Skip to content

L
linux
Commit 35d75fa0 authored by Jeb J. Cramer's avatar Jeb J. Cramer Committed by Jeff Garzik
Browse files
Options

[E1000] Modulus math removed 

* Removed modulus math; decreases CPU utilization, especially on PPC64
  [anton@samba.org]
parent 6497b3ff
Hide whitespace changes
Inline Side-by-side
Showing with 32 additions and 60 deletions
drivers/net/e1000/e1000.h
View file @ 35d75fa0
......@@ -117,7 +117,7 @@ struct e1000_adapter;
/* How many Tx Descriptors do we need to call netif_wake_queue ? */
#define E1000_TX_QUEUE_WAKE 16
/* How many Rx Buffers do we bundle into one write to the hardware ? */
#define E1000_RX_BUFFER_WRITE 16
#define E1000_RX_BUFFER_WRITE 16 /* Must be power of 2 */
#define AUTO_ALL_MODES 0
#define E1000_EEPROM_APME 4
......@@ -152,7 +152,8 @@ struct e1000_desc_ring {
};
#define E1000_DESC_UNUSED(R) \
((((R)->next_to_clean + (R)->count) - ((R)->next_to_use + 1)) % ((R)->count))
((((R)->next_to_clean > (R)->next_to_use) ? 0 : (R)->count) + \
(R)->next_to_clean - (R)->next_to_use - 1)
#define E1000_GET_DESC(R, i, type) (&(((struct type *)((R).desc))[i]))
#define E1000_RX_DESC(R, i) E1000_GET_DESC(R, i, e1000_rx_desc)
......@@ -186,7 +187,6 @@ struct e1000_adapter {
uint32_t txd_cmd;
uint32_t tx_int_delay;
uint32_t tx_abs_int_delay;
int max_data_per_txd;
uint32_t tx_fifo_head;
uint32_t tx_head_addr;
uint32_t tx_fifo_size;
......
drivers/net/e1000/e1000_main.c
View file @ 35d75fa0
......@@ -155,12 +155,8 @@ static void e1000_update_stats(struct e1000_adapter *adapter);
static inline void e1000_irq_disable(struct e1000_adapter *adapter);
static inline void e1000_irq_enable(struct e1000_adapter *adapter);
static void e1000_intr(int irq, void *data, struct pt_regs *regs);
static void e1000_clean_tx_irq(struct e1000_adapter *adapter);
#ifdef CONFIG_E1000_NAPI
static int e1000_poll(struct net_device *netdev, int *budget);
#else
static void e1000_clean_rx_irq(struct e1000_adapter *adapter);
#endif
static boolean_t e1000_clean_tx_irq(struct e1000_adapter *adapter);
static boolean_t e1000_clean_rx_irq(struct e1000_adapter *adapter);
static void e1000_alloc_rx_buffers(struct e1000_adapter *adapter);
static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd);
static int e1000_mii_ioctl(struct net_device *netdev, struct ifreq *ifr,
......@@ -420,10 +416,6 @@ e1000_probe(struct pci_dev *pdev,
netdev->change_mtu = &e1000_change_mtu;
netdev->do_ioctl = &e1000_ioctl;
netdev->tx_timeout = &e1000_tx_timeout;
#ifdef CONFIG_E1000_NAPI
netdev->poll = &e1000_poll;
netdev->weight = 64;
#endif
netdev->watchdog_timeo = 5 * HZ;
netdev->vlan_rx_register = e1000_vlan_rx_register;
netdev->vlan_rx_add_vid = e1000_vlan_rx_add_vid;
......@@ -478,13 +470,6 @@ e1000_probe(struct pci_dev *pdev,
e1000_get_bus_info(&adapter->hw);
if((adapter->hw.mac_type == e1000_82544) &&
(adapter->hw.bus_type == e1000_bus_type_pcix))
adapter->max_data_per_txd = 4096;
else
adapter->max_data_per_txd = MAX_JUMBO_FRAME_SIZE;
init_timer(&adapter->tx_fifo_stall_timer);
adapter->tx_fifo_stall_timer.function = &e1000_82547_tx_fifo_stall;
