock, flags);
}
/* Interrupt Handler for Transmit complete */
@@ -1952,6 +2151,7 @@ static int gfar_process_frame(struct net_device *dev, struct sk_buff *skb,
fcb = (struct rxfcb *)skb->data;
/* Remove the FCB from the skb */
+ skb_set_queue_mapping(skb, fcb->rq);
/* Remove the padded bytes, if there are any */
if (amount_pull)
skb_pull(skb, amount_pull);
@@ -2072,28 +2272,54 @@ int gfar_clean_rx_ring(struct gfar_priv_rx_q *rx_queue, int rx_work_limit)
static int gfar_poll(struct napi_struct *napi, int budget)
{
- struct gfar_priv_rx_q *rx_queue = container_of(napi,
- struct gfar_priv_rx_q, napi);
- struct net_device *dev = rx_queue->dev;
- struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_grp *gfargrp = container_of(napi,
+ struct gfar_priv_grp, napi);
+ struct gfar_private *priv = gfargrp->priv;
struct gfar __iomem *regs = priv->gfargrp.regs;
struct gfar_priv_tx_q *tx_queue = NULL;
- int tx_cleaned = 0;
- int rx_cleaned = 0;
+ struct gfar_priv_rx_q *rx_queue = NULL;
+ int rx_cleaned = 0, budget_per_queue = 0, rx_cleaned_per_queue = 0;
+ int tx_cleaned = 0, i, left_over_budget = budget, serviced_queues = 0;
+ int num_queues = 0;
unsigned long flags;
+ num_queues = gfargrp->num_rx_queues;
+ budget_per_queue = budget/num_queues;
+
/* Clear IEVENT, so interrupts aren't called again
* because of the packets that have already arrived */
gfar_write(®s->ievent, IEVENT_RTX_MASK);
- tx_queue = priv->tx_queue;
- /* If we fail to get the lock, don't bother with the TX BDs */
- if (spin_trylock_irqsave(&tx_queue->txlock, flags)) {
- tx_cleaned = gfar_clean_tx_ring(tx_queue);
- spin_unlock_irqrestore(&tx_queue->txlock, flags);
- }
+ while (num_queues && left_over_budget) {
- rx_cleaned = gfar_clean_rx_ring(rx_queue, budget);
+ budget_per_queue = left_over_budget/num_queues;
+ left_over_budget = 0;
+
+ for_each_bit(i, &gfargrp->rx_bit_map, priv->num_rx_queues) {
+ if (test_bit(i, &serviced_queues))
+ continue;
+ rx_queue = priv->rx_queue[i];
+ tx_queue = priv->tx_queue[rx_queue->qindex];
+
+ /* If we fail to get the lock,
+ * don't bother with the TX BDs */
+ if (spin_trylock_irqsave(&tx_queue->txlock, flags)) {
+ tx_cleaned += gfar_clean_tx_ring(tx_queue);
+ spin_unlock_irqrestore(&tx_queue->txlock,
+ flags);
+ }
+
+ rx_cleaned_per_queue = gfar_clean_rx_ring(rx_queue,
+ budget_per_queue);
+ rx_cleaned += rx_cleaned_per_queue;
+ if(rx_cleaned_per_queue < budget_per_queue) {
+ left_over_budget = left_over_budget +
+ (budget_per_queue - rx_cleaned_per_queue);
+ set_bit(i, &serviced_queues);
+ num_queues--;
+ }
+ }
+ }
if (tx_cleaned)
return budget;
@@ -2102,7 +2328,7 @@ static int gfar_poll(struct napi_struct *napi, int budget)
napi_complete(napi);
/* Clear the halt bit in RSTAT */
- gfar_write(®s->rstat, RSTAT_CLEAR_RHALT);
+ gfar_write(®s->rstat, gfargrp->rstat);
gfar_write(®s->imask, IMASK_DEFAULT);
@@ -2180,14 +2406,14 @@ static irqreturn_t gfar_interrupt(int irq, void *grp_id)
static void adjust_link(struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
- struct gfar_priv_tx_q *tx_queue = NULL;
struct gfar __iomem *regs = priv->gfargrp.regs;
unsigned long flags;
struct phy_device *phydev = priv->phydev;
int new_state = 0;
- tx_queue = priv->tx_queue;
- spin_lock_irqsave(&tx_queue->txlock, flags);
+ local_irq_save(flags);
+ lock_tx_qs(priv);
+
if (phydev->link) {
u32 tempval = gfar_read(®s->maccfg2);
u32 ecntrl = gfar_read(®s->ecntrl);
@@ -2252,8 +2478,8 @@ static void adjust_link(struct net_device *dev)
if (new_state && netif_msg_link(priv))
phy_print_status(phydev);
-
- spin_unlock_irqrestore(&tx_queue->txlock, flags);
+ unlock_tx_qs(priv);
+ local_irq_restore(flags);
}
/* Update the hash table based on the current list of multicast
@@ -2457,7 +2683,7 @@ static irqreturn_t gfar_error(int irq, void *grp_id)
priv->extra_stats.tx_underrun++;
/* Reactivate the Tx Queues */
- gfar_write(®s->tstat, TSTAT_CLEAR_THALT);
+ gfar_write(®s->tstat, gfargrp->tstat);
}
if (netif_msg_tx_err(priv))
printk(KERN_DEBUG "%s: Transmit Error\n", dev->name);
diff --git a/drivers/net/gianfar.h b/drivers/net/gianfar.h
index 79e8471..fc43cfd 100644
--- a/drivers/net/gianfar.h
+++ b/drivers/net/gianfar.h
@@ -75,6 +75,10 @@
extern const char gfar_driver_name[];
extern const char gfar_driver_version[];
+/* MAXIMUM NUMBER OF QUEUES SUPPORTED */
+#define MAX_TX_QS 0x8
+#define MAX_RX_QS 0x8
+
/* These need to be powers of 2 for this driver */
#define DEFAULT_TX_RING_SIZE 256
#define DEFAULT_RX_RING_SIZE 256
@@ -172,12 +176,63 @@ extern const char gfar_driver_version[];
#define MINFLR_INIT_SETTINGS 0x00000040
+/* Tqueue control */
+#define TQUEUE_EN0 0x00008000
+#define TQUEUE_EN1 0x00004000
+#define TQUEUE_EN2 0x00002000
+#define TQUEUE_EN3 0x00001000
+#define TQUEUE_EN4 0x00000800
+#define TQUEUE_EN5 0x00000400
+#define TQUEUE_EN6 0x00000200
+#define TQUEUE_EN7 0x00000100
+#define TQUEUE_EN_ALL 0x0000FF00
+
+#define TR03WT_WT0_MASK 0xFF000000
+#define TR03WT_WT1_MASK 0x00FF0000
+#define TR03WT_WT2_MASK 0x0000FF00
+#define TR03WT_WT3_MASK 0x000000FF
+
+#define TR47WT_WT4_MASK 0xFF000000
+#define TR47WT_WT5_MASK 0x00FF0000
+#define TR47WT_WT6_MASK 0x0000FF00
+#define TR47WT_WT7_MASK 0x000000FF
+
+/* Rqueue control */
+#define RQUEUE_EX0 0x00800000
+#define RQUEUE_EX1 0x00400000
+#define RQUEUE_EX2 0x00200000
+#define RQUEUE_EX3 0x00100000
+#define RQUEUE_EX4 0x00080000
+#define RQUEUE_EX5 0x00040000
+#define RQUEUE_EX6 0x00020000
+#define RQUEUE_EX7 0x00010000
+#define RQUEUE_EX_ALL 0x00FF0000
+
+#define RQUEUE_EN0 0x00000080
+#define RQUEUE_EN1 0x00000040
+#define RQUEUE_EN2 0x00000020
+#define RQUEUE_EN3 0x00000010
+#define RQUEUE_EN4 0x00000008
+#define RQUEUE_EN5 0x00000004
+#define RQUEUE_EN6 0x00000002
+#define RQUEUE_EN7 0x00000001
+#define RQUEUE_EN_ALL 0x000000FF
+
/* Init to do tx snooping for buffers and descriptors */
#define DMACTRL_INIT_SETTINGS 0x000000c3
#define DMACTRL_GRS 0x00000010
#define DMACTRL_GTS 0x00000008
-#define TSTAT_CLEAR_THALT 0x80000000
+#define TSTAT_CLEAR_THALT_ALL 0xFF000000
+#define TSTAT_CLEAR_THALT 0x80000000
+#define TSTAT_CLEAR_THALT0 0x80000000
+#define TSTAT_CLEAR_THALT1 0x40000000
+#define TSTAT_CLEAR_THALT2 0x20000000
+#define TSTAT_CLEAR_THALT3 0x10000000
+#define TSTAT_CLEAR_THALT4 0x08000000
+#define TSTAT_CLEAR_THALT5 0x04000000
+#define TSTAT_CLEAR_THALT6 0x02000000
+#define TSTAT_CLEAR_THALT7 0x01000000
/* Interrupt coalescing macros */
#define IC_ICEN 0x80000000
@@ -228,6 +283,13 @@ extern const char gfar_driver_version[];
#define TCTRL_IPCSEN 0x00004000
#define TCTRL_TUCSEN 0x00002000
#define TCTRL_VLINS 0x00001000
+#define TCTRL_THDF 0x00000800
+#define TCTRL_RFCPAUSE 0x00000010
+#define TCTRL_TFCPAUSE 0x00000008
+#define TCTRL_TXSCHED_MASK 0x00000006
+#define TCTRL_TXSCHED_INIT 0x00000000
+#define TCTRL_TXSCHED_PRIO 0x00000002
+#define TCTRL_TXSCHED_WRRS 0x00000004
#define TCTRL_INIT_CSUM (TCTRL_TUCSEN | TCTRL_IPCSEN)
#define IEVENT_INIT_CLEAR 0xffffffff
@@ -700,6 +762,8 @@ struct gfar {
#define FSL_GIANFAR_DEV_HAS_BD_STASHING 0x00000200
#define FSL_GIANFAR_DEV_HAS_BUF_STASHING 0x00000400
+#define DEFAULT_MAPPING 0xFF
+
/**
* struct gfar_priv_tx_q - per tx queue structure
* @txlock: per queue tx spin lock
@@ -743,7 +807,6 @@ struct gfar_priv_tx_q {
/**
* struct gfar_priv_rx_q - per rx queue structure
* @rxlock: per queue rx spin lock
- * @napi: the napi poll function
* @rx_skbuff: skb pointers
* @skb_currx: currently use skb pointer
* @rx_bd_base: First rx buffer descriptor
@@ -757,7 +820,6 @@ struct gfar_priv_tx_q {
struct gfar_priv_rx_q {
spinlock_t rxlock __attribute__ ((aligned (SMP_CACHE_BYTES)));
- struct napi_struct napi;
struct sk_buff ** rx_skbuff;
struct rxbd8 *rx_bd_base;
struct rxbd8 *cur_rx;
@@ -772,6 +834,7 @@ struct gfar_priv_rx_q {
/**
* struct gfar_priv_grp - per group structure
+ * @napi: the napi poll function
* @priv: back pointer to the priv structure
* @regs: the ioremapped register space for this group
* @grp_id: group id for this group
@@ -785,8 +848,17 @@ struct gfar_priv_rx_q {
struct gfar_priv_grp {
spinlock_t grplock __attribute__ ((aligned (SMP_CACHE_BYTES)));
+ struct napi_struct napi;
struct gfar_private *priv;
struct gfar __iomem *regs;
+ unsigned int rx_bit_map;
+ unsigned int tx_bit_map;
+ unsigned int num_tx_queues;
+ unsigned int num_rx_queues;
+ unsigned int rstat;
+ unsigned int tstat;
+ unsigned int imask;
+ unsigned int ievent;
unsigned int interruptTransmit;
unsigned int interruptReceive;
unsigned int interruptError;
@@ -807,13 +879,21 @@ struct gfar_priv_grp {
*/
struct gfar_private {
+ /* Indicates how many tx, rx queues are enabled */
+ unsigned int num_tx_queues;
+ unsigned int num_rx_queues;
+
+ /* The total tx and rx ring size for the enabled queues */
+ unsigned int total_tx_ring_size;
+ unsigned int total_rx_ring_size;
+
struct device_node *node;
struct net_device *ndev;
struct of_device *ofdev;
struct gfar_priv_grp gfargrp;
- struct gfar_priv_tx_q *tx_queue;
- struct gfar_priv_rx_q *rx_queue;
+ struct gfar_priv_tx_q *tx_queue[MAX_TX_QS];
+ struct gfar_priv_rx_q *rx_queue[MAX_RX_QS];
/* RX per device parameters */
unsigned int rx_buffer_size;
@@ -844,6 +924,7 @@ struct gfar_private {
unsigned char rx_csum_enable:1,
extended_hash:1,
bd_stash_en:1,
+ rx_filer_enable:1,
wol_en:1; /* Wake-on-LAN enabled */
unsigned short padding;
@@ -874,6 +955,10 @@ static inline void gfar_write(volatile unsigned __iomem *addr, u32 val)
out_be32(addr, val);
}
+extern inline void lock_rx_qs(struct gfar_private *priv);
+extern inline void lock_tx_qs(struct gfar_private *priv);
+extern inline void unlock_rx_qs(struct gfar_private *priv);
+extern inline void unlock_tx_qs(struct gfar_private *priv);
extern irqreturn_t gfar_receive(int irq, void *dev_id);
extern int startup_gfar(struct net_device *dev);
extern void stop_gfar(struct net_device *dev);
diff --git a/drivers/net/gianfar_ethtool.c b/drivers/net/gianfar_ethtool.c
index c681b41..d3d2623 100644
--- a/drivers/net/gianfar_ethtool.c
+++ b/drivers/net/gianfar_ethtool.c
@@ -204,9 +204,11 @@ static int gfar_gsettings(struct net_device *dev, struct ethtool_cmd *cmd)
if (NULL == phydev)
return -ENODEV;
- tx_queue = priv->tx_queue;
- rx_queue = priv->rx_queue;
+ tx_queue = priv->tx_queue[0];
+ rx_queue = priv->rx_queue[0];
+ /* etsec-1.7 and older versions have only one txic
+ * and rxic regs although they support multiple queues */
cmd->maxtxpkt = get_icft_value(tx_queue->txic);
cmd->maxrxpkt = get_icft_value(rx_queue->rxic);
@@ -298,8 +300,8 @@ static int gfar_gcoalesce(struct net_device *dev, struct ethtool_coalesce *cvals
if (NULL == priv->phydev)
return -ENODEV;
- rx_queue = priv->rx_queue;
- tx_queue = priv->tx_queue;
+ rx_queue = priv->rx_queue[0];
+ tx_queue = priv->tx_queue[0];
rxtime = get_ictt_value(rx_queue->rxic);
rxcount = get_icft_value(rx_queue->rxic);
@@ -357,8 +359,8 @@ static int gfar_scoalesce(struct net_device *dev, struct ethtool_coalesce *cvals
if (!(priv->device_flags & FSL_GIANFAR_DEV_HAS_COALESCE))
return -EOPNOTSUPP;
- tx_queue = priv->tx_queue;
- rx_queue = priv->rx_queue;
+ tx_queue = priv->tx_queue[0];
+ rx_queue = priv->rx_queue[0];
/* Set up rx coalescing */
if ((cvals->rx_coalesce_usecs == 0) ||
@@ -429,8 +431,8 @@ static void gfar_gringparam(struct net_device *dev, struct ethtool_ringparam *rv
struct gfar_priv_tx_q *tx_queue = NULL;
struct gfar_priv_rx_q *rx_queue = NULL;
- tx_queue = priv->tx_queue;
- rx_queue = priv->rx_queue;
+ tx_queue = priv->tx_queue[0];
+ rx_queue = priv->rx_queue[0];
rvals->rx_max_pending = GFAR_RX_MAX_RING_SIZE;
rvals->rx_mini_max_pending = GFAR_RX_MAX_RING_SIZE;
@@ -453,9 +455,7 @@ static void gfar_gringparam(struct net_device *dev, struct ethtool_ringparam *rv
static int gfar_sringparam(struct net_device *dev, struct ethtool_ringparam *rvals)
{
struct gfar_private *priv = netdev_priv(dev);
- struct gfar_priv_tx_q *tx_queue = NULL;
- struct gfar_priv_rx_q *rx_queue = NULL;
- int err = 0;
+ int err = 0, i = 0;
if (rvals->rx_pending > GFAR_RX_MAX_RING_SIZE)
return -EINVAL;
@@ -475,37 +475,41 @@ static int gfar_sringparam(struct net_device *dev, struct ethtool_ringparam *rva
return -EINVAL;
}
- tx_queue = priv->tx_queue;
- rx_queue = priv->rx_queue;
if (dev->flags & IFF_UP) {
unsigned long flags;
/* Halt TX and RX, and process the frames which
* have already been received */
- spin_lock_irqsave(&tx_queue->txlock, flags);
- spin_lock(&rx_queue->rxlock);
+ local_irq_save(flags);
+ lock_tx_qs(priv);
+ lock_rx_qs(priv);
gfar_halt(dev);
- spin_unlock(&rx_queue->rxlock);
- spin_unlock_irqrestore(&tx_queue->txlock, flags);
+ unlock_rx_qs(priv);
+ unlock_tx_qs(priv);
+ local_irq_restore(flags);
- gfar_clean_rx_ring(rx_queue, rx_queue->rx_ring_size);
+ for (i = 0; i < priv->num_rx_queues; i++)
+ gfar_clean_rx_ring(priv->rx_queue[i],
+ priv->rx_queue[i]->rx_ring_size);
/* Now we take down the rings to rebuild them */
stop_gfar(dev);
}
/* Change the size */
- rx_queue->rx_ring_size = rvals->rx_pending;
- tx_queue->tx_ring_size = rvals->tx_pending;
- tx_queue->num_txbdfree = tx_queue->tx_ring_size;
+ for (i = 0; i < priv->num_rx_queues; i++) {
+ priv->rx_queue[i]->rx_ring_size = rvals->rx_pending;
+ priv->tx_queue[i]->tx_ring_size = rvals->tx_pending;
+ priv->tx_queue[i]->num_txbdfree = priv->tx_queue[i]->tx_ring_size;
+ }
/* Rebuild the rings with the new size */
if (dev->flags & IFF_UP) {
err = startup_gfar(dev);
- netif_wake_queue(dev);
+ netif_tx_wake_all_queues(dev);
}
return err;
}
@@ -513,29 +517,29 @@ static int gfar_sringparam(struct net_device *dev, struct ethtool_ringparam *rva
static int gfar_set_rx_csum(struct net_device *dev, uint32_t data)
{
struct gfar_private *priv = netdev_priv(dev);
- struct gfar_priv_rx_q *rx_queue = NULL;
- struct gfar_priv_tx_q *tx_queue = NULL;
unsigned long flags;
- int err = 0;
+ int err = 0, i = 0;
if (!(priv->device_flags & FSL_GIANFAR_DEV_HAS_CSUM))
return -EOPNOTSUPP;
- tx_queue = priv->tx_queue;
- rx_queue = priv->rx_queue;
if (dev->flags & IFF_UP) {
/* Halt TX and RX, and process the frames which
* have already been received */
- spin_lock_irqsave(&tx_queue->txlock, flags);
- spin_lock(&rx_queue->rxlock);
+ local_irq_save(flags);
+ lock_tx_qs(priv);
+ lock_rx_qs(priv);
gfar_halt(dev);
- spin_unlock(&rx_queue->rxlock);
- spin_unlock_irqrestore(&tx_queue->txlock, flags);
+ unlock_tx_qs(priv);
+ unlock_rx_qs(priv);
+ local_irq_save(flags);
- gfar_clean_rx_ring(rx_queue, rx_queue->rx_ring_size);
+ for (i = 0; i < priv->num_rx_queues; i++)
+ gfar_clean_rx_ring(priv->rx_queue[i],
+ priv->rx_queue[i]->rx_ring_size);
/* Now we take down the rings to rebuild them */
stop_gfar(dev);
@@ -547,7 +551,7 @@ static int gfar_set_rx_csum(struct net_device *dev, uint32_t data)
if (dev->flags & IFF_UP) {
err = startup_gfar(dev);
- netif_wake_queue(dev);
+ netif_tx_wake_all_queues(dev);
}
return err;
}
diff --git a/drivers/net/gianfar_sysfs.c b/drivers/net/gianfar_sysfs.c
index adea11e..4b726f6 100644
--- a/drivers/net/gianfar_sysfs.c
+++ b/drivers/net/gianfar_sysfs.c
@@ -51,7 +51,6 @@ static ssize_t gfar_set_bd_stash(struct device *dev,
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
struct gfar __iomem *regs = priv->gfargrp.regs;
- struct gfar_priv_rx_q *rx_queue = NULL;
int new_setting = 0;
u32 temp;
unsigned long flags;
@@ -59,7 +58,6 @@ static ssize_t gfar_set_bd_stash(struct device *dev,
if (!(priv->device_flags & FSL_GIANFAR_DEV_HAS_BD_STASHING))
return count;
- rx_queue = priv->rx_queue;
/* Find out the new setting */
if (!strncmp("on", buf, count - 1) || !strncmp("1", buf, count - 1))
@@ -70,7 +68,9 @@ static ssize_t gfar_set_bd_stash(struct device *dev,
else
return count;
- spin_lock_irqsave(&rx_queue->rxlock, flags);
+
+ local_irq_save(flags);
+ lock_rx_qs(priv);
/* Set the new stashing value */
priv->bd_stash_en = new_setting;
@@ -84,7 +84,8 @@ static ssize_t gfar_set_bd_stash(struct device *dev,
gfar_write(®s->attr, temp);
- spin_unlock_irqrestore(&rx_queue->rxlock, flags);
+ unlock_rx_qs(priv);
+ local_irq_restore(flags);
return count;
}
@@ -105,7 +106,6 @@ static ssize_t gfar_set_rx_stash_size(struct device *dev,
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
struct gfar __iomem *regs = priv->gfargrp.regs;
- struct gfar_priv_rx_q *rx_queue = NULL;
unsigned int length = simple_strtoul(buf, NULL, 0);
u32 temp;
unsigned long flags;
@@ -113,9 +113,9 @@ static ssize_t gfar_set_rx_stash_size(struct device *dev,
if (!(priv->device_flags & FSL_GIANFAR_DEV_HAS_BUF_STASHING))
return count;
- rx_queue = priv->rx_queue;
+ local_irq_save(flags);
+ lock_rx_qs(priv);
- spin_lock_irqsave(&rx_queue->rxlock, flags);
if (length > priv->rx_buffer_size)
goto out;
@@ -140,7 +140,8 @@ static ssize_t gfar_set_rx_stash_size(struct device *dev,
gfar_write(®s->attr, temp);
out:
- spin_unlock_irqrestore(&rx_queue->rxlock, flags);
+ unlock_rx_qs(priv);
+ local_irq_restore(flags);
return count;
}
@@ -164,7 +165,6 @@ static ssize_t gfar_set_rx_stash_index(struct device *dev,
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
struct gfar __iomem *regs = priv->gfargrp.regs;
- struct gfar_priv_rx_q *rx_queue = NULL;
unsigned short index = simple_strtoul(buf, NULL, 0);
u32 temp;
unsigned long flags;
@@ -172,9 +172,9 @@ static ssize_t gfar_set_rx_stash_index(struct device *dev,
if (!(priv->device_flags & FSL_GIANFAR_DEV_HAS_BUF_STASHING))
return count;
- rx_queue = priv->rx_queue;
+ local_irq_save(flags);
+ lock_rx_qs(priv);
- spin_lock_irqsave(&rx_queue->rxlock, flags);
if (index > priv->rx_stash_size)
goto out;
@@ -189,7 +189,8 @@ static ssize_t gfar_set_rx_stash_index(struct device *dev,
gfar_write(®s->attreli, flags);
out:
- spin_unlock_irqrestore(&rx_queue->rxlock, flags);
+ unlock_rx_qs(priv);
+ local_irq_restore(flags);
return count;
}
@@ -212,7 +213,6 @@ static ssize_t gfar_set_fifo_threshold(struct device *dev,
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
struct gfar __iomem *regs = priv->gfargrp.regs;
- struct gfar_priv_tx_q *tx_queue = NULL;
unsigned int length = simple_strtoul(buf, NULL, 0);
u32 temp;
unsigned long flags;
@@ -220,9 +220,8 @@ static ssize_t gfar_set_fifo_threshold(struct device *dev,
if (length > GFAR_MAX_FIFO_THRESHOLD)
return count;
- tx_queue = priv->tx_queue;
-
- spin_lock_irqsave(&tx_queue->txlock, flags);
+ local_irq_save(flags);
+ lock_tx_qs(priv);
priv->fifo_threshold = length;
@@ -231,7 +230,8 @@ static ssize_t gfar_set_fifo_threshold(struct device *dev,
temp |= length;
gfar_write(®s->fifo_tx_thr, temp);
- spin_unlock_irqrestore(&tx_queue->txlock, flags);
+ unlock_tx_qs(priv);
+ local_irq_restore(flags);
return count;
}
@@ -253,7 +253,6 @@ static ssize_t gfar_set_fifo_starve(struct device *dev,
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
struct gfar __iomem *regs = priv->gfargrp.regs;
- struct gfar_priv_tx_q *tx_queue = NULL;
unsigned int num = simple_strtoul(buf, NULL, 0);
u32 temp;
unsigned long flags;
@@ -261,8 +260,8 @@ static ssize_t gfar_set_fifo_starve(struct device *dev,
if (num > GFAR_MAX_FIFO_STARVE)
return count;
- tx_queue = priv->tx_queue;
- spin_lock_irqsave(&tx_queue->txlock, flags);
+ local_irq_save(flags);
+ lock_tx_qs(priv);
priv->fifo_starve = num;
@@ -271,7 +270,8 @@ static ssize_t gfar_set_fifo_starve(struct device *dev,
temp |= num;
gfar_write(®s->fifo_tx_starve, temp);
- spin_unlock_irqrestore(&tx_queue->txlock, flags);
+ unlock_tx_qs(priv);
+ local_irq_restore(flags);
return count;
}
@@ -294,7 +294,6 @@ static ssize_t gfar_set_fifo_starve_off(struct device *dev,
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
struct gfar __iomem *regs = priv->gfargrp.regs;
- struct gfar_priv_tx_q *tx_queue = NULL;
unsigned int num = simple_strtoul(buf, NULL, 0);
u32 temp;
unsigned long flags;
@@ -302,8 +301,8 @@ static ssize_t gfar_set_fifo_starve_off(struct device *dev,
if (num > GFAR_MAX_FIFO_STARVE_OFF)
return count;
- tx_queue = priv->tx_queue;
- spin_lock_irqsave(&tx_queue->txlock, flags);
+ local_irq_save(flags);
+ lock_tx_qs(priv);
priv->fifo_starve_off = num;
@@ -312,7 +311,8 @@ static ssize_t gfar_set_fifo_starve_off(struct device *dev,
temp |= num;
gfar_write(®s->fifo_tx_starve_shutoff, temp);
- spin_unlock_irqrestore(&tx_queue->txlock, flags);
+ unlock_tx_qs(priv);
+ local_irq_restore(flags);
return count;
}
--
1.5.2.2
^ permalink raw reply related
* [PATCH v1 2/7] gianfar: Introduce logical group support.
From: Sandeep Gopalpet @ 2009-10-26 16:27 UTC (permalink / raw)
To: netdev; +Cc: Sandeep Gopalpet
This patch introduces the group structure. The elements of this
structure are the interrupt lines, their corresponding names,
the register memory map.
The elements for this group are factored out from the gfar_private
structure. The introduction of group structure will help in
providing support for newer versions of etsec.
Currently, the support is present only for single group and
single tx/rx queues.
Signed-off-by: Sandeep Gopalpet <sandeep.kumar@freescale.com>
---
drivers/net/gianfar.c | 364 ++++++++++++++++++++++-------------------
drivers/net/gianfar.h | 38 +++--
drivers/net/gianfar_ethtool.c | 14 +-
drivers/net/gianfar_sysfs.c | 34 +++--
4 files changed, 250 insertions(+), 200 deletions(-)
diff --git a/drivers/net/gianfar.c b/drivers/net/gianfar.c
index 354b2b5..fa0188e 100644
--- a/drivers/net/gianfar.c
+++ b/drivers/net/gianfar.c
@@ -151,7 +151,6 @@ MODULE_LICENSE("GPL");
static void gfar_init_rxbdp(struct gfar_priv_rx_q *rx_queue, struct rxbd8 *bdp,
dma_addr_t buf)
{
- struct net_device *dev = rx_queue->dev;
u32 lstatus;
bdp->bufPtr = buf;
@@ -290,9 +289,9 @@ cleanup:
static void gfar_init_mac(struct net_device *ndev)
{
struct gfar_private *priv = netdev_priv(ndev);
- struct gfar __iomem *regs = priv->regs;
struct gfar_priv_tx_q *tx_queue = NULL;
struct gfar_priv_rx_q *rx_queue = NULL;
+ struct gfar __iomem *regs = priv->gfargrp.regs;
u32 rctrl = 0;
u32 tctrl = 0;
u32 attrs = 0;
@@ -407,24 +406,25 @@ static int gfar_of_init(struct net_device *dev)
/* get a pointer to the register memory */
addr = of_translate_address(np, of_get_address(np, 0, &size, NULL));
- priv->regs = ioremap(addr, size);
+ priv->gfargrp.regs = ioremap(addr, size);
- if (priv->regs == NULL)
+ if (priv->gfargrp.regs == NULL)
return -ENOMEM;
- priv->interruptTransmit = irq_of_parse_and_map(np, 0);
+ priv->gfargrp.priv = priv; /* back pointer from group to priv */
+ priv->gfargrp.interruptTransmit = irq_of_parse_and_map(np, 0);
model = of_get_property(np, "model", NULL);
/* If we aren't the FEC we have multiple interrupts */
if (model && strcasecmp(model, "FEC")) {
- priv->interruptReceive = irq_of_parse_and_map(np, 1);
+ priv->gfargrp.interruptReceive = irq_of_parse_and_map(np, 1);
- priv->interruptError = irq_of_parse_and_map(np, 2);
+ priv->gfargrp.interruptError = irq_of_parse_and_map(np, 2);
- if (priv->interruptTransmit < 0 ||
- priv->interruptReceive < 0 ||
- priv->interruptError < 0) {
+ if (priv->gfargrp.interruptTransmit < 0 ||
+ priv->gfargrp.interruptReceive < 0 ||
+ priv->gfargrp.interruptError < 0) {
err = -EINVAL;
goto err_out;
}
@@ -491,7 +491,7 @@ static int gfar_of_init(struct net_device *dev)
return 0;
err_out:
- iounmap(priv->regs);
+ iounmap(priv->gfargrp.regs);
return err;
}
@@ -517,6 +517,7 @@ static int gfar_probe(struct of_device *ofdev,
u32 tempval;
struct net_device *dev = NULL;
struct gfar_private *priv = NULL;
+ struct gfar __iomem *regs = NULL;
int err = 0;
int len_devname;
@@ -549,32 +550,34 @@ static int gfar_probe(struct of_device *ofdev,
spin_lock_init(&priv->tx_queue->txlock);
spin_lock_init(&priv->rx_queue->rxlock);
+ spin_lock_init(&priv->gfargrp.grplock);
spin_lock_init(&priv->bflock);
INIT_WORK(&priv->reset_task, gfar_reset_task);
dev_set_drvdata(&ofdev->dev, priv);
+ regs = priv->gfargrp.regs;
/* Stop the DMA engine now, in case it was running before */
/* (The firmware could have used it, and left it running). */
gfar_halt(dev);
/* Reset MAC layer */
- gfar_write(&priv->regs->maccfg1, MACCFG1_SOFT_RESET);
+ gfar_write(®s->maccfg1, MACCFG1_SOFT_RESET);
/* We need to delay at least 3 TX clocks */
udelay(2);
tempval = (MACCFG1_TX_FLOW | MACCFG1_RX_FLOW);
- gfar_write(&priv->regs->maccfg1, tempval);
+ gfar_write(®s->maccfg1, tempval);
/* Initialize MACCFG2. */
- gfar_write(&priv->regs->maccfg2, MACCFG2_INIT_SETTINGS);
+ gfar_write(®s->maccfg2, MACCFG2_INIT_SETTINGS);
/* Initialize ECNTRL */
- gfar_write(&priv->regs->ecntrl, ECNTRL_INIT_SETTINGS);
+ gfar_write(®s->ecntrl, ECNTRL_INIT_SETTINGS);
/* Set the dev->base_addr to the gfar reg region */
- dev->base_addr = (unsigned long) (priv->regs);
+ dev->base_addr = (unsigned long) regs;
SET_NETDEV_DEV(dev, &ofdev->dev);
@@ -602,35 +605,35 @@ static int gfar_probe(struct of_device *ofdev,
priv->extended_hash = 1;
priv->hash_width = 9;
- priv->hash_regs[0] = &priv->regs->igaddr0;
- priv->hash_regs[1] = &priv->regs->igaddr1;
- priv->hash_regs[2] = &priv->regs->igaddr2;
- priv->hash_regs[3] = &priv->regs->igaddr3;
- priv->hash_regs[4] = &priv->regs->igaddr4;
- priv->hash_regs[5] = &priv->regs->igaddr5;
- priv->hash_regs[6] = &priv->regs->igaddr6;
- priv->hash_regs[7] = &priv->regs->igaddr7;
- priv->hash_regs[8] = &priv->regs->gaddr0;
- priv->hash_regs[9] = &priv->regs->gaddr1;
- priv->hash_regs[10] = &priv->regs->gaddr2;
- priv->hash_regs[11] = &priv->regs->gaddr3;
- priv->hash_regs[12] = &priv->regs->gaddr4;
- priv->hash_regs[13] = &priv->regs->gaddr5;
- priv->hash_regs[14] = &priv->regs->gaddr6;
- priv->hash_regs[15] = &priv->regs->gaddr7;
+ priv->hash_regs[0] = ®s->igaddr0;
+ priv->hash_regs[1] = ®s->igaddr1;
+ priv->hash_regs[2] = ®s->igaddr2;
+ priv->hash_regs[3] = ®s->igaddr3;
+ priv->hash_regs[4] = ®s->igaddr4;
+ priv->hash_regs[5] = ®s->igaddr5;
+ priv->hash_regs[6] = ®s->igaddr6;
+ priv->hash_regs[7] = ®s->igaddr7;
+ priv->hash_regs[8] = ®s->gaddr0;
+ priv->hash_regs[9] = ®s->gaddr1;
+ priv->hash_regs[10] = ®s->gaddr2;
+ priv->hash_regs[11] = ®s->gaddr3;
+ priv->hash_regs[12] = ®s->gaddr4;
+ priv->hash_regs[13] = ®s->gaddr5;
+ priv->hash_regs[14] = ®s->gaddr6;
+ priv->hash_regs[15] = ®s->gaddr7;
} else {
priv->extended_hash = 0;
priv->hash_width = 8;
- priv->hash_regs[0] = &priv->regs->gaddr0;
- priv->hash_regs[1] = &priv->regs->gaddr1;
- priv->hash_regs[2] = &priv->regs->gaddr2;
- priv->hash_regs[3] = &priv->regs->gaddr3;
- priv->hash_regs[4] = &priv->regs->gaddr4;
- priv->hash_regs[5] = &priv->regs->gaddr5;
- priv->hash_regs[6] = &priv->regs->gaddr6;
- priv->hash_regs[7] = &priv->regs->gaddr7;
+ priv->hash_regs[0] = ®s->gaddr0;
+ priv->hash_regs[1] = ®s->gaddr1;
+ priv->hash_regs[2] = ®s->gaddr2;
+ priv->hash_regs[3] = ®s->gaddr3;
+ priv->hash_regs[4] = ®s->gaddr4;
+ priv->hash_regs[5] = ®s->gaddr5;
+ priv->hash_regs[6] = ®s->gaddr6;
+ priv->hash_regs[7] = ®s->gaddr7;
}
if (priv->device_flags & FSL_GIANFAR_DEV_HAS_PADDING)
@@ -672,20 +675,20 @@ static int gfar_probe(struct of_device *ofdev,
/* fill out IRQ number and name fields */
len_devname = strlen(dev->name);
- strncpy(&priv->int_name_tx[0], dev->name, len_devname);
+ strncpy(&priv->gfargrp.int_name_tx[0], dev->name, len_devname);
if (priv->device_flags & FSL_GIANFAR_DEV_HAS_MULTI_INTR) {
- strncpy(&priv->int_name_tx[len_devname],
+ strncpy(&priv->gfargrp.int_name_tx[len_devname],
"_tx", sizeof("_tx") + 1);
- strncpy(&priv->int_name_rx[0], dev->name, len_devname);
- strncpy(&priv->int_name_rx[len_devname],
+ strncpy(&priv->gfargrp.int_name_rx[0], dev->name, len_devname);
+ strncpy(&priv->gfargrp.int_name_rx[len_devname],
"_rx", sizeof("_rx") + 1);
- strncpy(&priv->int_name_er[0], dev->name, len_devname);
- strncpy(&priv->int_name_er[len_devname],
+ strncpy(&priv->gfargrp.int_name_er[0], dev->name, len_devname);
+ strncpy(&priv->gfargrp.int_name_er[len_devname],
"_er", sizeof("_er") + 1);
} else
- priv->int_name_tx[len_devname] = '\0';
+ priv->gfargrp.int_name_tx[len_devname] = '\0';
/* Create all the sysfs files */
gfar_init_sysfs(dev);
@@ -702,7 +705,7 @@ static int gfar_probe(struct of_device *ofdev,
return 0;
register_fail:
- iounmap(priv->regs);
+ iounmap(priv->gfargrp.regs);
kfree(priv->rx_queue);
rx_queue_fail:
kfree(priv->tx_queue);
@@ -727,7 +730,7 @@ static int gfar_remove(struct of_device *ofdev)
dev_set_drvdata(&ofdev->dev, NULL);
unregister_netdev(priv->ndev);
- iounmap(priv->regs);
+ iounmap(priv->gfargrp.regs);
free_netdev(priv->ndev);
return 0;
@@ -741,6 +744,7 @@ static int gfar_suspend(struct device *dev)
struct net_device *ndev = priv->ndev;
struct gfar_priv_tx_q *tx_queue = NULL;
struct gfar_priv_rx_q *rx_queue = NULL;
+ struct gfar __iomem *regs = NULL;
unsigned long flags;
u32 tempval;
@@ -750,6 +754,7 @@ static int gfar_suspend(struct device *dev)
netif_device_detach(ndev);
tx_queue = priv->tx_queue;
rx_queue = priv->rx_queue;
+ regs = priv->gfargrp.regs;
if (netif_running(ndev)) {
spin_lock_irqsave(&tx_queue->txlock, flags);
@@ -758,14 +763,14 @@ static int gfar_suspend(struct device *dev)
gfar_halt_nodisable(ndev);
/* Disable Tx, and Rx if wake-on-LAN is disabled. */
- tempval = gfar_read(&priv->regs->maccfg1);
+ tempval = gfar_read(®s->maccfg1);
tempval &= ~MACCFG1_TX_EN;
if (!magic_packet)
tempval &= ~MACCFG1_RX_EN;
- gfar_write(&priv->regs->maccfg1, tempval);
+ gfar_write(®s->maccfg1, tempval);
spin_unlock(&rx_queue->rxlock);
spin_unlock_irqrestore(&tx_queue->txlock, flags);
@@ -774,12 +779,12 @@ static int gfar_suspend(struct device *dev)
if (magic_packet) {
/* Enable interrupt on Magic Packet */
- gfar_write(&priv->regs->imask, IMASK_MAG);
+ gfar_write(®s->imask, IMASK_MAG);
/* Enable Magic Packet mode */
- tempval = gfar_read(&priv->regs->maccfg2);
+ tempval = gfar_read(®s->maccfg2);
tempval |= MACCFG2_MPEN;
- gfar_write(&priv->regs->maccfg2, tempval);
+ gfar_write(®s->maccfg2, tempval);
} else {
phy_stop(priv->phydev);
}
@@ -794,6 +799,7 @@ static int gfar_resume(struct device *dev)
struct net_device *ndev = priv->ndev;
struct gfar_priv_tx_q *tx_queue = NULL;
struct gfar_priv_rx_q *rx_queue = NULL;
+ struct gfar __iomem *regs = NULL;
unsigned long flags;
u32 tempval;
int magic_packet = priv->wol_en &&
@@ -812,13 +818,14 @@ static int gfar_resume(struct device *dev)
*/
rx_queue = priv->rx_queue;
tx_queue = priv->tx_queue;
+ regs = priv->gfargrp.regs;
spin_lock_irqsave(&tx_queue->txlock, flags);
spin_lock(&rx_queue->rxlock);
- tempval = gfar_read(&priv->regs->maccfg2);
+ tempval = gfar_read(®s->maccfg2);
tempval &= ~MACCFG2_MPEN;
- gfar_write(&priv->regs->maccfg2, tempval);
+ gfar_write(®s->maccfg2, tempval);
gfar_start(ndev);
@@ -893,7 +900,11 @@ static int gfar_legacy_resume(struct of_device *ofdev)
static phy_interface_t gfar_get_interface(struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
- u32 ecntrl = gfar_read(&priv->regs->ecntrl);
+ struct gfar __iomem *regs = NULL;
+ u32 ecntrl;
+
+ regs = priv->gfargrp.regs;
+ ecntrl = gfar_read(®s->ecntrl);
if (ecntrl & ECNTRL_SGMII_MODE)
return PHY_INTERFACE_MODE_SGMII;
@@ -1015,46 +1026,48 @@ static void gfar_configure_serdes(struct net_device *dev)
static void init_registers(struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar __iomem *regs = NULL;
+ regs = priv->gfargrp.regs;
/* Clear IEVENT */
- gfar_write(&priv->regs->ievent, IEVENT_INIT_CLEAR);
+ gfar_write(®s->ievent, IEVENT_INIT_CLEAR);
/* Initialize IMASK */
- gfar_write(&priv->regs->imask, IMASK_INIT_CLEAR);
+ gfar_write(®s->imask, IMASK_INIT_CLEAR);
/* Init hash registers to zero */
- gfar_write(&priv->regs->igaddr0, 0);
- gfar_write(&priv->regs->igaddr1, 0);
- gfar_write(&priv->regs->igaddr2, 0);
- gfar_write(&priv->regs->igaddr3, 0);
- gfar_write(&priv->regs->igaddr4, 0);
- gfar_write(&priv->regs->igaddr5, 0);
- gfar_write(&priv->regs->igaddr6, 0);
- gfar_write(&priv->regs->igaddr7, 0);
-
- gfar_write(&priv->regs->gaddr0, 0);
- gfar_write(&priv->regs->gaddr1, 0);
- gfar_write(&priv->regs->gaddr2, 0);
- gfar_write(&priv->regs->gaddr3, 0);
- gfar_write(&priv->regs->gaddr4, 0);
- gfar_write(&priv->regs->gaddr5, 0);
- gfar_write(&priv->regs->gaddr6, 0);
- gfar_write(&priv->regs->gaddr7, 0);
+ gfar_write(®s->igaddr0, 0);
+ gfar_write(®s->igaddr1, 0);
+ gfar_write(®s->igaddr2, 0);
+ gfar_write(®s->igaddr3, 0);
+ gfar_write(®s->igaddr4, 0);
+ gfar_write(®s->igaddr5, 0);
+ gfar_write(®s->igaddr6, 0);
+ gfar_write(®s->igaddr7, 0);
+
+ gfar_write(®s->gaddr0, 0);
+ gfar_write(®s->gaddr1, 0);
+ gfar_write(®s->gaddr2, 0);
+ gfar_write(®s->gaddr3, 0);
+ gfar_write(®s->gaddr4, 0);
+ gfar_write(®s->gaddr5, 0);
+ gfar_write(®s->gaddr6, 0);
+ gfar_write(®s->gaddr7, 0);
/* Zero out the rmon mib registers if it has them */
if (priv->device_flags & FSL_GIANFAR_DEV_HAS_RMON) {
- memset_io(&(priv->regs->rmon), 0, sizeof (struct rmon_mib));
+ memset_io(&(regs->rmon), 0, sizeof (struct rmon_mib));
/* Mask off the CAM interrupts */
- gfar_write(&priv->regs->rmon.cam1, 0xffffffff);
- gfar_write(&priv->regs->rmon.cam2, 0xffffffff);
+ gfar_write(®s->rmon.cam1, 0xffffffff);
+ gfar_write(®s->rmon.cam2, 0xffffffff);
}
/* Initialize the max receive buffer length */
- gfar_write(&priv->regs->mrblr, priv->rx_buffer_size);
+ gfar_write(®s->mrblr, priv->rx_buffer_size);
/* Initialize the Minimum Frame Length Register */
- gfar_write(&priv->regs->minflr, MINFLR_INIT_SETTINGS);
+ gfar_write(®s->minflr, MINFLR_INIT_SETTINGS);
}
@@ -1062,7 +1075,7 @@ static void init_registers(struct net_device *dev)
static void gfar_halt_nodisable(struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
- struct gfar __iomem *regs = priv->regs;
+ struct gfar __iomem *regs = priv->gfargrp.regs;
u32 tempval;
/* Mask all interrupts */
@@ -1072,13 +1085,13 @@ static void gfar_halt_nodisable(struct net_device *dev)
gfar_write(®s->ievent, IEVENT_INIT_CLEAR);
/* Stop the DMA, and wait for it to stop */
- tempval = gfar_read(&priv->regs->dmactrl);
+ tempval = gfar_read(®s->dmactrl);
if ((tempval & (DMACTRL_GRS | DMACTRL_GTS))
!= (DMACTRL_GRS | DMACTRL_GTS)) {
tempval |= (DMACTRL_GRS | DMACTRL_GTS);
- gfar_write(&priv->regs->dmactrl, tempval);
+ gfar_write(®s->dmactrl, tempval);
- while (!(gfar_read(&priv->regs->ievent) &
+ while (!(gfar_read(®s->ievent) &
(IEVENT_GRSC | IEVENT_GTSC)))
cpu_relax();
}
@@ -1088,7 +1101,7 @@ static void gfar_halt_nodisable(struct net_device *dev)
void gfar_halt(struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
- struct gfar __iomem *regs = priv->regs;
+ struct gfar __iomem *regs = priv->gfargrp.regs;
u32 tempval;
gfar_halt_nodisable(dev);
@@ -1122,11 +1135,11 @@ void stop_gfar(struct net_device *dev)
/* Free the IRQs */
if (priv->device_flags & FSL_GIANFAR_DEV_HAS_MULTI_INTR) {
- free_irq(priv->interruptError, dev);
- free_irq(priv->interruptTransmit, dev);
- free_irq(priv->interruptReceive, dev);
+ free_irq(priv->gfargrp.interruptError, &priv->gfargrp);
+ free_irq(priv->gfargrp.interruptTransmit, &priv->gfargrp);
+ free_irq(priv->gfargrp.interruptReceive, &priv->gfargrp);
} else {
- free_irq(priv->interruptTransmit, dev);
+ free_irq(priv->gfargrp.interruptTransmit, &priv->gfargrp);
}
free_skb_resources(priv);
@@ -1201,9 +1214,7 @@ skip_rx_skbuff:
void gfar_start(struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
- struct gfar_priv_tx_q *tx_queue;
- struct gfar_priv_rx_q *rx_queue;
- struct gfar __iomem *regs = priv->regs;
+ struct gfar __iomem *regs = priv->gfargrp.regs;
u32 tempval;
/* Enable Rx and Tx in MACCFG1 */
@@ -1212,14 +1223,14 @@ void gfar_start(struct net_device *dev)
gfar_write(®s->maccfg1, tempval);
/* Initialize DMACTRL to have WWR and WOP */
- tempval = gfar_read(&priv->regs->dmactrl);
+ tempval = gfar_read(®s->dmactrl);
tempval |= DMACTRL_INIT_SETTINGS;
- gfar_write(&priv->regs->dmactrl, tempval);
+ gfar_write(®s->dmactrl, tempval);
/* Make sure we aren't stopped */
- tempval = gfar_read(&priv->regs->dmactrl);
+ tempval = gfar_read(®s->dmactrl);
tempval &= ~(DMACTRL_GRS | DMACTRL_GTS);
- gfar_write(&priv->regs->dmactrl, tempval);
+ gfar_write(®s->dmactrl, tempval);
/* Clear THLT/RHLT, so that the DMA starts polling now */
gfar_write(®s->tstat, TSTAT_CLEAR_THALT);
@@ -1235,7 +1246,7 @@ void gfar_start(struct net_device *dev)
int startup_gfar(struct net_device *ndev)
{
struct gfar_private *priv = netdev_priv(ndev);
- struct gfar __iomem *regs = priv->regs;
+ struct gfar __iomem *regs = priv->gfargrp.regs;
int err;
gfar_write(®s->imask, IMASK_INIT_CLEAR);
@@ -1251,39 +1262,46 @@ int startup_gfar(struct net_device *ndev)
if (priv->device_flags & FSL_GIANFAR_DEV_HAS_MULTI_INTR) {
/* Install our interrupt handlers for Error,
* Transmit, and Receive */
- err = request_irq(priv->interruptError, gfar_error, 0,
- priv->int_name_er, ndev);
+ err = request_irq(priv->gfargrp.interruptError, gfar_error, 0,
+ priv->gfargrp.int_name_er, &priv->gfargrp);
if (err) {
if (netif_msg_intr(priv))
pr_err("%s: Can't get IRQ %d\n", ndev->name,
- priv->interruptError);
+ priv->gfargrp.interruptError);
goto err_irq_fail;
}
- err = request_irq(priv->interruptTransmit, gfar_transmit, 0,
- priv->int_name_tx, ndev);
+ err = request_irq(priv->gfargrp.interruptTransmit,
+ gfar_transmit, 0,
+ priv->gfargrp.int_name_tx,
+ &priv->gfargrp);
if (err) {
if (netif_msg_intr(priv))
pr_err("%s: Can't get IRQ %d\n", ndev->name,
- priv->interruptTransmit);
+ priv->gfargrp.interruptTransmit);
goto tx_irq_fail;
}
- err = request_irq(priv->interruptReceive, gfar_receive, 0,
- priv->int_name_rx, ndev);
+ err = request_irq(priv->gfargrp.interruptReceive,
+ gfar_receive, 0,
+ priv->gfargrp.int_name_rx,
+ &priv->gfargrp);
if (err) {
if (netif_msg_intr(priv))
pr_err("%s: Can't get IRQ %d (receive0)\n",
- ndev->name, priv->interruptReceive);
+ ndev->name,
+ priv->gfargrp.interruptReceive);
goto rx_irq_fail;
}
} else {
- err = request_irq(priv->interruptTransmit, gfar_interrupt,
- 0, priv->int_name_tx, ndev);
+ err = request_irq(priv->gfargrp.interruptTransmit,
+ gfar_interrupt, 0,
+ priv->gfargrp.int_name_tx,
+ &priv->gfargrp);
if (err) {
if (netif_msg_intr(priv))
pr_err("%s: Can't get IRQ %d\n", ndev->name,
- priv->interruptTransmit);
+ priv->gfargrp.interruptTransmit);
goto err_irq_fail;
}
}
@@ -1296,9 +1314,9 @@ int startup_gfar(struct net_device *ndev)
return 0;
rx_irq_fail:
- free_irq(priv->interruptTransmit, ndev);
+ free_irq(priv->gfargrp.interruptTransmit, &priv->gfargrp);
tx_irq_fail:
- free_irq(priv->interruptError, ndev);
+ free_irq(priv->gfargrp.interruptError, &priv->gfargrp);
err_irq_fail:
free_skb_resources(priv);
return err;
@@ -1403,6 +1421,7 @@ static int gfar_start_xmit(struct sk_buff *skb, struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
struct gfar_priv_tx_q *tx_queue = NULL;
+ struct gfar __iomem *regs = NULL;
struct txfcb *fcb = NULL;
struct txbd8 *txbdp, *txbdp_start, *base;
u32 lstatus;
@@ -1413,6 +1432,7 @@ static int gfar_start_xmit(struct sk_buff *skb, struct net_device *dev)
tx_queue = priv->tx_queue;
base = tx_queue->tx_bd_base;
+ regs = priv->gfargrp.regs;
/* make space for additional header when fcb is needed */
if (((skb->ip_summed == CHECKSUM_PARTIAL) ||
@@ -1536,7 +1556,7 @@ static int gfar_start_xmit(struct sk_buff *skb, struct net_device *dev)
}
/* Tell the DMA to go go go */
- gfar_write(&priv->regs->tstat, TSTAT_CLEAR_THALT);
+ gfar_write(®s->tstat, TSTAT_CLEAR_THALT);
/* Unlock priv */
spin_unlock_irqrestore(&tx_queue->txlock, flags);
@@ -1579,40 +1599,42 @@ static void gfar_vlan_rx_register(struct net_device *dev,
{
struct gfar_private *priv = netdev_priv(dev);
struct gfar_priv_rx_q *rx_queue = NULL;
+ struct gfar __iomem *regs = NULL;
unsigned long flags;
u32 tempval;
rx_queue = priv->rx_queue;
+ regs = priv->gfargrp.regs;
spin_lock_irqsave(&rx_queue->rxlock, flags);
priv->vlgrp = grp;
if (grp) {
/* Enable VLAN tag insertion */
- tempval = gfar_read(&priv->regs->tctrl);
+ tempval = gfar_read(®s->tctrl);