adapter->tx_fifo_stall_timer.data = (unsigned long) adapter;
......@@ -1423,12 +1408,12 @@ e1000_tso(struct e1000_adapter *adapter, struct sk_buff *skb, int tx_flags)
E1000_TXD_CMD_IP | E1000_TXD_CMD_TCP |
(skb->len - (hdr_len)));
i = (i + 1) % adapter->tx_ring.count;
if(++i == adapter->tx_ring.count) i = 0;
adapter->tx_ring.next_to_use = i;
return TRUE;
}
return FALSE;
}
......@@ -1453,7 +1438,7 @@ e1000_tx_csum(struct e1000_adapter *adapter, struct sk_buff *skb)
context_desc->cmd_and_length =
cpu_to_le32(adapter->txd_cmd | E1000_TXD_CMD_DEXT);
i = (i + 1) % adapter->tx_ring.count;
if(++i == adapter->tx_ring.count) i = 0;
adapter->tx_ring.next_to_use = i;
return TRUE;
......@@ -1462,24 +1447,24 @@ e1000_tx_csum(struct e1000_adapter *adapter, struct sk_buff *skb)
return FALSE;
}
#define E1000_MAX_TXD_PWR 12
#define E1000_MAX_DATA_PER_TXD (1<<E1000_MAX_TXD_PWR)
static inline int
e1000_tx_map(struct e1000_adapter *adapter, struct sk_buff *skb)
{
struct e1000_desc_ring *tx_ring = &adapter->tx_ring;
int len, offset, size, count, i;
int len = skb->len, offset = 0, size, count = 0, i;
int tso = skb_shinfo(skb)->tso_size;
int nr_frags = skb_shinfo(skb)->nr_frags;
int f;
len = skb->len - skb->data_len;
i = (tx_ring->next_to_use + tx_ring->count - 1) % tx_ring->count;
count = 0;
len -= skb->data_len;
offset = 0;
i = tx_ring->next_to_use;
while(len) {
i = (i + 1) % tx_ring->count;
size = min(len, adapter->max_data_per_txd);
size = min(len, E1000_MAX_DATA_PER_TXD);
/* Workaround for premature desc write-backs
* in TSO mode. Append 4-byte sentinel desc */
if(tso && !nr_frags && size == len && size > 4)
......@@ -1495,6 +1480,7 @@ e1000_tx_map(struct e1000_adapter *adapter, struct sk_buff *skb)
len -= size;
offset += size;
count++;
if(++i == tx_ring->count) i = 0;
}
for(f = 0; f < nr_frags; f++) {
......@@ -1505,8 +1491,7 @@ e1000_tx_map(struct e1000_adapter *adapter, struct sk_buff *skb)
offset = 0;
while(len) {
i = (i + 1) % tx_ring->count;
size = min(len, adapter->max_data_per_txd);
size = min(len, E1000_MAX_DATA_PER_TXD);
/* Workaround for premature desc write-backs
* in TSO mode. Append 4-byte sentinel desc */
if(tso && f == (nr_frags-1) && size == len && size > 4)
......@@ -1523,8 +1508,10 @@ e1000_tx_map(struct e1000_adapter *adapter, struct sk_buff *skb)
len -= size;
offset += size;
count++;
if(++i == tx_ring->count) i = 0;
}
}
if(--i < 0) i = tx_ring->count - 1;
tx_ring->buffer_info[i].skb = skb;
return count;
......@@ -1565,7 +1552,7 @@ e1000_tx_queue(struct e1000_adapter *adapter, int count, int tx_flags)
tx_desc->lower.data =
cpu_to_le32(txd_lower | tx_ring->buffer_info[i].length);
tx_desc->upper.data = cpu_to_le32(txd_upper);
i = (i + 1) % tx_ring->count;
if(++i == tx_ring->count) i = 0;
}
tx_desc->lower.data |= cpu_to_le32(E1000_TXD_CMD_EOP);
......@@ -1580,7 +1567,6 @@ e1000_tx_queue(struct e1000_adapter *adapter, int count, int tx_flags)
E1000_WRITE_REG(&adapter->hw, TDT, i);
}
#define TXD_USE_COUNT(S, X) (((S) / (X)) + (((S) % (X)) ? 1 : 0))
/**
* 82547 workaround to avoid controller hang in half-duplex environment.