tempval |= TCTRL_VLINS;
- gfar_write(&priv->regs->tctrl, tempval);
+ gfar_write(®s->tctrl, tempval);
/* Enable VLAN tag extraction */
- tempval = gfar_read(&priv->regs->rctrl);
+ tempval = gfar_read(®s->rctrl);
tempval |= (RCTRL_VLEX | RCTRL_PRSDEP_INIT);
- gfar_write(&priv->regs->rctrl, tempval);
+ gfar_write(®s->rctrl, tempval);
} else {
/* Disable VLAN tag insertion */
- tempval = gfar_read(&priv->regs->tctrl);
+ tempval = gfar_read(®s->tctrl);
tempval &= ~TCTRL_VLINS;
- gfar_write(&priv->regs->tctrl, tempval);
+ gfar_write(®s->tctrl, tempval);
/* Disable VLAN tag extraction */
- tempval = gfar_read(&priv->regs->rctrl);
+ tempval = gfar_read(®s->rctrl);
tempval &= ~RCTRL_VLEX;
/* If parse is no longer required, then disable parser */
if (tempval & RCTRL_REQ_PARSER)
tempval |= RCTRL_PRSDEP_INIT;
else
tempval &= ~RCTRL_PRSDEP_INIT;
- gfar_write(&priv->regs->rctrl, tempval);
+ gfar_write(®s->rctrl, tempval);
}
gfar_change_mtu(dev, dev->mtu);
@@ -1624,6 +1646,7 @@ static int gfar_change_mtu(struct net_device *dev, int new_mtu)
{
int tempsize, tempval;
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar __iomem *regs = priv->gfargrp.regs;
int oldsize = priv->rx_buffer_size;
int frame_size = new_mtu + ETH_HLEN;
@@ -1655,20 +1678,20 @@ static int gfar_change_mtu(struct net_device *dev, int new_mtu)
dev->mtu = new_mtu;
- gfar_write(&priv->regs->mrblr, priv->rx_buffer_size);
- gfar_write(&priv->regs->maxfrm, priv->rx_buffer_size);
+ gfar_write(®s->mrblr, priv->rx_buffer_size);
+ gfar_write(®s->maxfrm, priv->rx_buffer_size);
/* If the mtu is larger than the max size for standard
* ethernet frames (ie, a jumbo frame), then set maccfg2
* to allow huge frames, and to check the length */
- tempval = gfar_read(&priv->regs->maccfg2);
+ tempval = gfar_read(®s->maccfg2);
if (priv->rx_buffer_size > DEFAULT_RX_BUFFER_SIZE)
tempval |= (MACCFG2_HUGEFRAME | MACCFG2_LENGTHCHECK);
else
tempval &= ~(MACCFG2_HUGEFRAME | MACCFG2_LENGTHCHECK);
- gfar_write(&priv->regs->maccfg2, tempval);
+ gfar_write(®s->maccfg2, tempval);
if ((oldsize != tempsize) && (dev->flags & IFF_UP))
startup_gfar(dev);
@@ -1787,9 +1810,9 @@ static int gfar_clean_tx_ring(struct gfar_priv_tx_q *tx_queue)
return howmany;
}
-static void gfar_schedule_cleanup(struct net_device *dev)
+static void gfar_schedule_cleanup(struct gfar_priv_grp *gfargrp)
{
- struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_private *priv = gfargrp->priv;
struct gfar_priv_tx_q *tx_queue = NULL;
struct gfar_priv_rx_q *rx_queue = NULL;
unsigned long flags;
@@ -1800,14 +1823,14 @@ static void gfar_schedule_cleanup(struct net_device *dev)
spin_lock(&rx_queue->rxlock);
if (napi_schedule_prep(&rx_queue->napi)) {
- gfar_write(&priv->regs->imask, IMASK_RTX_DISABLED);
+ gfar_write(&gfargrp->regs->imask, IMASK_RTX_DISABLED);
__napi_schedule(&rx_queue->napi);
} else {
/*
* Clear IEVENT, so interrupts aren't called again
* because of the packets that have already arrived.
*/
- gfar_write(&priv->regs->ievent, IEVENT_RTX_MASK);
+ gfar_write(&gfargrp->regs->ievent, IEVENT_RTX_MASK);
}
spin_unlock(&rx_queue->rxlock);
@@ -1815,9 +1838,9 @@ static void gfar_schedule_cleanup(struct net_device *dev)
}
/* Interrupt Handler for Transmit complete */
-static irqreturn_t gfar_transmit(int irq, void *dev_id)
+static irqreturn_t gfar_transmit(int irq, void *grp_id)
{
- gfar_schedule_cleanup((struct net_device *)dev_id);
+ gfar_schedule_cleanup((struct gfar_priv_grp *)grp_id);
return IRQ_HANDLED;
}
@@ -1897,9 +1920,9 @@ static inline void count_errors(unsigned short status, struct net_device *dev)
}
}
-irqreturn_t gfar_receive(int irq, void *dev_id)
+irqreturn_t gfar_receive(int irq, void *grp_id)
{
- gfar_schedule_cleanup((struct net_device *)dev_id);
+ gfar_schedule_cleanup((struct gfar_priv_grp *)grp_id);
return IRQ_HANDLED;
}
@@ -2053,6 +2076,7 @@ static int gfar_poll(struct napi_struct *napi, int budget)
struct gfar_priv_rx_q, napi);
struct net_device *dev = rx_queue->dev;
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar __iomem *regs = priv->gfargrp.regs;
struct gfar_priv_tx_q *tx_queue = NULL;
int tx_cleaned = 0;
int rx_cleaned = 0;
@@ -2060,7 +2084,7 @@ static int gfar_poll(struct napi_struct *napi, int budget)
/* Clear IEVENT, so interrupts aren't called again
* because of the packets that have already arrived */
- gfar_write(&priv->regs->ievent, IEVENT_RTX_MASK);
+ gfar_write(®s->ievent, IEVENT_RTX_MASK);
tx_queue = priv->tx_queue;
/* If we fail to get the lock, don't bother with the TX BDs */
@@ -2078,19 +2102,19 @@ static int gfar_poll(struct napi_struct *napi, int budget)
napi_complete(napi);
/* Clear the halt bit in RSTAT */
- gfar_write(&priv->regs->rstat, RSTAT_CLEAR_RHALT);
+ gfar_write(®s->rstat, RSTAT_CLEAR_RHALT);
- gfar_write(&priv->regs->imask, IMASK_DEFAULT);
+ gfar_write(®s->imask, IMASK_DEFAULT);
/* If we are coalescing interrupts, update the timer */
/* Otherwise, clear it */
if (likely(rx_queue->rxcoalescing)) {
- gfar_write(&priv->regs->rxic, 0);
- gfar_write(&priv->regs->rxic, rx_queue->rxic);
+ gfar_write(®s->rxic, 0);
+ gfar_write(®s->rxic, rx_queue->rxic);
}
if (likely(tx_queue->txcoalescing)) {
- gfar_write(&priv->regs->txic, 0);
- gfar_write(&priv->regs->txic, tx_queue->txic);
+ gfar_write(®s->txic, 0);
+ gfar_write(®s->txic, tx_queue->txic);
}
}
@@ -2109,41 +2133,40 @@ static void gfar_netpoll(struct net_device *dev)
/* If the device has multiple interrupts, run tx/rx */
if (priv->device_flags & FSL_GIANFAR_DEV_HAS_MULTI_INTR) {
- disable_irq(priv->interruptTransmit);
- disable_irq(priv->interruptReceive);
- disable_irq(priv->interruptError);
- gfar_interrupt(priv->interruptTransmit, dev);
- enable_irq(priv->interruptError);
- enable_irq(priv->interruptReceive);
- enable_irq(priv->interruptTransmit);
+ disable_irq(priv->gfargrp.interruptTransmit);
+ disable_irq(priv->gfargrp.interruptReceive);
+ disable_irq(priv->gfargrp.interruptError);
+ gfar_interrupt(priv->gfargrp.interruptTransmit, &priv->gfargrp);
+ enable_irq(priv->gfargrp.interruptError);
+ enable_irq(priv->gfargrp.interruptReceive);
+ enable_irq(priv->gfargrp.interruptTransmit);
} else {
- disable_irq(priv->interruptTransmit);
- gfar_interrupt(priv->interruptTransmit, dev);
- enable_irq(priv->interruptTransmit);
+ disable_irq(priv->gfargrp.interruptTransmit);
+ gfar_interrupt(priv->gfargrp.interruptTransmit, &priv->gfargrp);
+ enable_irq(priv->gfargrp.interruptTransmit);
}
}
#endif
/* The interrupt handler for devices with one interrupt */
-static irqreturn_t gfar_interrupt(int irq, void *dev_id)
+static irqreturn_t gfar_interrupt(int irq, void *grp_id)
{
- struct net_device *dev = dev_id;
- struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_grp *gfargrp = grp_id;
/* Save ievent for future reference */
- u32 events = gfar_read(&priv->regs->ievent);
+ u32 events = gfar_read(&gfargrp->regs->ievent);
/* Check for reception */
if (events & IEVENT_RX_MASK)
- gfar_receive(irq, dev_id);
+ gfar_receive(irq, grp_id);
/* Check for transmit completion */
if (events & IEVENT_TX_MASK)
- gfar_transmit(irq, dev_id);
+ gfar_transmit(irq, grp_id);
/* Check for errors */
if (events & IEVENT_ERR_MASK)
- gfar_error(irq, dev_id);
+ gfar_error(irq, grp_id);
return IRQ_HANDLED;
}
@@ -2158,7 +2181,7 @@ static void adjust_link(struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
struct gfar_priv_tx_q *tx_queue = NULL;
- struct gfar __iomem *regs = priv->regs;
+ struct gfar __iomem *regs = priv->gfargrp.regs;
unsigned long flags;
struct phy_device *phydev = priv->phydev;
int new_state = 0;
@@ -2241,7 +2264,7 @@ static void gfar_set_multi(struct net_device *dev)
{
struct dev_mc_list *mc_ptr;
struct gfar_private *priv = netdev_priv(dev);
- struct gfar __iomem *regs = priv->regs;
+ struct gfar __iomem *regs = priv->gfargrp.regs;
u32 tempval;
if (dev->flags & IFF_PROMISC) {
@@ -2374,10 +2397,11 @@ static void gfar_set_hash_for_addr(struct net_device *dev, u8 *addr)
static void gfar_set_mac_for_addr(struct net_device *dev, int num, u8 *addr)
{
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar __iomem *regs = priv->gfargrp.regs;
int idx;
char tmpbuf[MAC_ADDR_LEN];
u32 tempval;
- u32 __iomem *macptr = &priv->regs->macstnaddr1;
+ u32 __iomem *macptr = ®s->macstnaddr1;
macptr += num*2;
@@ -2394,16 +2418,18 @@ static void gfar_set_mac_for_addr(struct net_device *dev, int num, u8 *addr)
}
/* GFAR error interrupt handler */
-static irqreturn_t gfar_error(int irq, void *dev_id)
+static irqreturn_t gfar_error(int irq, void *grp_id)
{
- struct net_device *dev = dev_id;
- struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_grp *gfargrp = grp_id;
+ struct gfar __iomem *regs = gfargrp->regs;
+ struct gfar_private *priv= gfargrp->priv;
+ struct net_device *dev = priv->ndev;
/* Save ievent for future reference */
- u32 events = gfar_read(&priv->regs->ievent);
+ u32 events = gfar_read(®s->ievent);
/* Clear IEVENT */
- gfar_write(&priv->regs->ievent, events & IEVENT_ERR_MASK);
+ gfar_write(®s->ievent, events & IEVENT_ERR_MASK);
/* Magic Packet is not an error. */
if ((priv->device_flags & FSL_GIANFAR_DEV_HAS_MAGIC_PACKET) &&
@@ -2413,7 +2439,7 @@ static irqreturn_t gfar_error(int irq, void *dev_id)
/* Hmm... */
if (netif_msg_rx_err(priv) || netif_msg_tx_err(priv))
printk(KERN_DEBUG "%s: error interrupt (ievent=0x%08x imask=0x%08x)\n",
- dev->name, events, gfar_read(&priv->regs->imask));
+ dev->name, events, gfar_read(®s->imask));
/* Update the error counters */
if (events & IEVENT_TXE) {
@@ -2431,7 +2457,7 @@ static irqreturn_t gfar_error(int irq, void *dev_id)
priv->extra_stats.tx_underrun++;
/* Reactivate the Tx Queues */
- gfar_write(&priv->regs->tstat, TSTAT_CLEAR_THALT);
+ gfar_write(®s->tstat, TSTAT_CLEAR_THALT);
}
if (netif_msg_tx_err(priv))
printk(KERN_DEBUG "%s: Transmit Error\n", dev->name);
@@ -2440,11 +2466,11 @@ static irqreturn_t gfar_error(int irq, void *dev_id)
dev->stats.rx_errors++;
priv->extra_stats.rx_bsy++;
- gfar_receive(irq, dev_id);
+ gfar_receive(irq, grp_id);
if (netif_msg_rx_err(priv))
printk(KERN_DEBUG "%s: busy error (rstat: %x)\n",
- dev->name, gfar_read(&priv->regs->rstat));
+ dev->name, gfar_read(®s->rstat));
}
if (events & IEVENT_BABR) {
dev->stats.rx_errors++;
diff --git a/drivers/net/gianfar.h b/drivers/net/gianfar.h
index a60f93f..79e8471 100644
--- a/drivers/net/gianfar.h
+++ b/drivers/net/gianfar.h
@@ -770,6 +770,32 @@ struct gfar_priv_rx_q {
unsigned long rxic;
};
+/**
+ * struct gfar_priv_grp - per group structure
+ * @priv: back pointer to the priv structure
+ * @regs: the ioremapped register space for this group
+ * @grp_id: group id for this group
+ * @interruptTransmit: The TX interrupt number for this group
+ * @interruptReceive: The RX interrupt number for this group
+ * @interruptError: The ERROR interrupt number for this group
+ * @int_name_tx: tx interrupt name for this group
+ * @int_name_rx: rx interrupt name for this group
+ * @int_name_er: er interrupt name for this group
+ */
+
+struct gfar_priv_grp {
+ spinlock_t grplock __attribute__ ((aligned (SMP_CACHE_BYTES)));
+ struct gfar_private *priv;
+ struct gfar __iomem *regs;
+ unsigned int interruptTransmit;
+ unsigned int interruptReceive;
+ unsigned int interruptError;
+
+ char int_name_tx[GFAR_INT_NAME_MAX];
+ char int_name_rx[GFAR_INT_NAME_MAX];
+ char int_name_er[GFAR_INT_NAME_MAX];
+};
+
/* Struct stolen almost completely (and shamelessly) from the FCC enet source
* (Ok, that's not so true anymore, but there is a family resemblence)
* The GFAR buffer descriptors track the ring buffers. The rx_bd_base
@@ -785,6 +811,7 @@ struct gfar_private {
struct net_device *ndev;
struct of_device *ofdev;
+ struct gfar_priv_grp gfargrp;
struct gfar_priv_tx_q *tx_queue;
struct gfar_priv_rx_q *rx_queue;
@@ -797,9 +824,6 @@ struct gfar_private {
struct vlan_group *vlgrp;
- /* Unprotected fields */
- /* Pointer to the GFAR memory mapped Registers */
- struct gfar __iomem *regs;
/* Hash registers and their width */
u32 __iomem *hash_regs[16];
@@ -823,10 +847,6 @@ struct gfar_private {
wol_en:1; /* Wake-on-LAN enabled */
unsigned short padding;
- unsigned int interruptTransmit;
- unsigned int interruptReceive;
- unsigned int interruptError;
-
/* PHY stuff */
struct phy_device *phydev;
struct mii_bus *mii_bus;
@@ -838,10 +858,6 @@ struct gfar_private {
struct work_struct reset_task;
- char int_name_tx[GFAR_INT_NAME_MAX];
- char int_name_rx[GFAR_INT_NAME_MAX];
- char int_name_er[GFAR_INT_NAME_MAX];
-
/* Network Statistics */
struct gfar_extra_stats extra_stats;
};
diff --git a/drivers/net/gianfar_ethtool.c b/drivers/net/gianfar_ethtool.c
index 6d0d171..c681b41 100644
--- a/drivers/net/gianfar_ethtool.c
+++ b/drivers/net/gianfar_ethtool.c
@@ -137,10 +137,11 @@ static void gfar_fill_stats(struct net_device *dev, struct ethtool_stats *dummy,
{
int i;
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar __iomem *regs = priv->gfargrp.regs;
u64 *extra = (u64 *) & priv->extra_stats;
if (priv->device_flags & FSL_GIANFAR_DEV_HAS_RMON) {
- u32 __iomem *rmon = (u32 __iomem *) & priv->regs->rmon;
+ u32 __iomem *rmon = (u32 __iomem *) ®s->rmon;
struct gfar_stats *stats = (struct gfar_stats *) buf;
for (i = 0; i < GFAR_RMON_LEN; i++)
@@ -223,7 +224,7 @@ static void gfar_get_regs(struct net_device *dev, struct ethtool_regs *regs, voi
{
int i;
struct gfar_private *priv = netdev_priv(dev);
- u32 __iomem *theregs = (u32 __iomem *) priv->regs;
+ u32 __iomem *theregs = (u32 __iomem *) priv->gfargrp.regs;
u32 *buf = (u32 *) regbuf;
for (i = 0; i < sizeof (struct gfar) / sizeof (u32); i++)
@@ -349,6 +350,7 @@ static int gfar_gcoalesce(struct net_device *dev, struct ethtool_coalesce *cvals
static int gfar_scoalesce(struct net_device *dev, struct ethtool_coalesce *cvals)
{
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar __iomem *regs = priv->gfargrp.regs;
struct gfar_priv_tx_q *tx_queue = NULL;
struct gfar_priv_rx_q *rx_queue = NULL;
@@ -407,13 +409,13 @@ static int gfar_scoalesce(struct net_device *dev, struct ethtool_coalesce *cvals
tx_queue->txic = mk_ic_value(cvals->tx_max_coalesced_frames,
gfar_usecs2ticks(priv, cvals->tx_coalesce_usecs));
- gfar_write(&priv->regs->rxic, 0);
+ gfar_write(®s->rxic, 0);
if (rx_queue->rxcoalescing)
- gfar_write(&priv->regs->rxic, rx_queue->rxic);
+ gfar_write(®s->rxic, rx_queue->rxic);
- gfar_write(&priv->regs->txic, 0);
+ gfar_write(®s->txic, 0);
if (tx_queue->txcoalescing)
- gfar_write(&priv->regs->txic, tx_queue->txic);
+ gfar_write(®s->txic, tx_queue->txic);
return 0;
}
diff --git a/drivers/net/gianfar_sysfs.c b/drivers/net/gianfar_sysfs.c
index 9c664f8..adea11e 100644
--- a/drivers/net/gianfar_sysfs.c
+++ b/drivers/net/gianfar_sysfs.c
@@ -50,6 +50,7 @@ static ssize_t gfar_set_bd_stash(struct device *dev,
const char *buf, size_t count)
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
+ struct gfar __iomem *regs = priv->gfargrp.regs;
struct gfar_priv_rx_q *rx_queue = NULL;
int new_setting = 0;
u32 temp;
@@ -74,14 +75,14 @@ static ssize_t gfar_set_bd_stash(struct device *dev,
/* Set the new stashing value */
priv->bd_stash_en = new_setting;
- temp = gfar_read(&priv->regs->attr);
+ temp = gfar_read(®s->attr);
if (new_setting)
temp |= ATTR_BDSTASH;
else
temp &= ~(ATTR_BDSTASH);
- gfar_write(&priv->regs->attr, temp);
+ gfar_write(®s->attr, temp);
spin_unlock_irqrestore(&rx_queue->rxlock, flags);
@@ -103,6 +104,7 @@ static ssize_t gfar_set_rx_stash_size(struct device *dev,
const char *buf, size_t count)
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
+ struct gfar __iomem *regs = priv->gfargrp.regs;
struct gfar_priv_rx_q *rx_queue = NULL;
unsigned int length = simple_strtoul(buf, NULL, 0);
u32 temp;
@@ -122,20 +124,20 @@ static ssize_t gfar_set_rx_stash_size(struct device *dev,
priv->rx_stash_size = length;
- temp = gfar_read(&priv->regs->attreli);
+ temp = gfar_read(®s->attreli);
temp &= ~ATTRELI_EL_MASK;
temp |= ATTRELI_EL(length);
- gfar_write(&priv->regs->attreli, temp);
+ gfar_write(®s->attreli, temp);
/* Turn stashing on/off as appropriate */
- temp = gfar_read(&priv->regs->attr);
+ temp = gfar_read(®s->attr);
if (length)
temp |= ATTR_BUFSTASH;
else
temp &= ~(ATTR_BUFSTASH);
- gfar_write(&priv->regs->attr, temp);
+ gfar_write(®s->attr, temp);
out:
spin_unlock_irqrestore(&rx_queue->rxlock, flags);
@@ -161,6 +163,7 @@ static ssize_t gfar_set_rx_stash_index(struct device *dev,
const char *buf, size_t count)
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
+ struct gfar __iomem *regs = priv->gfargrp.regs;
struct gfar_priv_rx_q *rx_queue = NULL;
unsigned short index = simple_strtoul(buf, NULL, 0);
u32 temp;
@@ -180,10 +183,10 @@ static ssize_t gfar_set_rx_stash_index(struct device *dev,
priv->rx_stash_index = index;
- temp = gfar_read(&priv->regs->attreli);
+ temp = gfar_read(®s->attreli);
temp &= ~ATTRELI_EI_MASK;
temp |= ATTRELI_EI(index);
- gfar_write(&priv->regs->attreli, flags);
+ gfar_write(®s->attreli, flags);
out:
spin_unlock_irqrestore(&rx_queue->rxlock, flags);
@@ -208,6 +211,7 @@ static ssize_t gfar_set_fifo_threshold(struct device *dev,
const char *buf, size_t count)
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
+ struct gfar __iomem *regs = priv->gfargrp.regs;
struct gfar_priv_tx_q *tx_queue = NULL;
unsigned int length = simple_strtoul(buf, NULL, 0);
u32 temp;
@@ -222,10 +226,10 @@ static ssize_t gfar_set_fifo_threshold(struct device *dev,
priv->fifo_threshold = length;
- temp = gfar_read(&priv->regs->fifo_tx_thr);
+ temp = gfar_read(®s->fifo_tx_thr);
temp &= ~FIFO_TX_THR_MASK;
temp |= length;
- gfar_write(&priv->regs->fifo_tx_thr, temp);
+ gfar_write(®s->fifo_tx_thr, temp);
spin_unlock_irqrestore(&tx_queue->txlock, flags);
@@ -248,6 +252,7 @@ static ssize_t gfar_set_fifo_starve(struct device *dev,
const char *buf, size_t count)
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
+ struct gfar __iomem *regs = priv->gfargrp.regs;
struct gfar_priv_tx_q *tx_queue = NULL;
unsigned int num = simple_strtoul(buf, NULL, 0);
u32 temp;
@@ -261,10 +266,10 @@ static ssize_t gfar_set_fifo_starve(struct device *dev,
priv->fifo_starve = num;
- temp = gfar_read(&priv->regs->fifo_tx_starve);
+ temp = gfar_read(®s->fifo_tx_starve);
temp &= ~FIFO_TX_STARVE_MASK;
temp |= num;
- gfar_write(&priv->regs->fifo_tx_starve, temp);
+ gfar_write(®s->fifo_tx_starve, temp);
spin_unlock_irqrestore(&tx_queue->txlock, flags);
@@ -288,6 +293,7 @@ static ssize_t gfar_set_fifo_starve_off(struct device *dev,
const char *buf, size_t count)
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
+ struct gfar __iomem *regs = priv->gfargrp.regs;
struct gfar_priv_tx_q *tx_queue = NULL;
unsigned int num = simple_strtoul(buf, NULL, 0);
u32 temp;
@@ -301,10 +307,10 @@ static ssize_t gfar_set_fifo_starve_off(struct device *dev,
priv->fifo_starve_off = num;
- temp = gfar_read(&priv->regs->fifo_tx_starve_shutoff);
+ temp = gfar_read(®s->fifo_tx_starve_shutoff);
temp &= ~FIFO_TX_STARVE_OFF_MASK;
temp |= num;
- gfar_write(&priv->regs->fifo_tx_starve_shutoff, temp);
+ gfar_write(®s->fifo_tx_starve_shutoff, temp);
spin_unlock_irqrestore(&tx_queue->txlock, flags);
--
1.5.2.2
^ permalink raw reply related
* [PATCH v1 1/7] gianfar: Add per queue structure support
From: Sandeep Gopalpet @ 2009-10-26 16:27 UTC (permalink / raw)
To: netdev; +Cc: Sandeep Gopalpet
This patch introduces per tx and per rx queue structures.
Earlier the members of these structures were inside the
gfar_private structure.
Moving forward if we want to support multiple queues, we need
to refactor the gfar_private structure so that introduction of
multiple queues is easier.
Signed-off-by: Sandeep Gopalpet <sandeep.kumar@freescale.com>
---
drivers/net/gianfar.c | 384 ++++++++++++++++++++++++-----------------
drivers/net/gianfar.h | 116 ++++++++-----
drivers/net/gianfar_ethtool.c | 100 +++++++----
drivers/net/gianfar_sysfs.c | 43 ++++--
4 files changed, 398 insertions(+), 245 deletions(-)
diff --git a/drivers/net/gianfar.c b/drivers/net/gianfar.c
index f714186..354b2b5 100644
--- a/drivers/net/gianfar.c
+++ b/drivers/net/gianfar.c
@@ -8,9 +8,10 @@
*
* Author: Andy Fleming
* Maintainer: Kumar Gala
+ * Modifier: Sandeep Gopalpet <sandeep.kumar@freescale.com>
*
- * Copyright (c) 2002-2006 Freescale Semiconductor, Inc.
- * Copyright (c) 2007 MontaVista Software, Inc.
+ * Copyright 2002-2009 Freescale Semiconductor, Inc.
+ * Copyright 2007 MontaVista Software, Inc.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
@@ -109,7 +110,7 @@ static void gfar_reset_task(struct work_struct *work);
static void gfar_timeout(struct net_device *dev);
static int gfar_close(struct net_device *dev);
struct sk_buff *gfar_new_skb(struct net_device *dev);
-static void gfar_new_rxbdp(struct net_device *dev, struct rxbd8 *bdp,
+static void gfar_new_rxbdp(struct gfar_priv_rx_q *rx_queue, struct rxbd8 *bdp,
struct sk_buff *skb);
static int gfar_set_mac_address(struct net_device *dev);
static int gfar_change_mtu(struct net_device *dev, int new_mtu);
@@ -130,8 +131,8 @@ static int gfar_poll(struct napi_struct *napi, int budget);
#ifdef CONFIG_NET_POLL_CONTROLLER
static void gfar_netpoll(struct net_device *dev);
#endif
-int gfar_clean_rx_ring(struct net_device *dev, int rx_work_limit);
-static int gfar_clean_tx_ring(struct net_device *dev);
+int gfar_clean_rx_ring(struct gfar_priv_rx_q *rx_queue, int rx_work_limit);
+static int gfar_clean_tx_ring(struct gfar_priv_tx_q *tx_queue);
static int gfar_process_frame(struct net_device *dev, struct sk_buff *skb,
int amount_pull);
static void gfar_vlan_rx_register(struct net_device *netdev,
@@ -147,16 +148,16 @@ MODULE_AUTHOR("Freescale Semiconductor, Inc");
MODULE_DESCRIPTION("Gianfar Ethernet Driver");
MODULE_LICENSE("GPL");
-static void gfar_init_rxbdp(struct net_device *dev, struct rxbd8 *bdp,
+static void gfar_init_rxbdp(struct gfar_priv_rx_q *rx_queue, struct rxbd8 *bdp,
dma_addr_t buf)
{
- struct gfar_private *priv = netdev_priv(dev);
+ struct net_device *dev = rx_queue->dev;
u32 lstatus;
bdp->bufPtr = buf;
lstatus = BD_LFLAG(RXBD_EMPTY | RXBD_INTERRUPT);
- if (bdp == priv->rx_bd_base + priv->rx_ring_size - 1)
+ if (bdp == rx_queue->rx_bd_base + rx_queue->rx_ring_size - 1)
lstatus |= BD_LFLAG(RXBD_WRAP);
eieio();
@@ -167,20 +168,25 @@ static void gfar_init_rxbdp(struct net_device *dev, struct rxbd8 *bdp,
static int gfar_init_bds(struct net_device *ndev)
{
struct gfar_private *priv = netdev_priv(ndev);
+ struct gfar_priv_tx_q *tx_queue = NULL;
+ struct gfar_priv_rx_q *rx_queue = NULL;
struct txbd8 *txbdp;
struct rxbd8 *rxbdp;
int i;
+ tx_queue = priv->tx_queue;
+ rx_queue = priv->rx_queue;
+
/* Initialize some variables in our dev structure */
- priv->num_txbdfree = priv->tx_ring_size;
- priv->dirty_tx = priv->cur_tx = priv->tx_bd_base;
- priv->cur_rx = priv->rx_bd_base;
- priv->skb_curtx = priv->skb_dirtytx = 0;
- priv->skb_currx = 0;
+ tx_queue->num_txbdfree = tx_queue->tx_ring_size;
+ tx_queue->dirty_tx = tx_queue->cur_tx = tx_queue->tx_bd_base;
+ rx_queue->cur_rx = rx_queue->rx_bd_base;
+ tx_queue->skb_curtx = tx_queue->skb_dirtytx = 0;
+ rx_queue->skb_currx = 0;
/* Initialize Transmit Descriptor Ring */
- txbdp = priv->tx_bd_base;
- for (i = 0; i < priv->tx_ring_size; i++) {
+ txbdp = tx_queue->tx_bd_base;
+ for (i = 0; i < tx_queue->tx_ring_size; i++) {
txbdp->lstatus = 0;
txbdp->bufPtr = 0;
txbdp++;
@@ -190,12 +196,12 @@ static int gfar_init_bds(struct net_device *ndev)
txbdp--;
txbdp->status |= TXBD_WRAP;
- rxbdp = priv->rx_bd_base;
- for (i = 0; i < priv->rx_ring_size; i++) {
- struct sk_buff *skb = priv->rx_skbuff[i];
+ rxbdp = rx_queue->rx_bd_base;
+ for (i = 0; i < rx_queue->rx_ring_size; i++) {
+ struct sk_buff *skb = rx_queue->rx_skbuff[i];
if (skb) {
- gfar_init_rxbdp(ndev, rxbdp, rxbdp->bufPtr);
+ gfar_init_rxbdp(rx_queue, rxbdp, rxbdp->bufPtr);
} else {
skb = gfar_new_skb(ndev);
if (!skb) {
@@ -203,9 +209,9 @@ static int gfar_init_bds(struct net_device *ndev)
ndev->name);
return -ENOMEM;
}
- priv->rx_skbuff[i] = skb;
+ rx_queue->rx_skbuff[i] = skb;
- gfar_new_rxbdp(ndev, rxbdp, skb);
+ gfar_new_rxbdp(rx_queue, rxbdp, skb);
}
rxbdp++;
@@ -220,12 +226,17 @@ static int gfar_alloc_skb_resources(struct net_device *ndev)
int i;
struct gfar_private *priv = netdev_priv(ndev);
struct device *dev = &priv->ofdev->dev;
+ struct gfar_priv_tx_q *tx_queue = NULL;
+ struct gfar_priv_rx_q *rx_queue = NULL;
+
+ tx_queue = priv->tx_queue;
+ rx_queue = priv->rx_queue;
/* Allocate memory for the buffer descriptors */
vaddr = dma_alloc_coherent(dev,
- sizeof(*priv->tx_bd_base) * priv->tx_ring_size +
- sizeof(*priv->rx_bd_base) * priv->rx_ring_size,
- &priv->tx_bd_dma_base, GFP_KERNEL);
+ sizeof(*tx_queue->tx_bd_base) * tx_queue->tx_ring_size +
+ sizeof(*rx_queue->rx_bd_base) * rx_queue->rx_ring_size,
+ &tx_queue->tx_bd_dma_base, GFP_KERNEL);
if (!vaddr) {
if (netif_msg_ifup(priv))
pr_err("%s: Could not allocate buffer descriptors!\n",
@@ -233,36 +244,38 @@ static int gfar_alloc_skb_resources(struct net_device *ndev)
return -ENOMEM;
}
- priv->tx_bd_base = vaddr;
+ tx_queue->tx_bd_base = vaddr;
+ tx_queue->dev = ndev;
/* Start the rx descriptor ring where the tx ring leaves off */
- vaddr = vaddr + sizeof(*priv->tx_bd_base) * priv->tx_ring_size;
- priv->rx_bd_base = vaddr;
+ vaddr = vaddr + sizeof(*tx_queue->tx_bd_base) * tx_queue->tx_ring_size;
+ rx_queue->rx_bd_base = vaddr;
+ rx_queue->dev = ndev;
/* Setup the skbuff rings */
- priv->tx_skbuff = kmalloc(sizeof(*priv->tx_skbuff) *
- priv->tx_ring_size, GFP_KERNEL);
- if (!priv->tx_skbuff) {
+ tx_queue->tx_skbuff = kmalloc(sizeof(*tx_queue->tx_skbuff) *
+ tx_queue->tx_ring_size, GFP_KERNEL);
+ if (!tx_queue->tx_skbuff) {
if (netif_msg_ifup(priv))
pr_err("%s: Could not allocate tx_skbuff\n",
ndev->name);
goto cleanup;
}
- for (i = 0; i < priv->tx_ring_size; i++)
- priv->tx_skbuff[i] = NULL;
+ for (i = 0; i < tx_queue->tx_ring_size; i++)
+ tx_queue->tx_skbuff[i] = NULL;
- priv->rx_skbuff = kmalloc(sizeof(*priv->rx_skbuff) *
- priv->rx_ring_size, GFP_KERNEL);
- if (!priv->rx_skbuff) {
+ rx_queue->rx_skbuff = kmalloc(sizeof(*rx_queue->rx_skbuff) *
+ rx_queue->rx_ring_size, GFP_KERNEL);
+ if (!rx_queue->rx_skbuff) {
if (netif_msg_ifup(priv))
pr_err("%s: Could not allocate rx_skbuff\n",
ndev->name);
goto cleanup;
}
- for (i = 0; i < priv->rx_ring_size; i++)
- priv->rx_skbuff[i] = NULL;
+ for (i = 0; i < rx_queue->rx_ring_size; i++)
+ rx_queue->rx_skbuff[i] = NULL;
if (gfar_init_bds(ndev))
goto cleanup;
@@ -278,24 +291,29 @@ static void gfar_init_mac(struct net_device *ndev)
{
struct gfar_private *priv = netdev_priv(ndev);
struct gfar __iomem *regs = priv->regs;
+ struct gfar_priv_tx_q *tx_queue = NULL;
+ struct gfar_priv_rx_q *rx_queue = NULL;
u32 rctrl = 0;
u32 tctrl = 0;
u32 attrs = 0;
+ tx_queue = priv->tx_queue;
+ rx_queue = priv->rx_queue;
+
/* enet DMA only understands physical addresses */
- gfar_write(®s->tbase0, priv->tx_bd_dma_base);
- gfar_write(®s->rbase0, priv->tx_bd_dma_base +
- sizeof(*priv->tx_bd_base) *
- priv->tx_ring_size);
+ gfar_write(®s->tbase0, tx_queue->tx_bd_dma_base);
+ gfar_write(®s->rbase0, tx_queue->tx_bd_dma_base +
+ sizeof(*tx_queue->tx_bd_base) *
+ tx_queue->tx_ring_size);
/* Configure the coalescing support */
gfar_write(®s->txic, 0);
- if (priv->txcoalescing)
- gfar_write(®s->txic, priv->txic);
+ if (tx_queue->txcoalescing)
+ gfar_write(®s->txic, tx_queue->txic);
gfar_write(®s->rxic, 0);
- if (priv->rxcoalescing)
- gfar_write(®s->rxic, priv->rxic);
+ if (rx_queue->rxcoalescing)
+ gfar_write(®s->rxic, rx_queue->rxic);
if (priv->rx_csum_enable)
rctrl |= RCTRL_CHECKSUMMING;
@@ -414,7 +432,7 @@ static int gfar_of_init(struct net_device *dev)
stash = of_get_property(np, "bd-stash", NULL);
- if(stash) {
+ if (stash) {
priv->device_flags |= FSL_GIANFAR_DEV_HAS_BD_STASHING;
priv->bd_stash_en = 1;
}
@@ -519,8 +537,18 @@ static int gfar_probe(struct of_device *ofdev,
if (err)
goto regs_fail;
- spin_lock_init(&priv->txlock);
- spin_lock_init(&priv->rxlock);
+ priv->tx_queue = (struct gfar_priv_tx_q *)kmalloc(
+ sizeof (struct gfar_priv_tx_q), GFP_KERNEL);
+ if (!priv->tx_queue)
+ goto regs_fail;
+
+ priv->rx_queue = (struct gfar_priv_rx_q *)kmalloc(
+ sizeof (struct gfar_priv_rx_q), GFP_KERNEL);
+ if (!priv->rx_queue)
+ goto rx_queue_fail;
+
+ spin_lock_init(&priv->tx_queue->txlock);
+ spin_lock_init(&priv->rx_queue->rxlock);
spin_lock_init(&priv->bflock);
INIT_WORK(&priv->reset_task, gfar_reset_task);
@@ -552,12 +580,13 @@ static int gfar_probe(struct of_device *ofdev,
/* Fill in the dev structure */
dev->watchdog_timeo = TX_TIMEOUT;
- netif_napi_add(dev, &priv->napi, gfar_poll, GFAR_DEV_WEIGHT);
dev->mtu = 1500;
-
dev->netdev_ops = &gfar_netdev_ops;
dev->ethtool_ops = &gfar_ethtool_ops;
+ /* Register for napi ...NAPI is for each rx_queue */
+ netif_napi_add(dev, &priv->rx_queue->napi, gfar_poll, GFAR_DEV_WEIGHT);
+
if (priv->device_flags & FSL_GIANFAR_DEV_HAS_CSUM) {
priv->rx_csum_enable = 1;
dev->features |= NETIF_F_IP_CSUM | NETIF_F_SG | NETIF_F_HIGHDMA;
@@ -613,14 +642,16 @@ static int gfar_probe(struct of_device *ofdev,
dev->hard_header_len += GMAC_FCB_LEN;
priv->rx_buffer_size = DEFAULT_RX_BUFFER_SIZE;
- priv->tx_ring_size = DEFAULT_TX_RING_SIZE;
- priv->rx_ring_size = DEFAULT_RX_RING_SIZE;
- priv->num_txbdfree = DEFAULT_TX_RING_SIZE;
- priv->txcoalescing = DEFAULT_TX_COALESCE;
- priv->txic = DEFAULT_TXIC;
- priv->rxcoalescing = DEFAULT_RX_COALESCE;
- priv->rxic = DEFAULT_RXIC;
+ /* Initializing some of the rx/tx queue level parameters */
+ priv->tx_queue->tx_ring_size = DEFAULT_TX_RING_SIZE;
+ priv->tx_queue->num_txbdfree = DEFAULT_TX_RING_SIZE;
+ priv->tx_queue->txcoalescing = DEFAULT_TX_COALESCE;
+ priv->tx_queue->txic = DEFAULT_TXIC;
+
+ priv->rx_queue->rx_ring_size = DEFAULT_RX_RING_SIZE;
+ priv->rx_queue->rxcoalescing = DEFAULT_RX_COALESCE;
+ priv->rx_queue->rxic = DEFAULT_RXIC;
/* Enable most messages by default */
priv->msg_enable = (NETIF_MSG_IFUP << 1 ) - 1;
@@ -666,12 +697,15 @@ static int gfar_probe(struct of_device *ofdev,
/* provided which set of benchmarks. */
printk(KERN_INFO "%s: Running with NAPI enabled\n", dev->name);
printk(KERN_INFO "%s: %d/%d RX/TX BD ring size\n",
- dev->name, priv->rx_ring_size, priv->tx_ring_size);
+ dev->name, priv->rx_queue->rx_ring_size, priv->tx_queue->tx_ring_size);
return 0;
register_fail:
iounmap(priv->regs);
+ kfree(priv->rx_queue);
+rx_queue_fail:
+ kfree(priv->tx_queue);
regs_fail:
if (priv->phy_node)
of_node_put(priv->phy_node);
@@ -705,6 +739,8 @@ static int gfar_suspend(struct device *dev)
{
struct gfar_private *priv = dev_get_drvdata(dev);
struct net_device *ndev = priv->ndev;
+ struct gfar_priv_tx_q *tx_queue = NULL;
+ struct gfar_priv_rx_q *rx_queue = NULL;
unsigned long flags;
u32 tempval;
@@ -712,10 +748,12 @@ static int gfar_suspend(struct device *dev)
(priv->device_flags & FSL_GIANFAR_DEV_HAS_MAGIC_PACKET);
netif_device_detach(ndev);
+ tx_queue = priv->tx_queue;
+ rx_queue = priv->rx_queue;
if (netif_running(ndev)) {
- spin_lock_irqsave(&priv->txlock, flags);
- spin_lock(&priv->rxlock);
+ spin_lock_irqsave(&tx_queue->txlock, flags);
+ spin_lock(&rx_queue->rxlock);
gfar_halt_nodisable(ndev);
@@ -729,10 +767,10 @@ static int gfar_suspend(struct device *dev)
gfar_write(&priv->regs->maccfg1, tempval);
- spin_unlock(&priv->rxlock);
- spin_unlock_irqrestore(&priv->txlock, flags);
+ spin_unlock(&rx_queue->rxlock);
+ spin_unlock_irqrestore(&tx_queue->txlock, flags);
- napi_disable(&priv->napi);
+ napi_disable(&rx_queue->napi);
if (magic_packet) {
/* Enable interrupt on Magic Packet */
@@ -754,6 +792,8 @@ static int gfar_resume(struct device *dev)
{
struct gfar_private *priv = dev_get_drvdata(dev);
struct net_device *ndev = priv->ndev;
+ struct gfar_priv_tx_q *tx_queue = NULL;
+ struct gfar_priv_rx_q *rx_queue = NULL;
unsigned long flags;
u32 tempval;
int magic_packet = priv->wol_en &&
@@ -770,9 +810,11 @@ static int gfar_resume(struct device *dev)
/* Disable Magic Packet mode, in case something
* else woke us up.