* The workaround is to avoid queuing a large packet that would span
......@@ -1629,24 +1615,14 @@ static int
e1000_xmit_frame(struct sk_buff *skb, struct net_device *netdev)
{
struct e1000_adapter *adapter = netdev->priv;
int tx_flags = 0, count;
int f;
int tx_flags = 0;
count = TXD_USE_COUNT(skb->len - skb->data_len,
adapter->max_data_per_txd);
if(count == 0) {
if(skb->len <= 0) {
dev_kfree_skb_any(skb);
return 0;
}
for(f = 0; f < skb_shinfo(skb)->nr_frags; f++)
count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size,
adapter->max_data_per_txd);
if((skb_shinfo(skb)->tso_size) || (skb->ip_summed == CHECKSUM_HW))
count++;
if(E1000_DESC_UNUSED(&adapter->tx_ring) < count) {
if(E1000_DESC_UNUSED(&adapter->tx_ring) < DESC_NEEDED) {
netif_stop_queue(netdev);
return 1;
}
......@@ -1669,9 +1645,7 @@ e1000_xmit_frame(struct sk_buff *skb, struct net_device *netdev)
else if(e1000_tx_csum(adapter, skb))
tx_flags |= E1000_TX_FLAGS_CSUM;
count = e1000_tx_map(adapter, skb);
e1000_tx_queue(adapter, count, tx_flags);
e1000_tx_queue(adapter, e1000_tx_map(adapter, skb), tx_flags);
netdev->trans_start = jiffies;
......@@ -2041,7 +2015,7 @@ e1000_clean_tx_irq(struct e1000_adapter *adapter)
tx_desc->upper.data = 0;
i = (i + 1) % tx_ring->count;
if(++i == tx_ring->count) i = 0;
tx_desc = E1000_TX_DESC(*tx_ring, i);
}
......@@ -2225,7 +2199,7 @@ e1000_clean_rx_irq(struct e1000_adapter *adapter)
rx_desc->status = 0;
rx_ring->buffer_info[i].skb = NULL;
i = (i + 1) % rx_ring->count;
if(++i == rx_ring->count) i = 0;
rx_desc = E1000_RX_DESC(*rx_ring, i);
continue;
......@@ -2253,7 +2227,7 @@ e1000_clean_rx_irq(struct e1000_adapter *adapter)
rx_desc->status = 0;
rx_ring->buffer_info[i].skb = NULL;
i = (i + 1) % rx_ring->count;
if(++i == rx_ring->count) i = 0;
rx_desc = E1000_RX_DESC(*rx_ring, i);
continue;
......@@ -2278,7 +2252,7 @@ e1000_clean_rx_irq(struct e1000_adapter *adapter)
rx_desc->status = 0;
rx_ring->buffer_info[i].skb = NULL;
i = (i + 1) % rx_ring->count;
if(++i == rx_ring->count) i = 0;
rx_desc = E1000_RX_DESC(*rx_ring, i);
}
......@@ -2302,11 +2276,9 @@ e1000_alloc_rx_buffers(struct e1000_adapter *adapter)
struct pci_dev *pdev = adapter->pdev;
struct e1000_rx_desc *rx_desc;
struct sk_buff *skb;
int reserve_len;
int reserve_len = 2;
int i;
reserve_len = 2;
i = rx_ring->next_to_use;
while(!rx_ring->buffer_info[i].skb) {
......@@ -2337,7 +2309,7 @@ e1000_alloc_rx_buffers(struct e1000_adapter *adapter)
rx_desc->buffer_addr = cpu_to_le64(rx_ring->buffer_info[i].dma);
if(!(i % E1000_RX_BUFFER_WRITE)) {
if((i & ~(E1000_RX_BUFFER_WRITE - 1)) == i) {
/* Force memory writes to complete before letting h/w
* know there are new descriptors to fetch. (Only
* applicable for weak-ordered memory model archs,
......@@ -2347,7 +2319,7 @@ e1000_alloc_rx_buffers(struct e1000_adapter *adapter)
E1000_WRITE_REG(&adapter->hw, RDT, i);
}
i = (i + 1) % rx_ring->count;
if(++i == rx_ring->count) i = 0;
}
rx_ring->next_to_use = i;
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7