*/
+ rx_queue = priv->rx_queue;
+ tx_queue = priv->tx_queue;
- spin_lock_irqsave(&priv->txlock, flags);
- spin_lock(&priv->rxlock);
+ spin_lock_irqsave(&tx_queue->txlock, flags);
+ spin_lock(&rx_queue->rxlock);
tempval = gfar_read(&priv->regs->maccfg2);
tempval &= ~MACCFG2_MPEN;
@@ -780,12 +822,12 @@ static int gfar_resume(struct device *dev)
gfar_start(ndev);
- spin_unlock(&priv->rxlock);
- spin_unlock_irqrestore(&priv->txlock, flags);
+ spin_unlock(&rx_queue->rxlock);
+ spin_unlock_irqrestore(&tx_queue->txlock, flags);
netif_device_attach(ndev);
- napi_enable(&priv->napi);
+ napi_enable(&rx_queue->napi);
return 0;
}
@@ -1060,18 +1102,23 @@ void gfar_halt(struct net_device *dev)
void stop_gfar(struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_tx_q *tx_queue = NULL;
+ struct gfar_priv_rx_q *rx_queue = NULL;
unsigned long flags;
phy_stop(priv->phydev);
+ tx_queue = priv->tx_queue;
+ rx_queue = priv->rx_queue;
+
/* Lock it down */
- spin_lock_irqsave(&priv->txlock, flags);
- spin_lock(&priv->rxlock);
+ spin_lock_irqsave(&tx_queue->txlock, flags);
+ spin_lock(&rx_queue->rxlock);
gfar_halt(dev);
- spin_unlock(&priv->rxlock);
- spin_unlock_irqrestore(&priv->txlock, flags);
+ spin_unlock(&rx_queue->rxlock);
+ spin_unlock_irqrestore(&tx_queue->txlock, flags);
/* Free the IRQs */
if (priv->device_flags & FSL_GIANFAR_DEV_HAS_MULTI_INTR) {
@@ -1092,46 +1139,50 @@ static void free_skb_resources(struct gfar_private *priv)
struct device *dev = &priv->ofdev->dev;
struct rxbd8 *rxbdp;
struct txbd8 *txbdp;
+ struct gfar_priv_tx_q *tx_queue = NULL;
+ struct gfar_priv_rx_q *rx_queue = NULL;
int i, j;
/* Go through all the buffer descriptors and free their data buffers */
- txbdp = priv->tx_bd_base;
+ tx_queue = priv->tx_queue;
+ txbdp = tx_queue->tx_bd_base;
- if (!priv->tx_skbuff)
+ if (!tx_queue->tx_skbuff)
goto skip_tx_skbuff;
- for (i = 0; i < priv->tx_ring_size; i++) {
- if (!priv->tx_skbuff[i])
+ for (i = 0; i < tx_queue->tx_ring_size; i++) {
+ if (!tx_queue->tx_skbuff[i])
continue;
dma_unmap_single(&priv->ofdev->dev, txbdp->bufPtr,
txbdp->length, DMA_TO_DEVICE);
txbdp->lstatus = 0;
- for (j = 0; j < skb_shinfo(priv->tx_skbuff[i])->nr_frags; j++) {
+ for (j = 0; j < skb_shinfo(tx_queue->tx_skbuff[i])->nr_frags; j++) {
txbdp++;
dma_unmap_page(&priv->ofdev->dev, txbdp->bufPtr,
txbdp->length, DMA_TO_DEVICE);
}
txbdp++;
- dev_kfree_skb_any(priv->tx_skbuff[i]);
- priv->tx_skbuff[i] = NULL;
+ dev_kfree_skb_any(tx_queue->tx_skbuff[i]);
+ tx_queue->tx_skbuff[i] = NULL;
}
- kfree(priv->tx_skbuff);
+ kfree(tx_queue->tx_skbuff);
skip_tx_skbuff:
- rxbdp = priv->rx_bd_base;
+ rx_queue = priv->rx_queue;
+ rxbdp = rx_queue->rx_bd_base;
- if (!priv->rx_skbuff)
+ if (!rx_queue->rx_skbuff)
goto skip_rx_skbuff;
- for (i = 0; i < priv->rx_ring_size; i++) {
- if (priv->rx_skbuff[i]) {
+ for (i = 0; i < rx_queue->rx_ring_size; i++) {
+ if (rx_queue->rx_skbuff[i]) {
dma_unmap_single(&priv->ofdev->dev, rxbdp->bufPtr,
priv->rx_buffer_size,
DMA_FROM_DEVICE);
- dev_kfree_skb_any(priv->rx_skbuff[i]);
- priv->rx_skbuff[i] = NULL;
+ dev_kfree_skb_any(rx_queue->rx_skbuff[i]);
+ rx_queue->rx_skbuff[i] = NULL;
}
rxbdp->lstatus = 0;
@@ -1139,17 +1190,19 @@ skip_tx_skbuff:
rxbdp++;
}
- kfree(priv->rx_skbuff);
+ kfree(rx_queue->rx_skbuff);
skip_rx_skbuff:
- dma_free_coherent(dev, sizeof(*txbdp) * priv->tx_ring_size +
- sizeof(*rxbdp) * priv->rx_ring_size,
- priv->tx_bd_base, priv->tx_bd_dma_base);
+ dma_free_coherent(dev, sizeof(*txbdp) * tx_queue->tx_ring_size +
+ sizeof(*rxbdp) * rx_queue->rx_ring_size,
+ tx_queue->tx_bd_base, tx_queue->tx_bd_dma_base);
}
void gfar_start(struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_tx_q *tx_queue;
+ struct gfar_priv_rx_q *rx_queue;
struct gfar __iomem *regs = priv->regs;
u32 tempval;
@@ -1258,7 +1311,7 @@ static int gfar_enet_open(struct net_device *dev)
struct gfar_private *priv = netdev_priv(dev);
int err;
- napi_enable(&priv->napi);
+ napi_enable(&priv->rx_queue->napi);
skb_queue_head_init(&priv->rx_recycle);
@@ -1269,14 +1322,14 @@ static int gfar_enet_open(struct net_device *dev)
err = init_phy(dev);
- if(err) {
- napi_disable(&priv->napi);
+ if (err) {
+ napi_disable(&priv->rx_queue->napi);
return err;
}
err = startup_gfar(dev);
if (err) {
- napi_disable(&priv->napi);
+ napi_disable(&priv->rx_queue->napi);
return err;
}
@@ -1349,6 +1402,7 @@ static inline struct txbd8 *next_txbd(struct txbd8 *bdp, struct txbd8 *base,
static int gfar_start_xmit(struct sk_buff *skb, struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_tx_q *tx_queue = NULL;
struct txfcb *fcb = NULL;
struct txbd8 *txbdp, *txbdp_start, *base;
u32 lstatus;
@@ -1357,7 +1411,8 @@ static int gfar_start_xmit(struct sk_buff *skb, struct net_device *dev)
unsigned long flags;
unsigned int nr_frags, length;
- base = priv->tx_bd_base;
+ tx_queue = priv->tx_queue;
+ base = tx_queue->tx_bd_base;
/* make space for additional header when fcb is needed */
if (((skb->ip_summed == CHECKSUM_PARTIAL) ||
@@ -1378,21 +1433,21 @@ static int gfar_start_xmit(struct sk_buff *skb, struct net_device *dev)
/* total number of fragments in the SKB */
nr_frags = skb_shinfo(skb)->nr_frags;
- spin_lock_irqsave(&priv->txlock, flags);
+ spin_lock_irqsave(&tx_queue->txlock, flags);
/* check if there is space to queue this packet */
- if ((nr_frags+1) > priv->num_txbdfree) {
+ if ((nr_frags+1) > tx_queue->num_txbdfree) {
/* no space, stop the queue */
netif_stop_queue(dev);
dev->stats.tx_fifo_errors++;
- spin_unlock_irqrestore(&priv->txlock, flags);
+ spin_unlock_irqrestore(&tx_queue->txlock, flags);
return NETDEV_TX_BUSY;
}
/* Update transmit stats */
dev->stats.tx_bytes += skb->len;
- txbdp = txbdp_start = priv->cur_tx;
+ txbdp = txbdp_start = tx_queue->cur_tx;
if (nr_frags == 0) {
lstatus = txbdp->lstatus | BD_LFLAG(TXBD_LAST | TXBD_INTERRUPT);
@@ -1400,7 +1455,7 @@ static int gfar_start_xmit(struct sk_buff *skb, struct net_device *dev)
/* Place the fragment addresses and lengths into the TxBDs */
for (i = 0; i < nr_frags; i++) {
/* Point at the next BD, wrapping as needed */
- txbdp = next_txbd(txbdp, base, priv->tx_ring_size);
+ txbdp = next_txbd(txbdp, base, tx_queue->tx_ring_size);
length = skb_shinfo(skb)->frags[i].size;
@@ -1442,7 +1497,7 @@ static int gfar_start_xmit(struct sk_buff *skb, struct net_device *dev)
}
/* setup the TxBD length and buffer pointer for the first BD */
- priv->tx_skbuff[priv->skb_curtx] = skb;
+ tx_queue->tx_skbuff[tx_queue->skb_curtx] = skb;
txbdp_start->bufPtr = dma_map_single(&priv->ofdev->dev, skb->data,
skb_headlen(skb), DMA_TO_DEVICE);
@@ -1462,19 +1517,19 @@ static int gfar_start_xmit(struct sk_buff *skb, struct net_device *dev)
/* Update the current skb pointer to the next entry we will use
* (wrapping if necessary) */
- priv->skb_curtx = (priv->skb_curtx + 1) &
- TX_RING_MOD_MASK(priv->tx_ring_size);
+ tx_queue->skb_curtx = (tx_queue->skb_curtx + 1) &
+ TX_RING_MOD_MASK(tx_queue->tx_ring_size);
- priv->cur_tx = next_txbd(txbdp, base, priv->tx_ring_size);
+ tx_queue->cur_tx = next_txbd(txbdp, base, tx_queue->tx_ring_size);
/* reduce TxBD free count */
- priv->num_txbdfree -= (nr_frags + 1);
+ tx_queue->num_txbdfree -= (nr_frags + 1);
dev->trans_start = jiffies;
/* If the next BD still needs to be cleaned up, then the bds
are full. We need to tell the kernel to stop sending us stuff. */
- if (!priv->num_txbdfree) {
+ if (!tx_queue->num_txbdfree) {
netif_stop_queue(dev);
dev->stats.tx_fifo_errors++;
@@ -1484,7 +1539,7 @@ static int gfar_start_xmit(struct sk_buff *skb, struct net_device *dev)
gfar_write(&priv->regs->tstat, TSTAT_CLEAR_THALT);
/* Unlock priv */
- spin_unlock_irqrestore(&priv->txlock, flags);
+ spin_unlock_irqrestore(&tx_queue->txlock, flags);
return NETDEV_TX_OK;
}
@@ -1494,7 +1549,7 @@ static int gfar_close(struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
- napi_disable(&priv->napi);
+ napi_disable(&priv->rx_queue->napi);
skb_queue_purge(&priv->rx_recycle);
cancel_work_sync(&priv->reset_task);
@@ -1523,10 +1578,12 @@ static void gfar_vlan_rx_register(struct net_device *dev,
struct vlan_group *grp)
{
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_rx_q *rx_queue = NULL;
unsigned long flags;
u32 tempval;
- spin_lock_irqsave(&priv->rxlock, flags);
+ rx_queue = priv->rx_queue;
+ spin_lock_irqsave(&rx_queue->rxlock, flags);
priv->vlgrp = grp;
@@ -1560,7 +1617,7 @@ static void gfar_vlan_rx_register(struct net_device *dev,
gfar_change_mtu(dev, dev->mtu);
- spin_unlock_irqrestore(&priv->rxlock, flags);
+ spin_unlock_irqrestore(&rx_queue->rxlock, flags);
}
static int gfar_change_mtu(struct net_device *dev, int new_mtu)
@@ -1649,24 +1706,27 @@ static void gfar_timeout(struct net_device *dev)
}
/* Interrupt Handler for Transmit complete */
-static int gfar_clean_tx_ring(struct net_device *dev)
+static int gfar_clean_tx_ring(struct gfar_priv_tx_q *tx_queue)
{
+ struct net_device *dev = tx_queue->dev;
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_rx_q *rx_queue = NULL;
struct txbd8 *bdp;
struct txbd8 *lbdp = NULL;
- struct txbd8 *base = priv->tx_bd_base;
+ struct txbd8 *base = tx_queue->tx_bd_base;
struct sk_buff *skb;
int skb_dirtytx;
- int tx_ring_size = priv->tx_ring_size;
+ int tx_ring_size = tx_queue->tx_ring_size;
int frags = 0;
int i;
int howmany = 0;
u32 lstatus;
- bdp = priv->dirty_tx;
- skb_dirtytx = priv->skb_dirtytx;
+ rx_queue = priv->rx_queue;
+ bdp = tx_queue->dirty_tx;
+ skb_dirtytx = tx_queue->skb_dirtytx;
- while ((skb = priv->tx_skbuff[skb_dirtytx])) {
+ while ((skb = tx_queue->tx_skbuff[skb_dirtytx])) {
frags = skb_shinfo(skb)->nr_frags;
lbdp = skip_txbd(bdp, frags, base, tx_ring_size);
@@ -1698,29 +1758,29 @@ static int gfar_clean_tx_ring(struct net_device *dev)
* If there's room in the queue (limit it to rx_buffer_size)
* we add this skb back into the pool, if it's the right size
*/
- if (skb_queue_len(&priv->rx_recycle) < priv->rx_ring_size &&
+ if (skb_queue_len(&priv->rx_recycle) < rx_queue->rx_ring_size &&
skb_recycle_check(skb, priv->rx_buffer_size +
RXBUF_ALIGNMENT))
__skb_queue_head(&priv->rx_recycle, skb);
else
dev_kfree_skb_any(skb);
- priv->tx_skbuff[skb_dirtytx] = NULL;
+ tx_queue->tx_skbuff[skb_dirtytx] = NULL;
skb_dirtytx = (skb_dirtytx + 1) &
TX_RING_MOD_MASK(tx_ring_size);
howmany++;
- priv->num_txbdfree += frags + 1;
+ tx_queue->num_txbdfree += frags + 1;
}
/* If we freed a buffer, we can restart transmission, if necessary */
- if (netif_queue_stopped(dev) && priv->num_txbdfree)
+ if (netif_queue_stopped(dev) && tx_queue->num_txbdfree)
netif_wake_queue(dev);
/* Update dirty indicators */
- priv->skb_dirtytx = skb_dirtytx;
- priv->dirty_tx = bdp;
+ tx_queue->skb_dirtytx = skb_dirtytx;
+ tx_queue->dirty_tx = bdp;
dev->stats.tx_packets += howmany;
@@ -1730,14 +1790,18 @@ static int gfar_clean_tx_ring(struct net_device *dev)
static void gfar_schedule_cleanup(struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_tx_q *tx_queue = NULL;
+ struct gfar_priv_rx_q *rx_queue = NULL;
unsigned long flags;
- spin_lock_irqsave(&priv->txlock, flags);
- spin_lock(&priv->rxlock);
+ rx_queue = priv->rx_queue;
+ tx_queue = priv->tx_queue;
+ spin_lock_irqsave(&tx_queue->txlock, flags);
+ spin_lock(&rx_queue->rxlock);
- if (napi_schedule_prep(&priv->napi)) {
+ if (napi_schedule_prep(&rx_queue->napi)) {
gfar_write(&priv->regs->imask, IMASK_RTX_DISABLED);
- __napi_schedule(&priv->napi);
+ __napi_schedule(&rx_queue->napi);
} else {
/*
* Clear IEVENT, so interrupts aren't called again
@@ -1746,8 +1810,8 @@ static void gfar_schedule_cleanup(struct net_device *dev)
gfar_write(&priv->regs->ievent, IEVENT_RTX_MASK);
}
- spin_unlock(&priv->rxlock);
- spin_unlock_irqrestore(&priv->txlock, flags);
+ spin_unlock(&rx_queue->rxlock);
+ spin_unlock_irqrestore(&tx_queue->txlock, flags);
}
/* Interrupt Handler for Transmit complete */
@@ -1757,15 +1821,16 @@ static irqreturn_t gfar_transmit(int irq, void *dev_id)
return IRQ_HANDLED;
}
-static void gfar_new_rxbdp(struct net_device *dev, struct rxbd8 *bdp,
+static void gfar_new_rxbdp(struct gfar_priv_rx_q *rx_queue, struct rxbd8 *bdp,
struct sk_buff *skb)
{
+ struct net_device *dev = rx_queue->dev;
struct gfar_private *priv = netdev_priv(dev);
dma_addr_t buf;
buf = dma_map_single(&priv->ofdev->dev, skb->data,
priv->rx_buffer_size, DMA_FROM_DEVICE);
- gfar_init_rxbdp(dev, bdp, buf);
+ gfar_init_rxbdp(rx_queue, bdp, buf);
}
@@ -1890,8 +1955,9 @@ static int gfar_process_frame(struct net_device *dev, struct sk_buff *skb,
* until the budget/quota has been reached. Returns the number
* of frames handled
*/
-int gfar_clean_rx_ring(struct net_device *dev, int rx_work_limit)
+int gfar_clean_rx_ring(struct gfar_priv_rx_q *rx_queue, int rx_work_limit)
{
+ struct net_device *dev = rx_queue->dev;
struct rxbd8 *bdp, *base;
struct sk_buff *skb;
int pkt_len;
@@ -1900,8 +1966,8 @@ int gfar_clean_rx_ring(struct net_device *dev, int rx_work_limit)
struct gfar_private *priv = netdev_priv(dev);
/* Get the first full descriptor */
- bdp = priv->cur_rx;
- base = priv->rx_bd_base;
+ bdp = rx_queue->cur_rx;
+ base = rx_queue->rx_bd_base;
amount_pull = (gfar_uses_fcb(priv) ? GMAC_FCB_LEN : 0) +
priv->padding;
@@ -1913,7 +1979,7 @@ int gfar_clean_rx_ring(struct net_device *dev, int rx_work_limit)
/* Add another skb for the future */
newskb = gfar_new_skb(dev);
- skb = priv->rx_skbuff[priv->skb_currx];
+ skb = rx_queue->rx_skbuff[rx_queue->skb_currx];
dma_unmap_single(&priv->ofdev->dev, bdp->bufPtr,
priv->rx_buffer_size, DMA_FROM_DEVICE);
@@ -1961,30 +2027,33 @@ int gfar_clean_rx_ring(struct net_device *dev, int rx_work_limit)
}
- priv->rx_skbuff[priv->skb_currx] = newskb;
+ rx_queue->rx_skbuff[rx_queue->skb_currx] = newskb;
/* Setup the new bdp */
- gfar_new_rxbdp(dev, bdp, newskb);
+ gfar_new_rxbdp(rx_queue, bdp, newskb);
/* Update to the next pointer */
- bdp = next_bd(bdp, base, priv->rx_ring_size);
+ bdp = next_bd(bdp, base, rx_queue->rx_ring_size);
/* update to point at the next skb */
- priv->skb_currx =
- (priv->skb_currx + 1) &
- RX_RING_MOD_MASK(priv->rx_ring_size);
+ rx_queue->skb_currx =
+ (rx_queue->skb_currx + 1) &
+ RX_RING_MOD_MASK(rx_queue->rx_ring_size);
}
/* Update the current rxbd pointer to be the next one */
- priv->cur_rx = bdp;
+ rx_queue->cur_rx = bdp;
return howmany;
}
static int gfar_poll(struct napi_struct *napi, int budget)
{
- struct gfar_private *priv = container_of(napi, struct gfar_private, napi);
- struct net_device *dev = priv->ndev;
+ struct gfar_priv_rx_q *rx_queue = container_of(napi,
+ struct gfar_priv_rx_q, napi);
+ struct net_device *dev = rx_queue->dev;
+ struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_tx_q *tx_queue = NULL;
int tx_cleaned = 0;
int rx_cleaned = 0;
unsigned long flags;
@@ -1992,14 +2061,15 @@ static int gfar_poll(struct napi_struct *napi, int budget)
/* Clear IEVENT, so interrupts aren't called again
* because of the packets that have already arrived */
gfar_write(&priv->regs->ievent, IEVENT_RTX_MASK);
+ tx_queue = priv->tx_queue;
/* If we fail to get the lock, don't bother with the TX BDs */
- if (spin_trylock_irqsave(&priv->txlock, flags)) {
- tx_cleaned = gfar_clean_tx_ring(dev);
- spin_unlock_irqrestore(&priv->txlock, flags);
+ if (spin_trylock_irqsave(&tx_queue->txlock, flags)) {
+ tx_cleaned = gfar_clean_tx_ring(tx_queue);
+ spin_unlock_irqrestore(&tx_queue->txlock, flags);
}
- rx_cleaned = gfar_clean_rx_ring(dev, budget);
+ rx_cleaned = gfar_clean_rx_ring(rx_queue, budget);
if (tx_cleaned)
return budget;
@@ -2014,13 +2084,13 @@ static int gfar_poll(struct napi_struct *napi, int budget)
/* If we are coalescing interrupts, update the timer */
/* Otherwise, clear it */
- if (likely(priv->rxcoalescing)) {
+ if (likely(rx_queue->rxcoalescing)) {
gfar_write(&priv->regs->rxic, 0);
- gfar_write(&priv->regs->rxic, priv->rxic);
+ gfar_write(&priv->regs->rxic, rx_queue->rxic);
}
- if (likely(priv->txcoalescing)) {
+ if (likely(tx_queue->txcoalescing)) {
gfar_write(&priv->regs->txic, 0);
- gfar_write(&priv->regs->txic, priv->txic);
+ gfar_write(&priv->regs->txic, tx_queue->txic);
}
}
@@ -2087,12 +2157,14 @@ static irqreturn_t gfar_interrupt(int irq, void *dev_id)
static void adjust_link(struct net_device *dev)
{
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_tx_q *tx_queue = NULL;
struct gfar __iomem *regs = priv->regs;
unsigned long flags;
struct phy_device *phydev = priv->phydev;
int new_state = 0;
- spin_lock_irqsave(&priv->txlock, flags);
+ tx_queue = priv->tx_queue;
+ spin_lock_irqsave(&tx_queue->txlock, flags);
if (phydev->link) {
u32 tempval = gfar_read(®s->maccfg2);
u32 ecntrl = gfar_read(®s->ecntrl);
@@ -2158,7 +2230,7 @@ static void adjust_link(struct net_device *dev)
if (new_state && netif_msg_link(priv))
phy_print_status(phydev);
- spin_unlock_irqrestore(&priv->txlock, flags);
+ spin_unlock_irqrestore(&tx_queue->txlock, flags);
}
/* Update the hash table based on the current list of multicast
@@ -2172,7 +2244,7 @@ static void gfar_set_multi(struct net_device *dev)
struct gfar __iomem *regs = priv->regs;
u32 tempval;
- if(dev->flags & IFF_PROMISC) {
+ if (dev->flags & IFF_PROMISC) {
/* Set RCTRL to PROM */
tempval = gfar_read(®s->rctrl);
tempval |= RCTRL_PROM;
@@ -2184,7 +2256,7 @@ static void gfar_set_multi(struct net_device *dev)
gfar_write(®s->rctrl, tempval);
}
- if(dev->flags & IFF_ALLMULTI) {
+ if (dev->flags & IFF_ALLMULTI) {
/* Set the hash to rx all multicast frames */
gfar_write(®s->igaddr0, 0xffffffff);
gfar_write(®s->igaddr1, 0xffffffff);
@@ -2236,7 +2308,7 @@ static void gfar_set_multi(struct net_device *dev)
em_num = 0;
}
- if(dev->mc_count == 0)
+ if (dev->mc_count == 0)
return;
/* Parse the list, and set the appropriate bits */
diff --git a/drivers/net/gianfar.h b/drivers/net/gianfar.h
index 05732fa..a60f93f 100644
--- a/drivers/net/gianfar.h
+++ b/drivers/net/gianfar.h
@@ -7,8 +7,9 @@
*
* Author: Andy Fleming
* Maintainer: Kumar Gala
+ * Modifier: Sandeep Gopalpet <sandeep.kumar@freescale.com>
*
- * Copyright (c) 2002-2004 Freescale Semiconductor, Inc.
+ * Copyright 2002-2009 Freescale Semiconductor, Inc.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
@@ -699,6 +700,76 @@ struct gfar {
#define FSL_GIANFAR_DEV_HAS_BD_STASHING 0x00000200
#define FSL_GIANFAR_DEV_HAS_BUF_STASHING 0x00000400
+/**
+ * struct gfar_priv_tx_q - per tx queue structure
+ * @txlock: per queue tx spin lock
+ * @tx_skbuff:skb pointers
+ * @skb_curtx: to be used skb pointer
+ * @skb_dirtytx:the last used skb pointer
+ * @qindex: index of this queue
+ * @dev: back pointer to the dev structure
+ * @grp: back pointer to the group to which this queue belongs
+ * @tx_bd_base: First tx buffer descriptor
+ * @cur_tx: Next free ring entry
+ * @dirty_tx: First buffer in line to be transmitted
+ * @tx_ring_size: Tx ring size
+ * @num_txbdfree: number of free TxBds
+ * @txcoalescing: enable/disable tx coalescing
+ * @txic: transmit interrupt coalescing value
+ * @txcount: coalescing value if based on tx frame count
+ * @txtime: coalescing value if based on time
+ */
+struct gfar_priv_tx_q {
+ spinlock_t txlock __attribute__ ((aligned (SMP_CACHE_BYTES)));
+ struct sk_buff ** tx_skbuff;
+ /* Buffer descriptor pointers */
+ dma_addr_t tx_bd_dma_base;
+ struct txbd8 *tx_bd_base;
+ struct txbd8 *cur_tx;
+ struct txbd8 *dirty_tx;
+ struct net_device *dev;
+ u16 skb_curtx;
+ u16 skb_dirtytx;
+ u16 qindex;
+ unsigned int tx_ring_size;
+ unsigned int num_txbdfree;
+ /* Configuration info for the coalescing features */
+ unsigned char txcoalescing;
+ unsigned long txic;
+ unsigned short txcount;
+ unsigned short txtime;
+};
+
+/**
+ * struct gfar_priv_rx_q - per rx queue structure
+ * @rxlock: per queue rx spin lock
+ * @napi: the napi poll function
+ * @rx_skbuff: skb pointers
+ * @skb_currx: currently use skb pointer
+ * @rx_bd_base: First rx buffer descriptor
+ * @cur_rx: Next free rx ring entry
+ * @qindex: index of this queue
+ * @dev: back pointer to the dev structure
+ * @rx_ring_size: Rx ring size
+ * @rxcoalescing: enable/disable rx-coalescing
+ * @rxic: receive interrupt coalescing vlaue
+ */
+
+struct gfar_priv_rx_q {
+ spinlock_t rxlock __attribute__ ((aligned (SMP_CACHE_BYTES)));
+ struct napi_struct napi;
+ struct sk_buff ** rx_skbuff;
+ struct rxbd8 *rx_bd_base;
+ struct rxbd8 *cur_rx;
+ struct net_device *dev;
+ u16 skb_currx;
+ u16 qindex;
+ unsigned int rx_ring_size;
+ /* RX Coalescing values */
+ unsigned char rxcoalescing;
+ unsigned long rxic;
+};
+
/* Struct stolen almost completely (and shamelessly) from the FCC enet source
* (Ok, that's not so true anymore, but there is a family resemblence)
* The GFAR buffer descriptors track the ring buffers. The rx_bd_base
@@ -709,52 +780,15 @@ struct gfar {
* the buffer descriptor determines the actual condition.
*/
struct gfar_private {
- /* Fields controlled by TX lock */
- spinlock_t txlock;
-
- /* Pointer to the array of skbuffs */
- struct sk_buff ** tx_skbuff;
-
- /* next free skb in the array */
- u16 skb_curtx;
-
- /* First skb in line to be transmitted */
- u16 skb_dirtytx;
-
- /* Configuration info for the coalescing features */
- unsigned char txcoalescing;
- unsigned long txic;
-
- /* Buffer descriptor pointers */
- dma_addr_t tx_bd_dma_base;
- struct txbd8 *tx_bd_base; /* First tx buffer descriptor */
- struct txbd8 *cur_tx; /* Next free ring entry */
- struct txbd8 *dirty_tx; /* First buffer in line
- to be transmitted */
- unsigned int tx_ring_size;
- unsigned int num_txbdfree; /* number of TxBDs free */
-
- /* RX Locked fields */
- spinlock_t rxlock;
struct device_node *node;
struct net_device *ndev;
struct of_device *ofdev;
- struct napi_struct napi;
-
- /* skb array and index */
- struct sk_buff ** rx_skbuff;
- u16 skb_currx;
-
- /* RX Coalescing values */
- unsigned char rxcoalescing;
- unsigned long rxic;
- struct rxbd8 *rx_bd_base; /* First Rx buffers */
- struct rxbd8 *cur_rx; /* Next free rx ring entry */
+ struct gfar_priv_tx_q *tx_queue;
+ struct gfar_priv_rx_q *rx_queue;
- /* RX parameters */
- unsigned int rx_ring_size;
+ /* RX per device parameters */
unsigned int rx_buffer_size;
unsigned int rx_stash_size;
unsigned int rx_stash_index;
diff --git a/drivers/net/gianfar_ethtool.c b/drivers/net/gianfar_ethtool.c
index 6c144b5..6d0d171 100644
--- a/drivers/net/gianfar_ethtool.c
+++ b/drivers/net/gianfar_ethtool.c
@@ -7,8 +7,9 @@
*
* Author: Andy Fleming
* Maintainer: Kumar Gala
+ * Modifier: Sandeep Gopalpet <sandeep.kumar@freescale.com>
*
- * Copyright (c) 2003,2004 Freescale Semiconductor, Inc.
+ * Copyright 2003-2006, 2008-2009 Freescale Semiconductor, Inc.
*
* This software may be used and distributed according to
* the terms of the GNU Public License, Version 2, incorporated herein
@@ -41,7 +42,7 @@
#include "gianfar.h"
extern void gfar_start(struct net_device *dev);
-extern int gfar_clean_rx_ring(struct net_device *dev, int rx_work_limit);
+extern int gfar_clean_rx_ring(struct gfar_priv_rx_q *rx_queue, int rx_work_limit);
#define GFAR_MAX_COAL_USECS 0xffff
#define GFAR_MAX_COAL_FRAMES 0xff
@@ -197,12 +198,16 @@ static int gfar_gsettings(struct net_device *dev, struct ethtool_cmd *cmd)
{
struct gfar_private *priv = netdev_priv(dev);
struct phy_device *phydev = priv->phydev;
+ struct gfar_priv_rx_q *rx_queue = NULL;
+ struct gfar_priv_tx_q *tx_queue = NULL;
if (NULL == phydev)
return -ENODEV;
+ tx_queue = priv->tx_queue;
+ rx_queue = priv->rx_queue;
- cmd->maxtxpkt = get_icft_value(priv->txic);
- cmd->maxrxpkt = get_icft_value(priv->rxic);
+ cmd->maxtxpkt = get_icft_value(tx_queue->txic);
+ cmd->maxrxpkt = get_icft_value(rx_queue->rxic);
return phy_ethtool_gset(phydev, cmd);
}
@@ -279,6 +284,8 @@ static unsigned int gfar_ticks2usecs(struct gfar_private *priv, unsigned int tic
static int gfar_gcoalesce(struct net_device *dev, struct ethtool_coalesce *cvals)
{
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_rx_q *rx_queue = NULL;
+ struct gfar_priv_tx_q *tx_queue = NULL;
unsigned long rxtime;
unsigned long rxcount;
unsigned long txtime;
@@ -290,10 +297,13 @@ static int gfar_gcoalesce(struct net_device *dev, struct ethtool_coalesce *cvals
if (NULL == priv->phydev)
return -ENODEV;
- rxtime = get_ictt_value(priv->rxic);
- rxcount = get_icft_value(priv->rxic);
- txtime = get_ictt_value(priv->txic);
- txcount = get_icft_value(priv->txic);
+ rx_queue = priv->rx_queue;
+ tx_queue = priv->tx_queue;
+
+ rxtime = get_ictt_value(rx_queue->rxic);
+ rxcount = get_icft_value(rx_queue->rxic);
+ txtime = get_ictt_value(tx_queue->txic);
+ txcount = get_icft_value(tx_queue->txic);
cvals->rx_coalesce_usecs = gfar_ticks2usecs(priv, rxtime);
cvals->rx_max_coalesced_frames = rxcount;
@@ -339,16 +349,21 @@ static int gfar_gcoalesce(struct net_device *dev, struct ethtool_coalesce *cvals
static int gfar_scoalesce(struct net_device *dev, struct ethtool_coalesce *cvals)
{
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_tx_q *tx_queue = NULL;
+ struct gfar_priv_rx_q *rx_queue = NULL;
if (!(priv->device_flags & FSL_GIANFAR_DEV_HAS_COALESCE))
return -EOPNOTSUPP;
+ tx_queue = priv->tx_queue;
+ rx_queue = priv->rx_queue;
+
/* Set up rx coalescing */
if ((cvals->rx_coalesce_usecs == 0) ||
(cvals->rx_max_coalesced_frames == 0))
- priv->rxcoalescing = 0;
+ rx_queue->rxcoalescing = 0;
else
- priv->rxcoalescing = 1;
+ rx_queue->rxcoalescing = 1;
if (NULL == priv->phydev)
return -ENODEV;
@@ -366,15 +381,15 @@ static int gfar_scoalesce(struct net_device *dev, struct ethtool_coalesce *cvals
return -EINVAL;
}
- priv->rxic = mk_ic_value(cvals->rx_max_coalesced_frames,
+ rx_queue->rxic = mk_ic_value(cvals->rx_max_coalesced_frames,
gfar_usecs2ticks(priv, cvals->rx_coalesce_usecs));
/* Set up tx coalescing */
if ((cvals->tx_coalesce_usecs == 0) ||
(cvals->tx_max_coalesced_frames == 0))
- priv->txcoalescing = 0;
+ tx_queue->txcoalescing = 0;
else
- priv->txcoalescing = 1;
+ tx_queue->txcoalescing = 1;
/* Check the bounds of the values */
if (cvals->tx_coalesce_usecs > GFAR_MAX_COAL_USECS) {
@@ -389,16 +404,16 @@ static int gfar_scoalesce(struct net_device *dev, struct ethtool_coalesce *cvals
return -EINVAL;
}
- priv->txic = mk_ic_value(cvals->tx_max_coalesced_frames,
+ tx_queue->txic = mk_ic_value(cvals->tx_max_coalesced_frames,
gfar_usecs2ticks(priv, cvals->tx_coalesce_usecs));
gfar_write(&priv->regs->rxic, 0);
- if (priv->rxcoalescing)
- gfar_write(&priv->regs->rxic, priv->rxic);
+ if (rx_queue->rxcoalescing)
+ gfar_write(&priv->regs->rxic, rx_queue->rxic);
gfar_write(&priv->regs->txic, 0);
- if (priv->txcoalescing)
- gfar_write(&priv->regs->txic, priv->txic);
+ if (tx_queue->txcoalescing)
+ gfar_write(&priv->regs->txic, tx_queue->txic);
return 0;
}
@@ -409,6 +424,11 @@ static int gfar_scoalesce(struct net_device *dev, struct ethtool_coalesce *cvals
static void gfar_gringparam(struct net_device *dev, struct ethtool_ringparam *rvals)
{
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_tx_q *tx_queue = NULL;
+ struct gfar_priv_rx_q *rx_queue = NULL;
+
+ tx_queue = priv->tx_queue;
+ rx_queue = priv->rx_queue;
rvals->rx_max_pending = GFAR_RX_MAX_RING_SIZE;
rvals->rx_mini_max_pending = GFAR_RX_MAX_RING_SIZE;
@@ -418,10 +438,10 @@ static void gfar_gringparam(struct net_device *dev, struct ethtool_ringparam *rv
/* Values changeable by the user. The valid values are
* in the range 1 to the "*_max_pending" counterpart above.
*/
- rvals->rx_pending = priv->rx_ring_size;
- rvals->rx_mini_pending = priv->rx_ring_size;
- rvals->rx_jumbo_pending = priv->rx_ring_size;
- rvals->tx_pending = priv->tx_ring_size;
+ rvals->rx_pending = rx_queue->rx_ring_size;
+ rvals->rx_mini_pending = rx_queue->rx_ring_size;
+ rvals->rx_jumbo_pending = rx_queue->rx_ring_size;
+ rvals->tx_pending = tx_queue->tx_ring_size;
}
/* Change the current ring parameters, stopping the controller if
@@ -431,6 +451,8 @@ static void gfar_gringparam(struct net_device *dev, struct ethtool_ringparam *rv
static int gfar_sringparam(struct net_device *dev, struct ethtool_ringparam *rvals)
{
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_tx_q *tx_queue = NULL;
+ struct gfar_priv_rx_q *rx_queue = NULL;
int err = 0;
if (rvals->rx_pending > GFAR_RX_MAX_RING_SIZE)
@@ -451,29 +473,32 @@ static int gfar_sringparam(struct net_device *dev, struct ethtool_ringparam *rva
return -EINVAL;
}
+ tx_queue = priv->tx_queue;
+ rx_queue = priv->rx_queue;
+
if (dev->flags & IFF_UP) {
unsigned long flags;
/* Halt TX and RX, and process the frames which
* have already been received */
- spin_lock_irqsave(&priv->txlock, flags);
- spin_lock(&priv->rxlock);
+ spin_lock_irqsave(&tx_queue->txlock, flags);
+ spin_lock(&rx_queue->rxlock);
gfar_halt(dev);
- spin_unlock(&priv->rxlock);
- spin_unlock_irqrestore(&priv->txlock, flags);
+ spin_unlock(&rx_queue->rxlock);
+ spin_unlock_irqrestore(&tx_queue->txlock, flags);
- gfar_clean_rx_ring(dev, priv->rx_ring_size);
+ gfar_clean_rx_ring(rx_queue, rx_queue->rx_ring_size);
/* Now we take down the rings to rebuild them */
stop_gfar(dev);
}
/* Change the size */
- priv->rx_ring_size = rvals->rx_pending;
- priv->tx_ring_size = rvals->tx_pending;
- priv->num_txbdfree = priv->tx_ring_size;
+ rx_queue->rx_ring_size = rvals->rx_pending;
+ tx_queue->tx_ring_size = rvals->tx_pending;
+ tx_queue->num_txbdfree = tx_queue->tx_ring_size;
/* Rebuild the rings with the new size */
if (dev->flags & IFF_UP) {
@@ -486,24 +511,29 @@ static int gfar_sringparam(struct net_device *dev, struct ethtool_ringparam *rva
static int gfar_set_rx_csum(struct net_device *dev, uint32_t data)
{
struct gfar_private *priv = netdev_priv(dev);
+ struct gfar_priv_rx_q *rx_queue = NULL;
+ struct gfar_priv_tx_q *tx_queue = NULL;
unsigned long flags;
int err = 0;
if (!(priv->device_flags & FSL_GIANFAR_DEV_HAS_CSUM))
return -EOPNOTSUPP;
+ tx_queue = priv->tx_queue;
+ rx_queue = priv->rx_queue;
+
if (dev->flags & IFF_UP) {
/* Halt TX and RX, and process the frames which
* have already been received */
- spin_lock_irqsave(&priv->txlock, flags);
- spin_lock(&priv->rxlock);
+ spin_lock_irqsave(&tx_queue->txlock, flags);
+ spin_lock(&rx_queue->rxlock);
gfar_halt(dev);
- spin_unlock(&priv->rxlock);
- spin_unlock_irqrestore(&priv->txlock, flags);
+ spin_unlock(&rx_queue->rxlock);
+ spin_unlock_irqrestore(&tx_queue->txlock, flags);
- gfar_clean_rx_ring(dev, priv->rx_ring_size);
+ gfar_clean_rx_ring(rx_queue, rx_queue->rx_ring_size);
/* Now we take down the rings to rebuild them */
stop_gfar(dev);
diff --git a/drivers/net/gianfar_sysfs.c b/drivers/net/gianfar_sysfs.c
index dd26da7..9c664f8 100644
--- a/drivers/net/gianfar_sysfs.c
+++ b/drivers/net/gianfar_sysfs.c
@@ -8,8 +8,9 @@
*
* Author: Andy Fleming
* Maintainer: Kumar Gala (galak@kernel.crashing.org)
+ * Modifier: Sandeep Gopalpet <sandeep.kumar@freescale.com>
*
- * Copyright (c) 2002-2005 Freescale Semiconductor, Inc.
+ * Copyright 2002-2009 Freescale Semiconductor, Inc.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
@@ -49,6 +50,7 @@ static ssize_t gfar_set_bd_stash(struct device *dev,
const char *buf, size_t count)
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
+ struct gfar_priv_rx_q *rx_queue = NULL;
int new_setting = 0;
u32 temp;
unsigned long flags;
@@ -56,6 +58,8 @@ static ssize_t gfar_set_bd_stash(struct device *dev,
if (!(priv->device_flags & FSL_GIANFAR_DEV_HAS_BD_STASHING))
return count;
+ rx_queue = priv->rx_queue;
+
/* Find out the new setting */
if (!strncmp("on", buf, count - 1) || !strncmp("1", buf, count - 1))
new_setting = 1;
@@ -65,7 +69,7 @@ static ssize_t gfar_set_bd_stash(struct device *dev,
else
return count;
- spin_lock_irqsave(&priv->rxlock, flags);
+ spin_lock_irqsave(&rx_queue->rxlock, flags);
/* Set the new stashing value */
priv->bd_stash_en = new_setting;
@@ -79,7 +83,7 @@ static ssize_t gfar_set_bd_stash(struct device *dev,
gfar_write(&priv->regs->attr, temp);
- spin_unlock_irqrestore(&priv->rxlock, flags);
+ spin_unlock_irqrestore(&rx_queue->rxlock, flags);
return count;
}
@@ -99,6 +103,7 @@ static ssize_t gfar_set_rx_stash_size(struct device *dev,
const char *buf, size_t count)
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
+ struct gfar_priv_rx_q *rx_queue = NULL;
unsigned int length = simple_strtoul(buf, NULL, 0);
u32 temp;
unsigned long flags;
@@ -106,7 +111,9 @@ static ssize_t gfar_set_rx_stash_size(struct device *dev,
if (!(priv->device_flags & FSL_GIANFAR_DEV_HAS_BUF_STASHING))
return count;
- spin_lock_irqsave(&priv->rxlock, flags);
+ rx_queue = priv->rx_queue;
+
+ spin_lock_irqsave(&rx_queue->rxlock, flags);
if (length > priv->rx_buffer_size)
goto out;
@@ -131,7 +138,7 @@ static ssize_t gfar_set_rx_stash_size(struct device *dev,
gfar_write(&priv->regs->attr, temp);
out:
- spin_unlock_irqrestore(&priv->rxlock, flags);
+ spin_unlock_irqrestore(&rx_queue->rxlock, flags);
return count;
}
@@ -154,6 +161,7 @@ static ssize_t gfar_set_rx_stash_index(struct device *dev,
const char *buf, size_t count)
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
+ struct gfar_priv_rx_q *rx_queue = NULL;
unsigned short index = simple_strtoul(buf, NULL, 0);
u32 temp;
unsigned long flags;
@@ -161,7 +169,9 @@ static ssize_t gfar_set_rx_stash_index(struct device *dev,
if (!(priv->device_flags & FSL_GIANFAR_DEV_HAS_BUF_STASHING))
return count;
- spin_lock_irqsave(&priv->rxlock, flags);
+ rx_queue = priv->rx_queue;
+
+ spin_lock_irqsave(&rx_queue->rxlock, flags);
if (index > priv->rx_stash_size)
goto out;
@@ -176,7 +186,7 @@ static ssize_t gfar_set_rx_stash_index(struct device *dev,
gfar_write(&priv->regs->attreli, flags);
out:
- spin_unlock_irqrestore(&priv->rxlock, flags);
+ spin_unlock_irqrestore(&rx_queue->rxlock, flags);
return count;
}
@@ -198,6 +208,7 @@ static ssize_t gfar_set_fifo_threshold(struct device *dev,
const char *buf, size_t count)
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
+ struct gfar_priv_tx_q *tx_queue = NULL;
unsigned int length = simple_strtoul(buf, NULL, 0);
u32 temp;
unsigned long flags;
@@ -205,7 +216,9 @@ static ssize_t gfar_set_fifo_threshold(struct device *dev,
if (length > GFAR_MAX_FIFO_THRESHOLD)
return count;
- spin_lock_irqsave(&priv->txlock, flags);
+ tx_queue = priv->tx_queue;
+
+ spin_lock_irqsave(&tx_queue->txlock, flags);
priv->fifo_threshold = length;
@@ -214,7 +227,7 @@ static ssize_t gfar_set_fifo_threshold(struct device *dev,
temp |= length;
gfar_write(&priv->regs->fifo_tx_thr, temp);
- spin_unlock_irqrestore(&priv->txlock, flags);
+ spin_unlock_irqrestore(&tx_queue->txlock, flags);
return count;
}
@@ -235,6 +248,7 @@ static ssize_t gfar_set_fifo_starve(struct device *dev,
const char *buf, size_t count)
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
+ struct gfar_priv_tx_q *tx_queue = NULL;
unsigned int num = simple_strtoul(buf, NULL, 0);
u32 temp;
unsigned long flags;
@@ -242,7 +256,8 @@ static ssize_t gfar_set_fifo_starve(struct device *dev,
if (num > GFAR_MAX_FIFO_STARVE)
return count;
- spin_lock_irqsave(&priv->txlock, flags);
+ tx_queue = priv->tx_queue;
+ spin_lock_irqsave(&tx_queue->txlock, flags);
priv->fifo_starve = num;
@@ -251,7 +266,7 @@ static ssize_t gfar_set_fifo_starve(struct device *dev,
temp |= num;
gfar_write(&priv->regs->fifo_tx_starve, temp);
- spin_unlock_irqrestore(&priv->txlock, flags);
+ spin_unlock_irqrestore(&tx_queue->txlock, flags);
return count;
}
@@ -273,6 +288,7 @@ static ssize_t gfar_set_fifo_starve_off(struct device *dev,
const char *buf, size_t count)
{
struct gfar_private *priv = netdev_priv(to_net_dev(dev));
+ struct gfar_priv_tx_q *tx_queue = NULL;
unsigned int num = simple_strtoul(buf, NULL, 0);
u32 temp;
unsigned long flags;
@@ -280,7 +296,8 @@ static ssize_t gfar_set_fifo_starve_off(struct device *dev,
if (num > GFAR_MAX_FIFO_STARVE_OFF)
return count;
- spin_lock_irqsave(&priv->txlock, flags);
+ tx_queue = priv->tx_queue;
+ spin_lock_irqsave(&tx_queue->txlock, flags);
priv->fifo_starve_off = num;
@@ -289,7 +306,7 @@ static ssize_t gfar_set_fifo_starve_off(struct device *dev,
temp |= num;
gfar_write(&priv->regs->fifo_tx_starve_shutoff, temp);
- spin_unlock_irqrestore(&priv->txlock, flags);
+ spin_unlock_irqrestore(&tx_queue->txlock, flags);
return count;
}
--
1.5.2.2
^ permalink raw reply related
* Re: [PATCH next-next-2.6] netdev: better dev_name_hash
From: Eric Dumazet @ 2009-10-26 14:55 UTC (permalink / raw)
To: Octavian Purdila; +Cc: Krishna Kumar2, Hagen Paul Pfeifer, netdev
In-Reply-To: <200910261631.44666.opurdila@ixiacom.com>
Octavian Purdila a écrit :
> On Monday 26 October 2009 15:07:31 you wrote:
>> Eric Dumazet wrote on 10/26/2009 10:58:34 AM:
>>> In fact, new10 should be the 'perfect' hash for the "eth%d"
>>> netdev use, not jhash (way more expensive in cpu cycles BTW)
>>>
>>> Most linux machines in the world have less than 10 interfaces, jhash
>>> would be really overkill.
>>>
>>>
>>> Thanks
>>>
>>>
>>> [PATCH net-next-2.6] netdev: better dev_name_hash
>> Changing Eric's test program to pass a multiplier to string_hash()
>> and calling string_hash with multipler=<2 - 63> confirms that 10
>> is almost always the best number for varying netdev names. I print
>> the number of times perfect 64 was scored, and for most passed
>> device names, the best is for n=10, followed by n=5 and others.
>> Almost the worst was n=31, slightly better was n=17.
>>
>> But other variables matter too, like fewer entries (4K or 1K) but
>> above values for are still better compared to n=31.
>>
>
> Hmm, I've found out that it very much depends on the name as well:
>
> 0 - orig, 1 - jhash, 2 - new10, 3 - new17, 4 - new31
>
> $ ./dev_name_hash ixint 16000 0 16
> score 24741
> $ ./dev_name_hash ixint 16000 1 16
> score 17949
> $ ./dev_name_hash ixint 16000 2 16
> score 16000
> $ ./dev_name_hash ixint 16000 3 16
> score 16715
> $ ./dev_name_hash ixint 16000 4 16
> score 18125
>
>
> $ ./dev_name_hash ixunc 16000 0 16
> score 24741
> $ ./dev_name_hash ixunc 16000 1 16
> score 17904
> $ ./dev_name_hash ixunc 16000 2 16
> score 22180
> $ ./dev_name_hash ixunc 16000 3 16
> score 17065
> $ ./dev_name_hash ixunc 16000 4 16
> score 18038
This is because you chose a 65536 slots hash table, to store 16000 elements
The ideal function should be :
$ ./dev_name_hash ixunc 16000 5 16
score 16000
unsigned int dev_name_hash_new10bis(const char *name)
{
unsigned hash = 0;
int len = strnlen(name, IFNAMSIZ);
int i;
for (i = 0; i < len; ++i)
hash = 10 * hash + (name[i] - '0');
return hash;
}
But should we really care ?
^ permalink raw reply
* RE: [PATCH net-next] bnx2x: Do Tx handling in a separate tasklet.
From: Vladislav Zolotarov @ 2009-10-26 14:42 UTC (permalink / raw)
To: David Miller; +Cc: Eilon Greenstein, netdev@vger.kernel.org
In-Reply-To: <20091025.142059.175260055.davem@davemloft.net>
The separation of Tx and Rx interrupt handling gives us the possibility to properly affinitize the Rx (heavy CPU consuming task) and Tx (low CPU consuming task) and to ensure that Tx work is done not long after the Tx interrupt without interference of Rx work thus letting the user benefit from Tx coalescing configuration in order to achieve the best performance in each specific scenario. This is most important in heavy load scenarios with mixed traffic (UDP + TCP for instance). If we didn't separate Tx and Rx interrupt handling Tx coalescing configuration was not worth much.
The mentioned above lead us to conclusion that it's more efficient not to use NAPI for freeing Tx buffers but to use a separate tasklet instead.
Best regards,
vlad
-----Original Message-----
From: netdev-owner@vger.kernel.org [mailto:netdev-owner@vger.kernel.org] On Behalf Of David Miller
Sent: Sunday, October 25, 2009 11:21 PM
To: Vladislav Zolotarov
Cc: Eilon Greenstein; netdev@vger.kernel.org
Subject: Re: [PATCH net-next] bnx2x: Do Tx handling in a separate tasklet.
From: "Vladislav Zolotarov" <vladz@broadcom.com>
Date: Sun, 25 Oct 2009 14:19:42 +0200
> This patch moves the 'Tx interrupt work' of each Tx queue from the hardIRQ
> context to the separate low-latency tasklet. Otherwise there is a possibility
> of a software lockup situation in a Tx softIRQ as it handles freeing all skb's
> 'freed' in (hard)IRQ context.
>
> Signed-off-by: Vladislav Zolotarov <vladz@broadcom.com>
> Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Use NAPI, please...
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply
* Re: [PATCH next-next-2.6] netdev: better dev_name_hash
From: Octavian Purdila @ 2009-10-26 14:31 UTC (permalink / raw)
To: Krishna Kumar2; +Cc: Eric Dumazet, Hagen Paul Pfeifer, netdev
In-Reply-To: <OF703F980D.0971407E-ON6525765B.00458CCA-6525765B.00481976@in.ibm.com>
On Monday 26 October 2009 15:07:31 you wrote:
> Eric Dumazet wrote on 10/26/2009 10:58:34 AM:
> > In fact, new10 should be the 'perfect' hash for the "eth%d"
> > netdev use, not jhash (way more expensive in cpu cycles BTW)
> >
> > Most linux machines in the world have less than 10 interfaces, jhash
> > would be really overkill.
> >
> >
> > Thanks
> >
> >
> > [PATCH net-next-2.6] netdev: better dev_name_hash
>
> Changing Eric's test program to pass a multiplier to string_hash()
> and calling string_hash with multipler=<2 - 63> confirms that 10
> is almost always the best number for varying netdev names. I print
> the number of times perfect 64 was scored, and for most passed
> device names, the best is for n=10, followed by n=5 and others.
> Almost the worst was n=31, slightly better was n=17.
>
> But other variables matter too, like fewer entries (4K or 1K) but
> above values for are still better compared to n=31.
>
Hmm, I've found out that it very much depends on the name as well:
0 - orig, 1 - jhash, 2 - new10, 3 - new17, 4 - new31
$ ./dev_name_hash ixint 16000 0 16
score 24741
$ ./dev_name_hash ixint 16000 1 16
score 17949
$ ./dev_name_hash ixint 16000 2 16
score 16000
$ ./dev_name_hash ixint 16000 3 16
score 16715
$ ./dev_name_hash ixint 16000 4 16
score 18125
$ ./dev_name_hash ixunc 16000 0 16
score 24741
$ ./dev_name_hash ixunc 16000 1 16
score 17904
$ ./dev_name_hash ixunc 16000 2 16
score 22180
$ ./dev_name_hash ixunc 16000 3 16
score 17065
$ ./dev_name_hash ixunc 16000 4 16
score 18038
^ permalink raw reply
* Re: Increasing TCP initial cwnd
From: Ilpo Järvinen @ 2009-10-26 13:57 UTC (permalink / raw)
To: Yair Gottdenker; +Cc: Netdev
In-Reply-To: <alpine.DEB.2.00.0910261553180.19761@wel-95.cs.helsinki.fi>
[-- Attachment #1: Type: TEXT/PLAIN, Size: 1127 bytes --]
On Mon, 26 Oct 2009, Ilpo Järvinen wrote:
> On Mon, 26 Oct 2009, Yair Gottdenker wrote:
>
> > I am working on research project to control the sender initial
> > congestion window size. I am trying to allow user space to set the
> > initial congestion window size but with no luck. The sender always
> > sends just 4 packets disregarding from the snd_cwnd.
> >
> > I am working on kernel version 2.6.31.3.
> >
> > I made the following changes:
> > 1. tcp_ipv4.c -> in function tcp_v4_init_sock, changed from
> > tp->snd_cwnd = 2 to tp->snd_cwnd = user_space_initial_cwnd.
> > 2. tcp_output.c -> tcp_init_cwnd - to always return user_space_initial_cwnd.
> > 3. tcp_cong.c -> tcp_slow_start- to always return user_space_initial_cwnd.
> >
> > Any help will be more than welcome.
>
> How about trying tcp_init_metrics in tcp_input.c?
Also, now that I looked. I don't understand what interesting you find in
tcp_slow_start to modify? That certainly has nothing to do with initial
window. ....Then you also might want to play with RW but that
is out-of-scope of your question, just mentioning it for completeleness.
--
i.
^ permalink raw reply
* Re: Increasing TCP initial cwnd
From: Ilpo Järvinen @ 2009-10-26 13:54 UTC (permalink / raw)
To: Yair Gottdenker; +Cc: Netdev
In-Reply-To: <fa3cddca0910260640s22d0396ey5fc51c51047bc2c@mail.gmail.com>
On Mon, 26 Oct 2009, Yair Gottdenker wrote:
> I am working on research project to control the sender initial
> congestion window size. I am trying to allow user space to set the
> initial congestion window size but with no luck. The sender always
> sends just 4 packets disregarding from the snd_cwnd.
>
> I am working on kernel version 2.6.31.3.
>
> I made the following changes:
> 1. tcp_ipv4.c -> in function tcp_v4_init_sock, changed from
> tp->snd_cwnd = 2 to tp->snd_cwnd = user_space_initial_cwnd.
> 2. tcp_output.c -> tcp_init_cwnd - to always return user_space_initial_cwnd.
> 3. tcp_cong.c -> tcp_slow_start- to always return user_space_initial_cwnd.
>
> Any help will be more than welcome.
How about trying tcp_init_metrics in tcp_input.c?
--
i.
^ permalink raw reply
* Re: linux-next: manual merge of the net tree with the i2c tree
From: Ben Hutchings @ 2009-10-26 13:46 UTC (permalink / raw)
To: Stephen Rothwell
Cc: David Miller, netdev, linux-next, linux-kernel, Mika Kuoppala,
Jean Delvare
In-Reply-To: <20091026133757.7cf87e49.sfr@canb.auug.org.au>
On Mon, 2009-10-26 at 13:37 +1100, Stephen Rothwell wrote:
> Hi all,
>
> Today's linux-next merge of the net tree got a conflict in
> drivers/net/sfc/sfe4001.c between commit
> 3f7c0648f727a6d5baf6117653e4001dc877b90b ("i2c: Prevent priority
> inversion on top of bus lock") from the i2c tree and commit
> c9597d4f89565b6562bd3026adbe6eac6c317f47 ("sfc: Merge sfe4001.c into
> falcon_boards.c") from the net tree.
>
> I have applied the following merge fixup patch (after removing
> drivers/net/sfc/sfe4001.c) and can carry it as necessary.
>
> --
> Cheers,
> Stephen Rothwell sfr@canb.auug.org.au
>
> From: Stephen Rothwell <sfr@canb.auug.org.au>
> Date: Mon, 26 Oct 2009 12:24:55 +1100
> Subject: [PATCH] net: merge fixup for drivers/net/sfc/falcon_boards.c
>
> Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Ben Hutchings <bhutchings@solarflare.com>
Ben.
--
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply
* [PATCH 2/2] tc35815: Enable NAPI
From: Atsushi Nemoto @ 2009-10-26 13:46 UTC (permalink / raw)
To: netdev; +Cc: David Miller, Ralf Roesch
This driver has NAPI code but it has been disabled. Enable it now.
The non-napi code will be removed lator.
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
---
drivers/net/tc35815.c | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/drivers/net/tc35815.c b/drivers/net/tc35815.c
index 3d38479..0d621ca 100644
--- a/drivers/net/tc35815.c
+++ b/drivers/net/tc35815.c
@@ -22,6 +22,7 @@
* All Rights Reserved.
*/
+#define TC35815_NAPI
#ifdef TC35815_NAPI
#define DRV_VERSION "1.38-NAPI"
#else
--
1.5.6.5
^ permalink raw reply related
* [PATCH 1/2] tc35815: Fix return value of tc35815_do_interrupt when NAPI enabled
From: Atsushi Nemoto @ 2009-10-26 13:46 UTC (permalink / raw)
To: netdev; +Cc: David Miller, Ralf Roesch
Return received count correctly even if tx completed at the same time.
Currently NAPI is disabled for this driver so this patch does not fix
any real problem.
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
---
drivers/net/tc35815.c | 5 +++++
1 files changed, 5 insertions(+), 0 deletions(-)
diff --git a/drivers/net/tc35815.c b/drivers/net/tc35815.c
index d1298e5..3d38479 100644
--- a/drivers/net/tc35815.c
+++ b/drivers/net/tc35815.c
@@ -1592,7 +1592,12 @@ static int tc35815_do_interrupt(struct net_device *dev, u32 status)
lp->lstats.tx_ints++;
tc35815_txdone(dev);
netif_wake_queue(dev);
+#ifdef TC35815_NAPI
+ if (ret < 0)
+ ret = 0;
+#else
ret = 0;
+#endif
}
return ret;
}
--
1.5.6.5
^ permalink raw reply related
* Re: [PATCH net-next-2.6] net: sysfs: ethtool_ops can be NULL
From: Andy Gospodarek @ 2009-10-26 13:40 UTC (permalink / raw)
To: Eric Dumazet; +Cc: David S. Miller, Linux Netdev List, Andy Gospodarek
In-Reply-To: <4AE586B5.4030308@gmail.com>
On Mon, Oct 26, 2009 at 12:23:33PM +0100, Eric Dumazet wrote:
> commit d519e17e2d01a0ee9abe083019532061b4438065
> (net: export device speed and duplex via sysfs)
> made the wrong assumption that netdev->ethtool_ops was always set.
>
> This makes possible to crash kernel and let rtnl in locked state.
>
> modprobe dummy
> ip link set dummy0 up
> (udev runs and crash)
>
> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
> ---
> net/core/net-sysfs.c | 8 ++++++--
> 1 files changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/net/core/net-sysfs.c b/net/core/net-sysfs.c
> index 753c420..89de182 100644
> --- a/net/core/net-sysfs.c
> +++ b/net/core/net-sysfs.c
> @@ -139,7 +139,9 @@ static ssize_t show_speed(struct device *dev,
> if (!rtnl_trylock())
> return restart_syscall();
>
> - if (netif_running(netdev) && netdev->ethtool_ops->get_settings) {
> + if (netif_running(netdev) &&
> + netdev->ethtool_ops &&
> + netdev->ethtool_ops->get_settings) {
> struct ethtool_cmd cmd = { ETHTOOL_GSET };
>
> if (!netdev->ethtool_ops->get_settings(netdev, &cmd))
> @@ -158,7 +160,9 @@ static ssize_t show_duplex(struct device *dev,
> if (!rtnl_trylock())
> return restart_syscall();
>
> - if (netif_running(netdev) && netdev->ethtool_ops->get_settings) {
> + if (netif_running(netdev) &&
> + netdev->ethtool_ops &&
> + netdev->ethtool_ops->get_settings) {
> struct ethtool_cmd cmd = { ETHTOOL_GSET };
>
> if (!netdev->ethtool_ops->get_settings(netdev, &cmd))
Nice catch, Eric.
Acked-by: Andy Gospodarek <andy@greyhouse.net>
^ permalink raw reply
* Increasing TCP initial cwnd
From: Yair Gottdenker @ 2009-10-26 13:40 UTC (permalink / raw)
To: netdev
Hi all,
I am working on research project to control the sender initial
congestion window size. I am trying to allow user space to set the
initial congestion window size but with no luck. The sender always
sends just 4 packets disregarding from the snd_cwnd.
I am working on kernel version 2.6.31.3.
I made the following changes:
1. tcp_ipv4.c -> in function tcp_v4_init_sock, changed from
tp->snd_cwnd = 2 to tp->snd_cwnd = user_space_initial_cwnd.
2. tcp_output.c -> tcp_init_cwnd - to always return user_space_initial_cwnd.
3. tcp_cong.c -> tcp_slow_start- to always return user_space_initial_cwnd.
Any help will be more than welcome.
Yair
^ permalink raw reply
* Re: [PATCH next-next-2.6] netdev: better dev_name_hash
From: Krishna Kumar2 @ 2009-10-26 13:07 UTC (permalink / raw)
To: Eric Dumazet; +Cc: Hagen Paul Pfeifer, netdev, Octavian Purdila
In-Reply-To: <4AE53382.8020308@gmail.com>
Eric Dumazet wrote on 10/26/2009 10:58:34 AM:
> In fact, new10 should be the 'perfect' hash for the "eth%d"
> netdev use, not jhash (way more expensive in cpu cycles BTW)
>
> Most linux machines in the world have less than 10 interfaces, jhash
> would be really overkill.
>
>
> Thanks
>
>
> [PATCH net-next-2.6] netdev: better dev_name_hash
Changing Eric's test program to pass a multiplier to string_hash()
and calling string_hash with multipler=<2 - 63> confirms that 10
is almost always the best number for varying netdev names. I print
the number of times perfect 64 was scored, and for most passed
device names, the best is for n=10, followed by n=5 and others.
Almost the worst was n=31, slightly better was n=17.
But other variables matter too, like fewer entries (4K or 1K) but
above values for are still better compared to n=31.
Thanks,
- KK
#include <stdio.h>
#include <string.h>
#define NUM_ENTRIES 1024
static inline unsigned int
string_hash_n(const unsigned char *name, unsigned int len, int n)
{
unsigned hash = 0;
int i;
for (i = 0; i < len; ++i)
hash = n * hash + name[i];
return hash;
}
int best_n = -1, best_n_val = -1;
void print_dist(unsigned int *dist, int n)
{
unsigned int i;
int perfect = 0;
printf("-----------------------------------------------------------\n");
printf("n=%d[] = {\n", n);
for (i = 0; i < 256; i++) {
if (dist[i] == NUM_ENTRIES/256)
perfect++;
printf("%d, ", dist[i]);
if ((i & 15) == 15) putchar('\n');
}
if (perfect > best_n_val) {
best_n_val = perfect;
best_n = n;
}
printf("}; (PERFECT = %d (n=%d))\n", perfect, n);
}
int main(int argc, char *argv[])
{
int n, i;
char *str, name[16];
unsigned int hash, hashdist[256];
if (argc == 1)
str="eth";
else
str=argv[1];
for (n = 2; n < 63; n++) {
memset(hashdist, 0, 256*sizeof(int));
for (i = 0; i < NUM_ENTRIES; i++) {
unsigned int len = sprintf(name, "%s%d", str, i);
hash = string_hash_n(name, len, n);
hashdist[hash & 255]++;
}
print_dist(hashdist, n);
}
printf("Best N was %d, and perfect entries: %d\n",
best_n, best_n_val);
return 0;
}
^ permalink raw reply
* [PATCH net-next-2.6] net: sysfs: ethtool_ops can be NULL
From: Eric Dumazet @ 2009-10-26 11:23 UTC (permalink / raw)
To: David S. Miller; +Cc: Linux Netdev List, Andy Gospodarek
commit d519e17e2d01a0ee9abe083019532061b4438065
(net: export device speed and duplex via sysfs)
made the wrong assumption that netdev->ethtool_ops was always set.
This makes possible to crash kernel and let rtnl in locked state.
modprobe dummy
ip link set dummy0 up
(udev runs and crash)
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
---
net/core/net-sysfs.c | 8 ++++++--
1 files changed, 6 insertions(+), 2 deletions(-)
diff --git a/net/core/net-sysfs.c b/net/core/net-sysfs.c
index 753c420..89de182 100644
--- a/net/core/net-sysfs.c
+++ b/net/core/net-sysfs.c
@@ -139,7 +139,9 @@ static ssize_t show_speed(struct device *dev,
if (!rtnl_trylock())
return restart_syscall();
- if (netif_running(netdev) && netdev->ethtool_ops->get_settings) {
+ if (netif_running(netdev) &&
+ netdev->ethtool_ops &&
+ netdev->ethtool_ops->get_settings) {
struct ethtool_cmd cmd = { ETHTOOL_GSET };
if (!netdev->ethtool_ops->get_settings(netdev, &cmd))
@@ -158,7 +160,9 @@ static ssize_t show_duplex(struct device *dev,
if (!rtnl_trylock())
return restart_syscall();
- if (netif_running(netdev) && netdev->ethtool_ops->get_settings) {
+ if (netif_running(netdev) &&
+ netdev->ethtool_ops &&
+ netdev->ethtool_ops->get_settings) {
struct ethtool_cmd cmd = { ETHTOOL_GSET };
if (!netdev->ethtool_ops->get_settings(netdev, &cmd))
^ permalink raw reply related
* Re: [PATCH] virtio-net: fix data corruption with OOM
From: Michael S. Tsirkin @ 2009-10-26 9:07 UTC (permalink / raw)
To: Rusty Russell; +Cc: virtualization, kvm, netdev
In-Reply-To: <200910261211.52148.rusty@rustcorp.com.au>
On Mon, Oct 26, 2009 at 12:11:51PM +1030, Rusty Russell wrote:
> On Mon, 26 Oct 2009 03:33:40 am Michael S. Tsirkin wrote:
> > virtio net used to unlink skbs from send queues on error,
> > but ever since 48925e372f04f5e35fec6269127c62b2c71ab794
> > we do not do this. This causes guest data corruption and crashes
> > with vhost since net core can requeue the skb or free it without
> > it being taken off the list.
> >
> > This patch fixes this by queueing the skb after successfull
> > transmit.
>
> I originally thought that this was racy: as soon as we do add_buf, we need to
> make sure we're ready for the callback (for virtio_pci, it's ->kick, but we
> shouldn't rely on that).
>
> So a comment would be nice. How's this?
>
> Subject: virtio-net: fix data corruption with OOM
> Date: Sun, 25 Oct 2009 19:03:40 +0200
> From: "Michael S. Tsirkin" <mst@redhat.com>
Another, and hopefully the last, note, is that
git-am can only handle Subject/From lines
at the beginning of the message.
So git style of the mail would be
Subject: XXX
From: YYY
Description
---
Discussion and notes.
I think it's weird. We could invent some kind of separator
that would make git-am accept Subject/From/Date lines in
the middle of the message, so that discussion can come before
the description. Worth it?
>
> virtio net used to unlink skbs from send queues on error,
> but ever since 48925e372f04f5e35fec6269127c62b2c71ab794
> we do not do this. This causes guest data corruption and crashes
> with vhost since net core can requeue the skb or free it without
> it being taken off the list.
>
> This patch fixes this by queueing the skb after successful
> transmit.
>
> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (+ comment)
> ---
>
> Rusty, here's a fix for another data corrupter I saw.
> This fixes a regression from 2.6.31, so definitely
> 2.6.32 I think. Comments?
>
> drivers/net/virtio_net.c | 8 +++++---
> 1 files changed, 5 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -516,8 +516,7 @@ again:
> /* Free up any pending old buffers before queueing new ones. */
> free_old_xmit_skbs(vi);
>
> - /* Put new one in send queue and do transmit */
> - __skb_queue_head(&vi->send, skb);
> + /* Try to transmit */
> capacity = xmit_skb(vi, skb);
>
> /* This can happen with OOM and indirect buffers. */
> @@ -531,8 +530,17 @@ again:
> }
> return NETDEV_TX_BUSY;
> }
> + vi->svq->vq_ops->kick(vi->svq);
>
> - vi->svq->vq_ops->kick(vi->svq);
> + /*
> + * Put new one in send queue. You'd expect we'd need this before
> + * xmit_skb calls add_buf(), since the callback can be triggered
> + * immediately after that. But since the callback just triggers
> + * another call back here, normal network xmit locking prevents the
> + * race.
> + */
> + __skb_queue_head(&vi->send, skb);
> +
> /* Don't wait up for transmitted skbs to be freed. */
> skb_orphan(skb);
> nf_reset(skb);
^ permalink raw reply
* Re: [RFC] make per interface sysctl entries configurable
From: Cosmin Ratiu @ 2009-10-26 9:01 UTC (permalink / raw)
To: Octavian Purdila
Cc: Eric Dumazet, Benjamin LaHaise, Stephen Hemminger, netdev
In-Reply-To: <200910260021.48785.opurdila@ixiacom.com>
On Monday 26 October 2009 00:21:48 Octavian Purdila wrote:
> On Sunday 25 October 2009 23:37:19 you wrote:
> > Octavian Purdila a écrit :
> > > RFC patches are attached.
> > >
> > > Another possible approach: add an interface flag and use it to decide
> > > whether we want per interface sysctl entries or not.
> >
> > Hmm, could we speedup sysctl instead, adding rbtree or something ?
>
> Very good point, I think this is the best solution for people using a
> moderately high number of interfaces (a few thousand).
>
> But for really large setups there is another issue: memory consumption. In
> fact, in order to be able to scale to 128K interfaces and still have a
> significant amount of memory available to applications we also had to
> disable sysfs and #ifdef CONFIG_SYSFS struct device from net_device.
>
> I would also argue that when you have such a large number of interfaces you
> don't need to change setting on a per interface basis. Or at least this is
> our case :) and I suspect that the case with a large number of PPP
> interfaces is similar.
>
Another possible approach: shared settings for an interface group. If you have
a large number of interfaces of the same type it would be nice if you could
change some setting for the whole group instead of globally or individually.
Is this approach feasible anyway? Or I'm talking rubbish.
Cosmin.
^ permalink raw reply
* Re: [PATCH] virtio-net: fix data corruption with OOM
From: Michael S. Tsirkin @ 2009-10-26 9:00 UTC (permalink / raw)
To: Rusty Russell; +Cc: virtualization, kvm, netdev
In-Reply-To: <200910261211.52148.rusty@rustcorp.com.au>
On Mon, Oct 26, 2009 at 12:11:51PM +1030, Rusty Russell wrote:
> On Mon, 26 Oct 2009 03:33:40 am Michael S. Tsirkin wrote:
> > virtio net used to unlink skbs from send queues on error,
> > but ever since 48925e372f04f5e35fec6269127c62b2c71ab794
> > we do not do this. This causes guest data corruption and crashes
> > with vhost since net core can requeue the skb or free it without
> > it being taken off the list.
> >
> > This patch fixes this by queueing the skb after successfull
> > transmit.
>
> I originally thought that this was racy: as soon as we do add_buf, we need to
> make sure we're ready for the callback (for virtio_pci, it's ->kick, but we
> shouldn't rely on that).
BTW, wanted to note that unlink on error would *also* be racy if we did any
processing in the callback.
> So a comment would be nice. How's this?
>
> Subject: virtio-net: fix data corruption with OOM
> Date: Sun, 25 Oct 2009 19:03:40 +0200
> From: "Michael S. Tsirkin" <mst@redhat.com>
>
> virtio net used to unlink skbs from send queues on error,
> but ever since 48925e372f04f5e35fec6269127c62b2c71ab794
> we do not do this. This causes guest data corruption and crashes
> with vhost since net core can requeue the skb or free it without
> it being taken off the list.
>
> This patch fixes this by queueing the skb after successful
> transmit.
>
> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (+ comment)
> ---
>
> Rusty, here's a fix for another data corrupter I saw.
> This fixes a regression from 2.6.31, so definitely
> 2.6.32 I think. Comments?
>
> drivers/net/virtio_net.c | 8 +++++---
> 1 files changed, 5 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -516,8 +516,7 @@ again:
> /* Free up any pending old buffers before queueing new ones. */
> free_old_xmit_skbs(vi);
>
> - /* Put new one in send queue and do transmit */
> - __skb_queue_head(&vi->send, skb);
> + /* Try to transmit */
> capacity = xmit_skb(vi, skb);
>
> /* This can happen with OOM and indirect buffers. */
> @@ -531,8 +530,17 @@ again:
> }
> return NETDEV_TX_BUSY;
> }
> + vi->svq->vq_ops->kick(vi->svq);
>
> - vi->svq->vq_ops->kick(vi->svq);
> + /*
> + * Put new one in send queue. You'd expect we'd need this before
> + * xmit_skb calls add_buf(), since the callback can be triggered
> + * immediately after that. But since the callback just triggers
> + * another call back here, normal network xmit locking prevents the
> + * race.
> + */
> + __skb_queue_head(&vi->send, skb);
> +
> /* Don't wait up for transmitted skbs to be freed. */
> skb_orphan(skb);
> nf_reset(skb);
^ permalink raw reply
* Re: [PATCH 00/13] TProxy IPv6 support 2nd round
From: Balazs Scheidler @ 2009-10-26 9:00 UTC (permalink / raw)
To: Harald Welte; +Cc: netfilter-devel, netdev
In-Reply-To: <20091025101612.GS29852@prithivi.gnumonks.org>
On Sun, 2009-10-25 at 11:16 +0100, Harald Welte wrote:
> Dear Balazs,
>
> as you might have read from other mails (and by the long period of silence),
> Patrick McHardy is currently unavailable to perform his usual maintainer
> role.
>
> I personally am too much out of touch with recent developments in
> netfitler-land to be able to confidently review your patches...
>
> So unless somebody else from the team (Jozsef?, Pablo?) feels confident in
> ACKing your patchset, I will have to ask for your patience until Patrick is
> back and can do it by himself.
Thanks for letting me know, hopefully Patrick gets better soon. I've
planned another round of the TProxy patches as I got some comments at
the previous round I'm yet to address.
So no need to hurry.
--
Bazsi
^ permalink raw reply
* Re: [PATCH] net: Adjust softirq raising in __napi_schedule
From: Johannes Berg @ 2009-10-26 8:56 UTC (permalink / raw)
To: Tilman Schmidt
Cc: Jarek Poplawski, David Miller, hidave.darkstar, linux-kernel,
tglx, linux-wireless, linux-ppp, netdev, paulus, Michael Buesch,
Oliver Hartkopp
In-Reply-To: <4AE56217.3040708@imap.cc>
[-- Attachment #1: Type: text/plain, Size: 1582 bytes --]
On Mon, 2009-10-26 at 10:47 +0200, Tilman Schmidt wrote:
> > However, I lost track now of why we're discussing this.
>
> The starting point were several reports of the kernel message:
>
> NOHZ: local_softirq_pending 08
>
> Originally most if not all of them came from wireless networking,
> but I muddied the waters by adding to the mix a case involving ISDN.
> You stated that all the solutions proposed so far were wrong, so
> we're naturally turning to you for guidance on what the right
> solution might be.
Right. Sorry about getting lost here.
> > Basically it boils down to using netif_rx() when in (soft)irq, and
> > netif_rx_ni() when in process context. That could just be an
> > optimisation, but it's a very valid one.
>
> Hmmm. That seems to contradict your earlier statement to me that
> simply replacing a call to netif_rx() by one to netif_rx_ni()
> when not in interrupt context isn't a valid solution either.
> What am I missing?
Well, I think you misunderstood me. It would be correct to do this, if
and only if the code that calls it doesn't need the extra guarantee.
Any code (say ISDN code) that calls netif_rx() is clearly assuming to
always be running in (soft)irq context, otherwise it couldn't call
netif_rx() unconditionally. Agree so far?
So now if you change the ISDN code to call netif_rx_ni(), you've changed
the assumption that the ISDN code makes -- that it is running in
(soft)irq context. Therefore, you need to verify that this is actually a
correct change, which is what I tried to say.
johannes
[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 801 bytes --]
^ permalink raw reply
* Re: [PATCH] virtio-net: fix data corruption with OOM
From: Michael S. Tsirkin @ 2009-10-26 8:54 UTC (permalink / raw)
To: Rusty Russell; +Cc: virtualization, kvm, netdev
In-Reply-To: <200910261211.52148.rusty@rustcorp.com.au>
On Mon, Oct 26, 2009 at 12:11:51PM +1030, Rusty Russell wrote:
> On Mon, 26 Oct 2009 03:33:40 am Michael S. Tsirkin wrote:
> > virtio net used to unlink skbs from send queues on error,
> > but ever since 48925e372f04f5e35fec6269127c62b2c71ab794
> > we do not do this. This causes guest data corruption and crashes
> > with vhost since net core can requeue the skb or free it without
> > it being taken off the list.
> >
> > This patch fixes this by queueing the skb after successfull
> > transmit.
>
> I originally thought that this was racy: as soon as we do add_buf, we need to
> make sure we're ready for the callback (for virtio_pci, it's ->kick, but we
> shouldn't rely on that).
>
> So a comment would be nice. How's this?
Acked-by: Michael S. Tsirkin <mst@redhat.com>
> Subject: virtio-net: fix data corruption with OOM
> Date: Sun, 25 Oct 2009 19:03:40 +0200
> From: "Michael S. Tsirkin" <mst@redhat.com>
>
> virtio net used to unlink skbs from send queues on error,
> but ever since 48925e372f04f5e35fec6269127c62b2c71ab794
> we do not do this. This causes guest data corruption and crashes
> with vhost since net core can requeue the skb or free it without
> it being taken off the list.
>
> This patch fixes this by queueing the skb after successful
> transmit.
>
> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (+ comment)
> ---
>
> Rusty, here's a fix for another data corrupter I saw.
> This fixes a regression from 2.6.31, so definitely
> 2.6.32 I think. Comments?
>
> drivers/net/virtio_net.c | 8 +++++---
> 1 files changed, 5 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -516,8 +516,7 @@ again:
> /* Free up any pending old buffers before queueing new ones. */
> free_old_xmit_skbs(vi);
>
> - /* Put new one in send queue and do transmit */
> - __skb_queue_head(&vi->send, skb);
> + /* Try to transmit */
> capacity = xmit_skb(vi, skb);
>
> /* This can happen with OOM and indirect buffers. */
> @@ -531,8 +530,17 @@ again:
> }
> return NETDEV_TX_BUSY;
> }
> + vi->svq->vq_ops->kick(vi->svq);
>
> - vi->svq->vq_ops->kick(vi->svq);
> + /*
> + * Put new one in send queue. You'd expect we'd need this before
> + * xmit_skb calls add_buf(), since the callback can be triggered
> + * immediately after that. But since the callback just triggers
> + * another call back here, normal network xmit locking prevents the
> + * race.
> + */
> + __skb_queue_head(&vi->send, skb);
> +
> /* Don't wait up for transmitted skbs to be freed. */
> skb_orphan(skb);
> nf_reset(skb);
^ permalink raw reply
* Re: VLAN and ARP failure on tg3 drivers
From: Eric Dumazet @ 2009-10-26 8:54 UTC (permalink / raw)
To: Benny Amorsen; +Cc: Gertjan Hofman, Matt Carlson, netdev@vger.kernel.org
In-Reply-To: <m34opmr2fz.fsf@ursa.amorsen.dk>
Benny Amorsen a écrit :
> Gertjan Hofman <gertjan_hofman@yahoo.com> writes:
>
>> Dear Matt, Eric, Benny,
>>
>> Sorry about the slow response to your fast replies. I think Benny is
>> correct, the 'problem' lies in the fact that we were using a VLAN ID
>> of 0, without knowing its special significance. User error.
>>
>> I tested it with other VLAN id's (>0) and it appears to work fine. We
>> are not entirely sure we understand why it used to work with VLAN ID
>> 0 on the Broadcom chips and still does with a number of different
>> cards (with >2.6.27 kernels). What is the 'correct' behaviour for
>> this incorrect usage ?
>
> VLAN 0 isn't incorrect, it's just surprising. When you send a packet
> tagged with VLAN 0, it means that the packet should be interpreted as
> being the same VLAN as a completely untagged packet.
>
> So in theory, if both ends are using VLAN 0 and you aren't using eth0
> for anything, traffic should flow, at least if both ends are on the same
> kernel version. Feel free to debug why that isn't the case for you, of
> course...
>
VLAN id 0 is not usable on current kernel because we use 16 bits in skb to
store vlan_tci, and vlan_tci = 0 means there is no VLAN tagging.
We could use high order bit (0x8000) to tell if vlan tagging is set or not.
^ permalink raw reply
* Re: [PATCH] net: Adjust softirq raising in __napi_schedule
From: Tilman Schmidt @ 2009-10-26 8:47 UTC (permalink / raw)
To: Johannes Berg
Cc: Jarek Poplawski, David Miller, hidave.darkstar, linux-kernel,
tglx, linux-wireless, linux-ppp, netdev, paulus, Michael Buesch,
Oliver Hartkopp
In-Reply-To: <1256543932.28230.9.camel@johannes.local>
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Am 26.10.2009 09:58 schrieb Johannes Berg:
> On Mon, 2009-10-26 at 07:54 +0000, Jarek Poplawski wrote:
>
>>> No, I wrote that I didn't know. I suppose now that I looked at it I do
>>> know, and only disabling preemption is required.
>> But netif_rx has preemption disabled most of the time (by hardirqs
>> disabling). So maybe disabling preemption isn't the main reason here
>> either?
>
> Not for netpoll though, which may or may not be relevant (if I were to
> venture a guess I'd say it isn't and it disables preemption to be able
> to do the softirq thing)
>
> However, I lost track now of why we're discussing this.
The starting point were several reports of the kernel message:
NOHZ: local_softirq_pending 08
Originally most if not all of them came from wireless networking,
but I muddied the waters by adding to the mix a case involving ISDN.
You stated that all the solutions proposed so far were wrong, so
we're naturally turning to you for guidance on what the right
solution might be.
> Basically it boils down to using netif_rx() when in (soft)irq, and
> netif_rx_ni() when in process context. That could just be an
> optimisation, but it's a very valid one.
Hmmm. That seems to contradict your earlier statement to me that
simply replacing a call to netif_rx() by one to netif_rx_ni()
when not in interrupt context isn't a valid solution either.
What am I missing?
Thanks,
Tilman
- --
Tilman Schmidt E-Mail: tilman@imap.cc
Bonn, Germany
Diese Nachricht besteht zu 100% aus wiederverwerteten Bits.
Ungeöffnet mindestens haltbar bis: (siehe Rückseite)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.4 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
iD8DBQFK5WIXQ3+did9BuFsRAsj1AJ0e4VJ7Nsp69ROXCiT4oM/Q6lIpSwCfZvXS
4nV4tWTIzgk4IRlCt0CCF3Y=
=r15I
-----END PGP SIGNATURE-----
^ permalink raw reply
* Re: VLAN and ARP failure on tg3 drivers
From: Benny Amorsen @ 2009-10-26 8:20 UTC (permalink / raw)
To: Gertjan Hofman; +Cc: Matt Carlson, netdev@vger.kernel.org, Eric Dumazet
In-Reply-To: <477963.52849.qm@web32605.mail.mud.yahoo.com>
Gertjan Hofman <gertjan_hofman@yahoo.com> writes:
> Dear Matt, Eric, Benny,
>
> Sorry about the slow response to your fast replies. I think Benny is
> correct, the 'problem' lies in the fact that we were using a VLAN ID
> of 0, without knowing its special significance. User error.
>
> I tested it with other VLAN id's (>0) and it appears to work fine. We
> are not entirely sure we understand why it used to work with VLAN ID
> 0 on the Broadcom chips and still does with a number of different
> cards (with >2.6.27 kernels). What is the 'correct' behaviour for
> this incorrect usage ?
VLAN 0 isn't incorrect, it's just surprising. When you send a packet
tagged with VLAN 0, it means that the packet should be interpreted as
being the same VLAN as a completely untagged packet.
So in theory, if both ends are using VLAN 0 and you aren't using eth0
for anything, traffic should flow, at least if both ends are on the same
kernel version. Feel free to debug why that isn't the case for you, of
course...
/Benny
^ permalink raw reply
* [PATCH v3 2/7] Allow tcp_parse_options to consult dst entry
From: Gilad Ben-Yossef @ 2009-10-26 8:06 UTC (permalink / raw)
To: netdev; +Cc: ori, Gilad Ben-Yossef, Gilad Ben-Yossef
In-Reply-To: <1256544393-12450-1-git-send-email-gilad@codefidence.com>
From: Gilad Ben-Yossef <gby@watson.codefidence.com>
We need tcp_parse_options to be aware of dst_entry to
take into account per dst_entry TCP options settings
Signed-off-by: Gilad Ben-Yossef <gilad@codefidence.com>
Sigend-off-by: Ori Finkelman <ori@comsleep.com>
Sigend-off-by: Yony Amit <yony@comsleep.com>
---
include/net/tcp.h | 3 ++-
net/ipv4/syncookies.c | 27 ++++++++++++++-------------
net/ipv4/tcp_input.c | 9 ++++++---
net/ipv4/tcp_ipv4.c | 21 ++++++++++++---------
net/ipv4/tcp_minisocks.c | 7 +++++--
net/ipv6/syncookies.c | 28 +++++++++++++++-------------
net/ipv6/tcp_ipv6.c | 3 ++-
7 files changed, 56 insertions(+), 42 deletions(-)
diff --git a/include/net/tcp.h b/include/net/tcp.h
index 03a49c7..740d09b 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -409,7 +409,8 @@ extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
extern void tcp_parse_options(struct sk_buff *skb,
struct tcp_options_received *opt_rx,
- int estab);
+ int estab,
+ struct dst_entry *dst);
extern u8 *tcp_parse_md5sig_option(struct tcphdr *th);
diff --git a/net/ipv4/syncookies.c b/net/ipv4/syncookies.c
index a6e0e07..4990dd4 100644
--- a/net/ipv4/syncookies.c
+++ b/net/ipv4/syncookies.c
@@ -276,13 +276,6 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_SYNCOOKIESRECV);
- /* check for timestamp cookie support */
- memset(&tcp_opt, 0, sizeof(tcp_opt));
- tcp_parse_options(skb, &tcp_opt, 0);
-
- if (tcp_opt.saw_tstamp)
- cookie_check_timestamp(&tcp_opt);
-
ret = NULL;
req = inet_reqsk_alloc(&tcp_request_sock_ops); /* for safety */
if (!req)
@@ -298,12 +291,6 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
ireq->loc_addr = ip_hdr(skb)->daddr;
ireq->rmt_addr = ip_hdr(skb)->saddr;
ireq->ecn_ok = 0;
- ireq->snd_wscale = tcp_opt.snd_wscale;
- ireq->rcv_wscale = tcp_opt.rcv_wscale;
- ireq->sack_ok = tcp_opt.sack_ok;
- ireq->wscale_ok = tcp_opt.wscale_ok;
- ireq->tstamp_ok = tcp_opt.saw_tstamp;
- req->ts_recent = tcp_opt.saw_tstamp ? tcp_opt.rcv_tsval : 0;
/* We throwed the options of the initial SYN away, so we hope
* the ACK carries the same options again (see RFC1122 4.2.3.8)
@@ -351,6 +338,20 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
}
}
+ /* check for timestamp cookie support */
+ memset(&tcp_opt, 0, sizeof(tcp_opt));
+ tcp_parse_options(skb, &tcp_opt, 0, &rt->u.dst);
+
+ if (tcp_opt.saw_tstamp)
+ cookie_check_timestamp(&tcp_opt);
+
+ ireq->snd_wscale = tcp_opt.snd_wscale;
+ ireq->rcv_wscale = tcp_opt.rcv_wscale;
+ ireq->sack_ok = tcp_opt.sack_ok;
+ ireq->wscale_ok = tcp_opt.wscale_ok;
+ ireq->tstamp_ok = tcp_opt.saw_tstamp;
+ req->ts_recent = tcp_opt.saw_tstamp ? tcp_opt.rcv_tsval : 0;
+
/* Try to redo what tcp_v4_send_synack did. */
req->window_clamp = tp->window_clamp ? :dst_metric(&rt->u.dst, RTAX_WINDOW);
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index d86784b..d502f49 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -3698,12 +3698,14 @@ old_ack:
* the fast version below fails.
*/
void tcp_parse_options(struct sk_buff *skb, struct tcp_options_received *opt_rx,
- int estab)
+ int estab, struct dst_entry *dst)
{
unsigned char *ptr;
struct tcphdr *th = tcp_hdr(skb);
int length = (th->doff * 4) - sizeof(struct tcphdr);
+ BUG_ON(!estab && !dst);
+
ptr = (unsigned char *)(th + 1);
opt_rx->saw_tstamp = 0;
@@ -3820,7 +3822,7 @@ static int tcp_fast_parse_options(struct sk_buff *skb, struct tcphdr *th,
if (tcp_parse_aligned_timestamp(tp, th))
return 1;
}
- tcp_parse_options(skb, &tp->rx_opt, 1);
+ tcp_parse_options(skb, &tp->rx_opt, 1, NULL);
return 1;
}
@@ -5364,8 +5366,9 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
struct tcp_sock *tp = tcp_sk(sk);
struct inet_connection_sock *icsk = inet_csk(sk);
int saved_clamp = tp->rx_opt.mss_clamp;
+ struct dst_entry *dst = __sk_dst_get(sk);
- tcp_parse_options(skb, &tp->rx_opt, 0);
+ tcp_parse_options(skb, &tp->rx_opt, 0, dst);
if (th->ack) {
/* rfc793:
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 7cda24b..1d611e3 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -1256,11 +1256,21 @@ int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
tcp_rsk(req)->af_specific = &tcp_request_sock_ipv4_ops;
#endif
+ ireq = inet_rsk(req);
+ ireq->loc_addr = daddr;
+ ireq->rmt_addr = saddr;
+ ireq->no_srccheck = inet_sk(sk)->transparent;
+ ireq->opt = tcp_v4_save_options(sk, skb);
+
+ dst = inet_csk_route_req(sk, req);
+ if(!dst)
+ goto drop_and_free;
+
tcp_clear_options(&tmp_opt);
tmp_opt.mss_clamp = 536;
tmp_opt.user_mss = tcp_sk(sk)->rx_opt.user_mss;
- tcp_parse_options(skb, &tmp_opt, 0);
+ tcp_parse_options(skb, &tmp_opt, 0, dst);
if (want_cookie && !tmp_opt.saw_tstamp)
tcp_clear_options(&tmp_opt);
@@ -1269,14 +1279,8 @@ int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
tcp_openreq_init(req, &tmp_opt, skb);
- ireq = inet_rsk(req);
- ireq->loc_addr = daddr;
- ireq->rmt_addr = saddr;
- ireq->no_srccheck = inet_sk(sk)->transparent;
- ireq->opt = tcp_v4_save_options(sk, skb);
-
if (security_inet_conn_request(sk, skb, req))
- goto drop_and_free;
+ goto drop_and_release;
if (!want_cookie)
TCP_ECN_create_request(req, tcp_hdr(skb));
@@ -1301,7 +1305,6 @@ int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
*/
if (tmp_opt.saw_tstamp &&
tcp_death_row.sysctl_tw_recycle &&
- (dst = inet_csk_route_req(sk, req)) != NULL &&
(peer = rt_get_peer((struct rtable *)dst)) != NULL &&
peer->v4daddr == saddr) {
if (get_seconds() < peer->tcp_ts_stamp + TCP_PAWS_MSL &&
diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c
index c49a550..70ff955 100644
--- a/net/ipv4/tcp_minisocks.c
+++ b/net/ipv4/tcp_minisocks.c
@@ -101,7 +101,7 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb,
int paws_reject = 0;
if (th->doff > (sizeof(*th) >> 2) && tcptw->tw_ts_recent_stamp) {
- tcp_parse_options(skb, &tmp_opt, 1);
+ tcp_parse_options(skb, &tmp_opt, 1, NULL);
if (tmp_opt.saw_tstamp) {
tmp_opt.ts_recent = tcptw->tw_ts_recent;
@@ -499,10 +499,11 @@ struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb,
int paws_reject = 0;
struct tcp_options_received tmp_opt;
struct sock *child;
+ struct dst_entry *dst = inet_csk_route_req(sk, req);
tmp_opt.saw_tstamp = 0;
if (th->doff > (sizeof(struct tcphdr)>>2)) {
- tcp_parse_options(skb, &tmp_opt, 0);
+ tcp_parse_options(skb, &tmp_opt, 0, dst);
if (tmp_opt.saw_tstamp) {
tmp_opt.ts_recent = req->ts_recent;
@@ -515,6 +516,8 @@ struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb,
}
}
+ dst_release(dst);
+
/* Check for pure retransmitted SYN. */
if (TCP_SKB_CB(skb)->seq == tcp_rsk(req)->rcv_isn &&
flg == TCP_FLAG_SYN &&
diff --git a/net/ipv6/syncookies.c b/net/ipv6/syncookies.c
index 6b6ae91..6ece408 100644
--- a/net/ipv6/syncookies.c
+++ b/net/ipv6/syncookies.c
@@ -184,13 +184,6 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)
NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_SYNCOOKIESRECV);
- /* check for timestamp cookie support */
- memset(&tcp_opt, 0, sizeof(tcp_opt));
- tcp_parse_options(skb, &tcp_opt, 0);
-
- if (tcp_opt.saw_tstamp)
- cookie_check_timestamp(&tcp_opt);
-
ret = NULL;
req = inet6_reqsk_alloc(&tcp6_request_sock_ops);
if (!req)
@@ -224,12 +217,6 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)
req->expires = 0UL;
req->retrans = 0;
ireq->ecn_ok = 0;
- ireq->snd_wscale = tcp_opt.snd_wscale;
- ireq->rcv_wscale = tcp_opt.rcv_wscale;
- ireq->sack_ok = tcp_opt.sack_ok;
- ireq->wscale_ok = tcp_opt.wscale_ok;
- ireq->tstamp_ok = tcp_opt.saw_tstamp;
- req->ts_recent = tcp_opt.saw_tstamp ? tcp_opt.rcv_tsval : 0;
treq->rcv_isn = ntohl(th->seq) - 1;
treq->snt_isn = cookie;
@@ -264,6 +251,21 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)
goto out_free;
}
+ /* check for timestamp cookie support */
+ memset(&tcp_opt, 0, sizeof(tcp_opt));
+ tcp_parse_options(skb, &tcp_opt, 0, dst);
+
+ if (tcp_opt.saw_tstamp)
+ cookie_check_timestamp(&tcp_opt);
+
+ req->ts_recent = tcp_opt.saw_tstamp ? tcp_opt.rcv_tsval : 0;
+
+ ireq->snd_wscale = tcp_opt.snd_wscale;
+ ireq->rcv_wscale = tcp_opt.rcv_wscale;
+ ireq->sack_ok = tcp_opt.sack_ok;
+ ireq->wscale_ok = tcp_opt.wscale_ok;
+ ireq->tstamp_ok = tcp_opt.saw_tstamp;
+
req->window_clamp = tp->window_clamp ? :dst_metric(dst, RTAX_WINDOW);
tcp_select_initial_window(tcp_full_space(sk), req->mss,
&req->rcv_wnd, &req->window_clamp,
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 21d100b..2eebab5 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -1165,6 +1165,7 @@ static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
struct tcp_sock *tp = tcp_sk(sk);
struct request_sock *req = NULL;
__u32 isn = TCP_SKB_CB(skb)->when;
+ struct dst_entry *dst = __sk_dst_get(sk);
#ifdef CONFIG_SYN_COOKIES
int want_cookie = 0;
#else
@@ -1203,7 +1204,7 @@ static int tcp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
tmp_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr);
tmp_opt.user_mss = tp->rx_opt.user_mss;
- tcp_parse_options(skb, &tmp_opt, 0);
+ tcp_parse_options(skb, &tmp_opt, 0, dst);
if (want_cookie && !tmp_opt.saw_tstamp)
tcp_clear_options(&tmp_opt);
--
1.5.6.3
^ permalink raw reply related
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox