* Re: Hight speed data sending from custom IP out of kernel
From: juice @ 2011-04-19 11:34 UTC (permalink / raw)
To: monstr, netdev
Hi Michal.
How fast do you need to send the data?
I have an application where I send test stream out to GE line and can fill
the total capacity of the ethernet regardless of the packet size.
The test stream I am sending is stored in kernel memory, and therefore is
limited by the amount of free memory. 200M is no problem.
The solution I am using is loosely based on the pktgen module, except that
my module can load a wireshark capture from userland program and then send
it from ethernet interface in wire speed.
- Juice -
> Hi,
> I would like to create demo for high speed data sending from custom IP
through
> the ethernet. I think the best description is that there are dmaable memory
> mapped registers or just memory which store data I want to send (for
example 200MB).
> Linux should handle all communication between target(probably server)
and
> host
> (client) but data in the packets should go from that custom IP and can't go
> through the kernel because of performance issue.
> Ethernet core have own DMA which I could use but the question is if
there
> is any
> option how to convince the kernel that data will go directly from memory
mapped
> registers and the kernel/driver/... just setup dma BD for headers and
second for
> data.
> Do you have any experience with any solution with passing data
completely
> out of
> kernel?
> Thanks,
> Michal
> --
> Michal Simek, Ing. (M.Eng)
> w: www.monstr.eu p: +42-0-721842854
> Maintainer of Linux kernel 2.6 Microblaze Linux -
> http://www.monstr.eu/fdt/
> Microblaze U-BOOT custodian
> --
> 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: Hight speed data sending from custom IP out of kernel
From: Michal Simek @ 2011-04-19 11:50 UTC (permalink / raw)
To: juice; +Cc: netdev
In-Reply-To: <53f539ea861e1a24bde4aadceff0b3bb.squirrel@www.liukuma.net>
Hi Juice,
juice wrote:
> Hi Michal.
>
> How fast do you need to send the data?
It sounds weird but as fast as possible. There is no specific limit because I
want to create demo and test it on various hw configuration which I can easily
create on FPGA. For now the bottleneck is Microblaze cpu. It can run from 50MHz
till 170-180MHz. We also support both endians and have two hw IP
cores(10/100/1000) which I can use.
> I have an application where I send test stream out to GE line and can fill
> the total capacity of the ethernet regardless of the packet size.
What cpu do you use?
>
> The test stream I am sending is stored in kernel memory, and therefore is
> limited by the amount of free memory. 200M is no problem.
Is it UDP or TCP?
>
> The solution I am using is loosely based on the pktgen module, except that
> my module can load a wireshark capture from userland program and then send
> it from ethernet interface in wire speed.
Sound good. Would it be possible to see it and test it?
Thanks,
Michal
>
> - Juice -
>
>
>> Hi,
>> I would like to create demo for high speed data sending from custom IP
> through
>> the ethernet. I think the best description is that there are dmaable memory
>> mapped registers or just memory which store data I want to send (for
> example 200MB).
>> Linux should handle all communication between target(probably server)
> and
>> host
>> (client) but data in the packets should go from that custom IP and can't go
>> through the kernel because of performance issue.
>> Ethernet core have own DMA which I could use but the question is if
> there
>> is any
>> option how to convince the kernel that data will go directly from memory
> mapped
>> registers and the kernel/driver/... just setup dma BD for headers and
> second for
>> data.
>> Do you have any experience with any solution with passing data
> completely
>> out of
>> kernel?
>> Thanks,
>> Michal
>> --
>> Michal Simek, Ing. (M.Eng)
>> w: www.monstr.eu p: +42-0-721842854
>> Maintainer of Linux kernel 2.6 Microblaze Linux -
>> http://www.monstr.eu/fdt/
>> Microblaze U-BOOT custodian
>> --
>> 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
>
>
>
>
--
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian
^ permalink raw reply
* [PATCH] net: pch_gbe: convert to hw_features
From: Michał Mirosław @ 2011-04-19 11:56 UTC (permalink / raw)
To: netdev; +Cc: Toshiharu Okada, Masayuki Ohtake
This also fixes bug in xmit path, where TX checksum offload state was used
instead of skb->ip_summed to decide if the offload was needed.
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
---
drivers/net/pch_gbe/pch_gbe.h | 4 --
drivers/net/pch_gbe/pch_gbe_ethtool.c | 54 ---------------------------------
drivers/net/pch_gbe/pch_gbe_main.c | 42 +++++++++++++++++++------
drivers/net/pch_gbe/pch_gbe_param.c | 16 ++++++----
4 files changed, 42 insertions(+), 74 deletions(-)
diff --git a/drivers/net/pch_gbe/pch_gbe.h b/drivers/net/pch_gbe/pch_gbe.h
index bf126e7..59fac77 100644
--- a/drivers/net/pch_gbe/pch_gbe.h
+++ b/drivers/net/pch_gbe/pch_gbe.h
@@ -597,8 +597,6 @@ struct pch_gbe_hw_stats {
* @rx_ring: Pointer of Rx descriptor ring structure
* @rx_buffer_len: Receive buffer length
* @tx_queue_len: Transmit queue length
- * @rx_csum: Receive TCP/IP checksum enable/disable
- * @tx_csum: Transmit TCP/IP checksum enable/disable
* @have_msi: PCI MSI mode flag
*/
@@ -623,8 +621,6 @@ struct pch_gbe_adapter {
struct pch_gbe_rx_ring *rx_ring;
unsigned long rx_buffer_len;
unsigned long tx_queue_len;
- bool rx_csum;
- bool tx_csum;
bool have_msi;
};
diff --git a/drivers/net/pch_gbe/pch_gbe_ethtool.c b/drivers/net/pch_gbe/pch_gbe_ethtool.c
index d2174a4..c35d105 100644
--- a/drivers/net/pch_gbe/pch_gbe_ethtool.c
+++ b/drivers/net/pch_gbe/pch_gbe_ethtool.c
@@ -434,57 +434,6 @@ static int pch_gbe_set_pauseparam(struct net_device *netdev,
}
/**
- * pch_gbe_get_rx_csum - Report whether receive checksums are turned on or off
- * @netdev: Network interface device structure
- * Returns
- * true(1): Checksum On
- * false(0): Checksum Off
- */
-static u32 pch_gbe_get_rx_csum(struct net_device *netdev)
-{
- struct pch_gbe_adapter *adapter = netdev_priv(netdev);
-
- return adapter->rx_csum;
-}
-
-/**
- * pch_gbe_set_rx_csum - Turn receive checksum on or off
- * @netdev: Network interface device structure
- * @data: Checksum On[true] or Off[false]
- * Returns
- * 0: Successful.
- * Negative value: Failed.
- */
-static int pch_gbe_set_rx_csum(struct net_device *netdev, u32 data)
-{
- struct pch_gbe_adapter *adapter = netdev_priv(netdev);
-
- adapter->rx_csum = data;
- if ((netif_running(netdev)))
- pch_gbe_reinit_locked(adapter);
- else
- pch_gbe_reset(adapter);
-
- return 0;
-}
-
-/**
- * pch_gbe_set_tx_csum - Turn transmit checksums on or off
- * @netdev: Network interface device structure
- * @data: Checksum on[true] or off[false]
- * Returns
- * 0: Successful.
- * Negative value: Failed.
- */
-static int pch_gbe_set_tx_csum(struct net_device *netdev, u32 data)
-{
- struct pch_gbe_adapter *adapter = netdev_priv(netdev);
-
- adapter->tx_csum = data;
- return ethtool_op_set_tx_ipv6_csum(netdev, data);
-}
-
-/**
* pch_gbe_get_strings - Return a set of strings that describe the requested
* objects
* @netdev: Network interface device structure
@@ -554,9 +503,6 @@ static const struct ethtool_ops pch_gbe_ethtool_ops = {
.set_ringparam = pch_gbe_set_ringparam,
.get_pauseparam = pch_gbe_get_pauseparam,
.set_pauseparam = pch_gbe_set_pauseparam,
- .get_rx_csum = pch_gbe_get_rx_csum,
- .set_rx_csum = pch_gbe_set_rx_csum,
- .set_tx_csum = pch_gbe_set_tx_csum,
.get_strings = pch_gbe_get_strings,
.get_ethtool_stats = pch_gbe_get_ethtool_stats,
.get_sset_count = pch_gbe_get_sset_count,
diff --git a/drivers/net/pch_gbe/pch_gbe_main.c b/drivers/net/pch_gbe/pch_gbe_main.c
index 2ef2f9c..4cc9872 100644
--- a/drivers/net/pch_gbe/pch_gbe_main.c
+++ b/drivers/net/pch_gbe/pch_gbe_main.c
@@ -656,6 +656,7 @@ static void pch_gbe_configure_tx(struct pch_gbe_adapter *adapter)
*/
static void pch_gbe_setup_rctl(struct pch_gbe_adapter *adapter)
{
+ struct net_device *netdev = adapter->netdev;
struct pch_gbe_hw *hw = &adapter->hw;
u32 rx_mode, tcpip;
@@ -666,7 +667,7 @@ static void pch_gbe_setup_rctl(struct pch_gbe_adapter *adapter)
tcpip = ioread32(&hw->reg->TCPIP_ACC);
- if (adapter->rx_csum) {
+ if (netdev->features & NETIF_F_RXCSUM) {
tcpip &= ~PCH_GBE_RX_TCPIPACC_OFF;
tcpip |= PCH_GBE_RX_TCPIPACC_EN;
} else {
@@ -950,7 +951,7 @@ static void pch_gbe_tx_queue(struct pch_gbe_adapter *adapter,
frame_ctrl = 0;
if (unlikely(skb->len < PCH_GBE_SHORT_PKT))
frame_ctrl |= PCH_GBE_TXD_CTRL_APAD;
- if (unlikely(!adapter->tx_csum))
+ if (skb->ip_summed == CHECKSUM_NONE)
frame_ctrl |= PCH_GBE_TXD_CTRL_TCPIP_ACC_OFF;
/* Performs checksum processing */
@@ -958,7 +959,7 @@ static void pch_gbe_tx_queue(struct pch_gbe_adapter *adapter,
* It is because the hardware accelerator does not support a checksum,
* when the received data size is less than 64 bytes.
*/
- if ((skb->len < PCH_GBE_SHORT_PKT) && (adapter->tx_csum)) {
+ if (skb->len < PCH_GBE_SHORT_PKT && skb->ip_summed != CHECKSUM_NONE) {
frame_ctrl |= PCH_GBE_TXD_CTRL_APAD |
PCH_GBE_TXD_CTRL_TCPIP_ACC_OFF;
if (skb->protocol == htons(ETH_P_IP)) {
@@ -1426,7 +1427,7 @@ pch_gbe_clean_rx(struct pch_gbe_adapter *adapter,
length = (rx_desc->rx_words_eob) - 3;
/* Decide the data conversion method */
- if (!adapter->rx_csum) {
+ if (!(netdev->features & NETIF_F_RXCSUM)) {
/* [Header:14][payload] */
if (NET_IP_ALIGN) {
/* Because alignment differs,
@@ -2030,6 +2031,29 @@ static int pch_gbe_change_mtu(struct net_device *netdev, int new_mtu)
}
/**
+ * pch_gbe_set_features - Reset device after features changed
+ * @netdev: Network interface device structure
+ * @features: New features
+ * Returns
+ * 0: HW state updated successfully
+ */
+static int pch_gbe_set_features(struct net_device *netdev, u32 features)
+{
+ struct pch_gbe_adapter *adapter = netdev_priv(netdev);
+ u32 changed = features ^ netdev->features;
+
+ if (!(changed & NETIF_F_RXCSUM))
+ return 0;
+
+ if (netif_running(netdev))
+ pch_gbe_reinit_locked(adapter);
+ else
+ pch_gbe_reset(adapter);
+
+ return 0;
+}
+
+/**
* pch_gbe_ioctl - Controls register through a MII interface
* @netdev: Network interface device structure
* @ifr: Pointer to ifr structure
@@ -2129,6 +2153,7 @@ static const struct net_device_ops pch_gbe_netdev_ops = {
.ndo_set_mac_address = pch_gbe_set_mac,
.ndo_tx_timeout = pch_gbe_tx_timeout,
.ndo_change_mtu = pch_gbe_change_mtu,
+ .ndo_set_features = pch_gbe_set_features,
.ndo_do_ioctl = pch_gbe_ioctl,
.ndo_set_multicast_list = &pch_gbe_set_multi,
#ifdef CONFIG_NET_POLL_CONTROLLER
@@ -2334,7 +2359,9 @@ static int pch_gbe_probe(struct pci_dev *pdev,
netdev->watchdog_timeo = PCH_GBE_WATCHDOG_PERIOD;
netif_napi_add(netdev, &adapter->napi,
pch_gbe_napi_poll, PCH_GBE_RX_WEIGHT);
- netdev->features = NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | NETIF_F_GRO;
+ netdev->hw_features = NETIF_F_RXCSUM |
+ NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
+ netdev->features = netdev->hw_features;
pch_gbe_set_ethtool_ops(netdev);
pch_gbe_mac_load_mac_addr(&adapter->hw);
@@ -2373,11 +2400,6 @@ static int pch_gbe_probe(struct pci_dev *pdev,
pch_gbe_check_options(adapter);
- if (adapter->tx_csum)
- netdev->features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
- else
- netdev->features &= ~(NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM);
-
/* initialize the wol settings based on the eeprom settings */
adapter->wake_up_evt = PCH_GBE_WL_INIT_SETTING;
dev_info(&pdev->dev, "MAC address : %pM\n", netdev->dev_addr);
diff --git a/drivers/net/pch_gbe/pch_gbe_param.c b/drivers/net/pch_gbe/pch_gbe_param.c
index ef0996a..5b5d90a 100644
--- a/drivers/net/pch_gbe/pch_gbe_param.c
+++ b/drivers/net/pch_gbe/pch_gbe_param.c
@@ -426,6 +426,8 @@ full_duplex_only:
void pch_gbe_check_options(struct pch_gbe_adapter *adapter)
{
struct pch_gbe_hw *hw = &adapter->hw;
+ struct net_device *dev = adapter->netdev;
+ int val;
{ /* Transmit Descriptor Count */
static const struct pch_gbe_option opt = {
@@ -466,9 +468,10 @@ void pch_gbe_check_options(struct pch_gbe_adapter *adapter)
.err = "defaulting to Enabled",
.def = PCH_GBE_DEFAULT_RX_CSUM
};
- adapter->rx_csum = XsumRX;
- pch_gbe_validate_option((int *)(&adapter->rx_csum),
- &opt, adapter);
+ val = XsumRX;
+ pch_gbe_validate_option(&val, &opt, adapter);
+ if (!val)
+ dev->features &= ~NETIF_F_RXCSUM;
}
{ /* Checksum Offload Enable/Disable */
static const struct pch_gbe_option opt = {
@@ -477,9 +480,10 @@ void pch_gbe_check_options(struct pch_gbe_adapter *adapter)
.err = "defaulting to Enabled",
.def = PCH_GBE_DEFAULT_TX_CSUM
};
- adapter->tx_csum = XsumTX;
- pch_gbe_validate_option((int *)(&adapter->tx_csum),
- &opt, adapter);
+ val = XsumTX;
+ pch_gbe_validate_option(&val, &opt, adapter);
+ if (!val)
+ dev->features &= ~NETIF_F_ALL_CSUM;
}
{ /* Flow Control */
static const struct pch_gbe_option opt = {
--
1.7.2.5
^ permalink raw reply related
* [PATCH] net: xen-netback: convert to hw_features
From: Michał Mirosław @ 2011-04-19 11:56 UTC (permalink / raw)
To: netdev; +Cc: Ian Campbell, xen-devel
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
---
drivers/net/xen-netback/common.h | 3 -
drivers/net/xen-netback/interface.c | 84 ++++++----------------------------
2 files changed, 15 insertions(+), 72 deletions(-)
diff --git a/drivers/net/xen-netback/common.h b/drivers/net/xen-netback/common.h
index 5d7bbf2..8753e6d 100644
--- a/drivers/net/xen-netback/common.h
+++ b/drivers/net/xen-netback/common.h
@@ -73,9 +73,6 @@ struct xenvif {
struct vm_struct *tx_comms_area;
struct vm_struct *rx_comms_area;
- /* Flags that must not be set in dev->features */
- u32 features_disabled;
-
/* Frontend feature information. */
u8 can_sg:1;
u8 gso:1;
diff --git a/drivers/net/xen-netback/interface.c b/drivers/net/xen-netback/interface.c
index de569cc..fe25308 100644
--- a/drivers/net/xen-netback/interface.c
+++ b/drivers/net/xen-netback/interface.c
@@ -165,69 +165,18 @@ static int xenvif_change_mtu(struct net_device *dev, int mtu)
return 0;
}
-static void xenvif_set_features(struct xenvif *vif)
+static u32 xenvif_fix_features(struct xenvif *vif, u32 features)
{
struct net_device *dev = vif->dev;
- u32 features = dev->features;
- if (vif->can_sg)
- features |= NETIF_F_SG;
- if (vif->gso || vif->gso_prefix)
- features |= NETIF_F_TSO;
- if (vif->csum)
- features |= NETIF_F_IP_CSUM;
+ if (!vif->can_sg)
+ features &= ~NETIF_F_SG;
+ if (!vif->gso && !vif->gso_prefix)
+ features &= ~NETIF_F_TSO;
+ if (!vif->csum)
+ features &= ~NETIF_F_IP_CSUM;
- features &= ~(vif->features_disabled);
-
- if (!(features & NETIF_F_SG) && dev->mtu > ETH_DATA_LEN)
- dev->mtu = ETH_DATA_LEN;
-
- dev->features = features;
-}
-
-static int xenvif_set_tx_csum(struct net_device *dev, u32 data)
-{
- struct xenvif *vif = netdev_priv(dev);
- if (data) {
- if (!vif->csum)
- return -EOPNOTSUPP;
- vif->features_disabled &= ~NETIF_F_IP_CSUM;
- } else {
- vif->features_disabled |= NETIF_F_IP_CSUM;
- }
-
- xenvif_set_features(vif);
- return 0;
-}
-
-static int xenvif_set_sg(struct net_device *dev, u32 data)
-{
- struct xenvif *vif = netdev_priv(dev);
- if (data) {
- if (!vif->can_sg)
- return -EOPNOTSUPP;
- vif->features_disabled &= ~NETIF_F_SG;
- } else {
- vif->features_disabled |= NETIF_F_SG;
- }
-
- xenvif_set_features(vif);
- return 0;
-}
-
-static int xenvif_set_tso(struct net_device *dev, u32 data)
-{
- struct xenvif *vif = netdev_priv(dev);
- if (data) {
- if (!vif->gso && !vif->gso_prefix)
- return -EOPNOTSUPP;
- vif->features_disabled &= ~NETIF_F_TSO;
- } else {
- vif->features_disabled |= NETIF_F_TSO;
- }
-
- xenvif_set_features(vif);
- return 0;
+ return features;
}
static const struct xenvif_stat {
@@ -274,12 +223,6 @@ static void xenvif_get_strings(struct net_device *dev, u32 stringset, u8 * data)
}
static struct ethtool_ops xenvif_ethtool_ops = {
- .get_tx_csum = ethtool_op_get_tx_csum,
- .set_tx_csum = xenvif_set_tx_csum,
- .get_sg = ethtool_op_get_sg,
- .set_sg = xenvif_set_sg,
- .get_tso = ethtool_op_get_tso,
- .set_tso = xenvif_set_tso,
.get_link = ethtool_op_get_link,
.get_sset_count = xenvif_get_sset_count,
@@ -293,6 +236,7 @@ static struct net_device_ops xenvif_netdev_ops = {
.ndo_open = xenvif_open,
.ndo_stop = xenvif_close,
.ndo_change_mtu = xenvif_change_mtu,
+ .ndo_fix_features = xenvif_fix_features,
};
struct xenvif *xenvif_alloc(struct device *parent, domid_t domid,
@@ -331,7 +275,8 @@ struct xenvif *xenvif_alloc(struct device *parent, domid_t domid,
vif->credit_timeout.expires = jiffies;
dev->netdev_ops = &xenvif_netdev_ops;
- xenvif_set_features(vif);
+ dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_TSO;
+ dev->features = dev->hw_features;
SET_ETHTOOL_OPS(dev, &xenvif_ethtool_ops);
dev->tx_queue_len = XENVIF_QUEUE_LENGTH;
@@ -367,8 +312,6 @@ int xenvif_connect(struct xenvif *vif, unsigned long tx_ring_ref,
if (vif->irq)
return 0;
- xenvif_set_features(vif);
-
err = xen_netbk_map_frontend_rings(vif, tx_ring_ref, rx_ring_ref);
if (err < 0)
goto err;
@@ -384,9 +327,12 @@ int xenvif_connect(struct xenvif *vif, unsigned long tx_ring_ref,
xenvif_get(vif);
rtnl_lock();
- netif_carrier_on(vif->dev);
if (netif_running(vif->dev))
xenvif_up(vif);
+ if (!vif->can_sg && vif->dev->mtu > ETH_DATA_LEN)
+ dev_set_mtu(vif->dev, ETH_DATA_LEN);
+ netdev_update_features(vif->dev);
+ netif_carrier_on(vif->dev);
rtnl_unlock();
return 0;
--
1.7.2.5
^ permalink raw reply related
* [PATCH] bonding: 802.3ad - fix agg_device_up
From: Jiri Bohac @ 2011-04-19 12:09 UTC (permalink / raw)
To: netdev; +Cc: Jay Vosburgh, Andy Gospodarek, Stephen Hemminger
The slave member of struct aggregator does not necessarily point
to a slave which is part of the aggregator. It points to the
slave structure containing the aggregator structure, while
completely different slaves (or no slaves at all) may be part of
the aggregator.
The agg_device_up() function wrongly uses agg->slave to find the state of the
aggregator. Use agg->lag_ports->slave instead. The bug has been
introduced by commit 4cd6fe1c6483cde93e2ec91f58b7af9c9eea51ad.
Signed-off-by: Jiri Bohac <jbohac@suse.cz>
diff --git a/drivers/net/bonding/bond_3ad.c b/drivers/net/bonding/bond_3ad.c
index 494bf96..31912f1 100644
--- a/drivers/net/bonding/bond_3ad.c
+++ b/drivers/net/bonding/bond_3ad.c
@@ -1482,8 +1482,11 @@ static struct aggregator *ad_agg_selection_test(struct aggregator *best,
static int agg_device_up(const struct aggregator *agg)
{
- return (netif_running(agg->slave->dev) &&
- netif_carrier_ok(agg->slave->dev));
+ struct port *port = agg->lag_ports;
+ if (!port)
+ return 0;
+ return (netif_running(port->slave->dev) &&
+ netif_carrier_ok(port->slave->dev));
}
/**
--
Jiri Bohac <jbohac@suse.cz>
SUSE Labs, SUSE CZ
^ permalink raw reply related
* [PATCH] net: ibmveth: convert to hw_features
From: Michał Mirosław @ 2011-04-19 12:14 UTC (permalink / raw)
To: netdev; +Cc: Santiago Leon
A minimal conversion.
ibmveth_set_csum_offload() can be folded into ibmveth_set_features()
and adapter->rx_csum removed - left for later cleanup.
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
---
drivers/net/ibmveth.c | 96 ++++++++++++++-----------------------------------
1 files changed, 27 insertions(+), 69 deletions(-)
diff --git a/drivers/net/ibmveth.c b/drivers/net/ibmveth.c
index 5522d45..4855f1f 100644
--- a/drivers/net/ibmveth.c
+++ b/drivers/net/ibmveth.c
@@ -729,45 +729,24 @@ static void netdev_get_drvinfo(struct net_device *dev,
sizeof(info->version) - 1);
}
-static void ibmveth_set_rx_csum_flags(struct net_device *dev, u32 data)
+static u32 ibmveth_fix_features(struct net_device *dev, u32 features)
{
- struct ibmveth_adapter *adapter = netdev_priv(dev);
+ /*
+ * Since the ibmveth firmware interface does not have the
+ * concept of separate tx/rx checksum offload enable, if rx
+ * checksum is disabled we also have to disable tx checksum
+ * offload. Once we disable rx checksum offload, we are no
+ * longer allowed to send tx buffers that are not properly
+ * checksummed.
+ */
- if (data) {
- adapter->rx_csum = 1;
- } else {
- /*
- * Since the ibmveth firmware interface does not have the
- * concept of separate tx/rx checksum offload enable, if rx
- * checksum is disabled we also have to disable tx checksum
- * offload. Once we disable rx checksum offload, we are no
- * longer allowed to send tx buffers that are not properly
- * checksummed.
- */
- adapter->rx_csum = 0;
- dev->features &= ~NETIF_F_IP_CSUM;
- dev->features &= ~NETIF_F_IPV6_CSUM;
- }
-}
-
-static void ibmveth_set_tx_csum_flags(struct net_device *dev, u32 data)
-{
- struct ibmveth_adapter *adapter = netdev_priv(dev);
+ if (!(features & NETIF_F_RXCSUM))
+ features &= ~NETIF_F_ALL_CSUM;
- if (data) {
- if (adapter->fw_ipv4_csum_support)
- dev->features |= NETIF_F_IP_CSUM;
- if (adapter->fw_ipv6_csum_support)
- dev->features |= NETIF_F_IPV6_CSUM;
- adapter->rx_csum = 1;
- } else {
- dev->features &= ~NETIF_F_IP_CSUM;
- dev->features &= ~NETIF_F_IPV6_CSUM;
- }
+ return features;
}
-static int ibmveth_set_csum_offload(struct net_device *dev, u32 data,
- void (*done) (struct net_device *, u32))
+static int ibmveth_set_csum_offload(struct net_device *dev, u32 data)
{
struct ibmveth_adapter *adapter = netdev_priv(dev);
unsigned long set_attr, clr_attr, ret_attr;
@@ -827,8 +806,8 @@ static int ibmveth_set_csum_offload(struct net_device *dev, u32 data,
} else
adapter->fw_ipv6_csum_support = data;
- if (ret == H_SUCCESS || ret6 == H_SUCCESS)
- done(dev, data);
+ if (ret != H_SUCCESS || ret6 != H_SUCCESS)
+ adapter->rx_csum = data;
else
rc1 = -EIO;
} else {
@@ -844,41 +823,22 @@ static int ibmveth_set_csum_offload(struct net_device *dev, u32 data,
return rc1 ? rc1 : rc2;
}
-static int ibmveth_set_rx_csum(struct net_device *dev, u32 data)
+static int ibmveth_set_features(struct net_device *dev, u32 features)
{
struct ibmveth_adapter *adapter = netdev_priv(dev);
+ int rx_csum = !!(features & NETIF_F_RXCSUM);
+ int rc;
- if ((data && adapter->rx_csum) || (!data && !adapter->rx_csum))
+ if (rx_csum == adapter->rx_csum)
return 0;
- return ibmveth_set_csum_offload(dev, data, ibmveth_set_rx_csum_flags);
-}
-
-static int ibmveth_set_tx_csum(struct net_device *dev, u32 data)
-{
- struct ibmveth_adapter *adapter = netdev_priv(dev);
- int rc = 0;
-
- if (data && (dev->features & (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
- return 0;
- if (!data && !(dev->features & (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
- return 0;
-
- if (data && !adapter->rx_csum)
- rc = ibmveth_set_csum_offload(dev, data,
- ibmveth_set_tx_csum_flags);
- else
- ibmveth_set_tx_csum_flags(dev, data);
+ rc = ibmveth_set_csum_offload(dev, rx_csum);
+ if (rc && !adapter->rx_csum)
+ dev->features = features & ~(NETIF_F_ALL_CSUM | NETIF_F_RXCSUM);
return rc;
}
-static u32 ibmveth_get_rx_csum(struct net_device *dev)
-{
- struct ibmveth_adapter *adapter = netdev_priv(dev);
- return adapter->rx_csum;
-}
-
static void ibmveth_get_strings(struct net_device *dev, u32 stringset, u8 *data)
{
int i;
@@ -914,13 +874,9 @@ static const struct ethtool_ops netdev_ethtool_ops = {
.get_drvinfo = netdev_get_drvinfo,
.get_settings = netdev_get_settings,
.get_link = ethtool_op_get_link,
- .set_tx_csum = ibmveth_set_tx_csum,
- .get_rx_csum = ibmveth_get_rx_csum,
- .set_rx_csum = ibmveth_set_rx_csum,
.get_strings = ibmveth_get_strings,
.get_sset_count = ibmveth_get_sset_count,
.get_ethtool_stats = ibmveth_get_ethtool_stats,
- .set_sg = ethtool_op_set_sg,
};
static int ibmveth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
@@ -1345,6 +1301,8 @@ static const struct net_device_ops ibmveth_netdev_ops = {
.ndo_set_multicast_list = ibmveth_set_multicast_list,
.ndo_do_ioctl = ibmveth_ioctl,
.ndo_change_mtu = ibmveth_change_mtu,
+ .ndo_fix_features = ibmveth_fix_features,
+ .ndo_set_features = ibmveth_set_features,
.ndo_validate_addr = eth_validate_addr,
.ndo_set_mac_address = eth_mac_addr,
#ifdef CONFIG_NET_POLL_CONTROLLER
@@ -1412,7 +1370,9 @@ static int __devinit ibmveth_probe(struct vio_dev *dev,
netdev->netdev_ops = &ibmveth_netdev_ops;
netdev->ethtool_ops = &netdev_ethtool_ops;
SET_NETDEV_DEV(netdev, &dev->dev);
- netdev->features |= NETIF_F_SG;
+ netdev->hw_features = NETIF_F_SG | NETIF_F_RXCSUM |
+ NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
+ netdev->features |= netdev->hw_features;
memcpy(netdev->dev_addr, &adapter->mac_addr, netdev->addr_len);
@@ -1437,8 +1397,6 @@ static int __devinit ibmveth_probe(struct vio_dev *dev,
netdev_dbg(netdev, "registering netdev...\n");
- ibmveth_set_csum_offload(netdev, 1, ibmveth_set_tx_csum_flags);
-
rc = register_netdev(netdev);
if (rc) {
--
1.7.2.5
^ permalink raw reply related
* Re: Add NAPI support to ll_temac driver
From: Ben Hutchings @ 2011-04-19 12:25 UTC (permalink / raw)
To: Eric Dumazet; +Cc: monstr, netdev
In-Reply-To: <1303209787.3480.9.camel@edumazet-laptop>
On Tue, 2011-04-19 at 12:43 +0200, Eric Dumazet wrote:
[...]
> One possible way to get better performance is to change driver to
> allocate skbs only right before calling netif_rx(), so that you dont
> have to access cold sk_buff data twice (once when allocating skb and put
> it in ring buffer, a second time when receiving frame)
>
> drivers/net/niu.c is a good example for this (NAPI + netdev_alloc_skb()
> just in time + pull in skbhead only first cache line of packet)
[...]
If the hardware can do RX checksumming (it's not clear) then the driver
should pass the paged buffers into GRO and that will take care of skb
allocation as necessary.
Ben.
--
Ben Hutchings, Senior Software Engineer, Solarflare
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
* Re: Add NAPI support to ll_temac driver
From: Michal Simek @ 2011-04-19 12:26 UTC (permalink / raw)
To: Eric Dumazet; +Cc: netdev
In-Reply-To: <1303209787.3480.9.camel@edumazet-laptop>
Eric Dumazet wrote:
> Le mardi 19 avril 2011 à 11:35 +0200, Michal Simek a écrit :
>> Hi,
>>
>> I would like to try to add NAPI support for ll_temac and look if help us to
>> improve performance on Microblaze system. I would expect that bandwidth should
>> be increased.
>> We have the second non mainline driver which use tasklets and it provides better
>> performance than mainline driver but not so big that's why I think that NAPI
>> can increase performance.
>>
>> Can you please point me to any driver which I could use as a template?
>> Or any developer guide to do so.
>>
>> Do you know any other option how to improve driver performance on low speed cpu?
>>
>> I have found that driver spends a lot of time on skb allocation and preallocated
>> SKBs help a little bit. I have done a test where I increased number of
>> preallocated BDs(SKBs) for rx to 35000 and disable new BD(SKB) allocation in
>> rx_irq. 35000 BDs is setup because I need them to successfully finish netperf
>> test. I have got 25% bandwidth increasing.
>>
>> It will be also nice to be able to allocate several BDs(SKBs) which could be
>> faster than allocate them in sequence.
>
> Depends if your cpu has some cache. The best performance is to try to
> get high cache hit ratios.
Yes it has icache and dcache (write-back or write-through).
>
> One possible way to get better performance is to change driver to
> allocate skbs only right before calling netif_rx(), so that you dont
> have to access cold sk_buff data twice (once when allocating skb and put
> it in ring buffer, a second time when receiving frame)
ok. But I need to allocate BD for dma with pointer to skb where dma should copy
data to. I could do it in irq but I would have to wait till dma copy data from
ethernet controller to memory. I haven't measure how slow/fast is that copying.
>
> drivers/net/niu.c is a good example for this (NAPI + netdev_alloc_skb()
> just in time + pull in skbhead only first cache line of packet)
>
> drivers/net/ftmac100.c is also a recent driver (and probably a better
> start with less complex hardware than NIU) using these tricks
>
> { skb = netdev_alloc_skb_ip_align(netdev, 128);
> __pskb_pull_tail(skb, min(length, 64));
> }
I have change rx for napi but need to debug it a little bit. It works for some
packets but I am not able to run any test right now.
Thanks,
Michal
--
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian
^ permalink raw reply
* [PATCH] net: qlcnic: convert to hw_features
From: Michał Mirosław @ 2011-04-19 12:42 UTC (permalink / raw)
To: netdev; +Cc: Amit Kumar Salecha, Anirban Chakraborty, linux-driver
Bit more than minimal conversion. There might be some issues because
of qlcnic_set_netdev_features() if it's called after netdev init.
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
---
drivers/net/qlcnic/qlcnic.h | 3 +-
drivers/net/qlcnic/qlcnic_ethtool.c | 120 -----------------------------------
drivers/net/qlcnic/qlcnic_hw.c | 35 ++++++++++
drivers/net/qlcnic/qlcnic_init.c | 4 +-
drivers/net/qlcnic/qlcnic_main.c | 35 +++++------
5 files changed, 55 insertions(+), 142 deletions(-)
diff --git a/drivers/net/qlcnic/qlcnic.h b/drivers/net/qlcnic/qlcnic.h
index e5d3053..fa5b15c 100644
--- a/drivers/net/qlcnic/qlcnic.h
+++ b/drivers/net/qlcnic/qlcnic.h
@@ -926,7 +926,6 @@ struct qlcnic_adapter {
u8 max_rds_rings;
u8 max_sds_rings;
u8 msix_supported;
- u8 rx_csum;
u8 portnum;
u8 physical_port;
u8 reset_context;
@@ -1247,6 +1246,8 @@ void qlcnic_advert_link_change(struct qlcnic_adapter *adapter, int linkup);
int qlcnic_fw_cmd_set_mtu(struct qlcnic_adapter *adapter, int mtu);
int qlcnic_change_mtu(struct net_device *netdev, int new_mtu);
+u32 qlcnic_fix_features(struct net_device *netdev, u32 features);
+int qlcnic_set_features(struct net_device *netdev, u32 features);
int qlcnic_config_hw_lro(struct qlcnic_adapter *adapter, int enable);
int qlcnic_config_bridged_mode(struct qlcnic_adapter *adapter, u32 enable);
int qlcnic_send_lro_cleanup(struct qlcnic_adapter *adapter);
diff --git a/drivers/net/qlcnic/qlcnic_ethtool.c b/drivers/net/qlcnic/qlcnic_ethtool.c
index 3cd8a16..615a5ab 100644
--- a/drivers/net/qlcnic/qlcnic_ethtool.c
+++ b/drivers/net/qlcnic/qlcnic_ethtool.c
@@ -764,73 +764,6 @@ qlcnic_get_ethtool_stats(struct net_device *dev,
qlcnic_fill_device_stats(&index, data, &port_stats.tx);
}
-static int qlcnic_set_tx_csum(struct net_device *dev, u32 data)
-{
- struct qlcnic_adapter *adapter = netdev_priv(dev);
-
- if ((adapter->flags & QLCNIC_ESWITCH_ENABLED))
- return -EOPNOTSUPP;
- if (data)
- dev->features |= (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM);
- else
- dev->features &= ~(NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM);
-
- return 0;
-
-}
-static u32 qlcnic_get_tx_csum(struct net_device *dev)
-{
- return dev->features & NETIF_F_IP_CSUM;
-}
-
-static u32 qlcnic_get_rx_csum(struct net_device *dev)
-{
- struct qlcnic_adapter *adapter = netdev_priv(dev);
- return adapter->rx_csum;
-}
-
-static int qlcnic_set_rx_csum(struct net_device *dev, u32 data)
-{
- struct qlcnic_adapter *adapter = netdev_priv(dev);
-
- if ((adapter->flags & QLCNIC_ESWITCH_ENABLED))
- return -EOPNOTSUPP;
- if (!!data) {
- adapter->rx_csum = !!data;
- return 0;
- }
-
- if (dev->features & NETIF_F_LRO) {
- if (qlcnic_config_hw_lro(adapter, QLCNIC_LRO_DISABLED))
- return -EIO;
-
- dev->features &= ~NETIF_F_LRO;
- qlcnic_send_lro_cleanup(adapter);
- dev_info(&adapter->pdev->dev,
- "disabling LRO as rx_csum is off\n");
- }
- adapter->rx_csum = !!data;
- return 0;
-}
-
-static u32 qlcnic_get_tso(struct net_device *dev)
-{
- return (dev->features & (NETIF_F_TSO | NETIF_F_TSO6)) != 0;
-}
-
-static int qlcnic_set_tso(struct net_device *dev, u32 data)
-{
- struct qlcnic_adapter *adapter = netdev_priv(dev);
- if (!(adapter->capabilities & QLCNIC_FW_CAPABILITY_TSO))
- return -EOPNOTSUPP;
- if (data)
- dev->features |= (NETIF_F_TSO | NETIF_F_TSO6);
- else
- dev->features &= ~(NETIF_F_TSO | NETIF_F_TSO6);
-
- return 0;
-}
-
static int qlcnic_set_led(struct net_device *dev,
enum ethtool_phys_id_state state)
{
@@ -993,50 +926,6 @@ static int qlcnic_get_intr_coalesce(struct net_device *netdev,
return 0;
}
-static int qlcnic_set_flags(struct net_device *netdev, u32 data)
-{
- struct qlcnic_adapter *adapter = netdev_priv(netdev);
- int hw_lro;
-
- if (ethtool_invalid_flags(netdev, data, ETH_FLAG_LRO))
- return -EINVAL;
-
- if (data & ETH_FLAG_LRO) {
-
- if (netdev->features & NETIF_F_LRO)
- return 0;
-
- if (!(adapter->capabilities & QLCNIC_FW_CAPABILITY_HW_LRO))
- return -EINVAL;
-
- if (!adapter->rx_csum) {
- dev_info(&adapter->pdev->dev, "rx csum is off, "
- "cannot toggle lro\n");
- return -EINVAL;
- }
-
- hw_lro = QLCNIC_LRO_ENABLED;
- netdev->features |= NETIF_F_LRO;
-
- } else {
-
- if (!(netdev->features & NETIF_F_LRO))
- return 0;
-
- hw_lro = 0;
- netdev->features &= ~NETIF_F_LRO;
- }
-
- if (qlcnic_config_hw_lro(adapter, hw_lro))
- return -EIO;
-
- if ((hw_lro == 0) && qlcnic_send_lro_cleanup(adapter))
- return -EIO;
-
-
- return 0;
-}
-
static u32 qlcnic_get_msglevel(struct net_device *netdev)
{
struct qlcnic_adapter *adapter = netdev_priv(netdev);
@@ -1064,23 +953,14 @@ const struct ethtool_ops qlcnic_ethtool_ops = {
.set_ringparam = qlcnic_set_ringparam,
.get_pauseparam = qlcnic_get_pauseparam,
.set_pauseparam = qlcnic_set_pauseparam,
- .get_tx_csum = qlcnic_get_tx_csum,
- .set_tx_csum = qlcnic_set_tx_csum,
- .set_sg = ethtool_op_set_sg,
- .get_tso = qlcnic_get_tso,
- .set_tso = qlcnic_set_tso,
.get_wol = qlcnic_get_wol,
.set_wol = qlcnic_set_wol,
.self_test = qlcnic_diag_test,
.get_strings = qlcnic_get_strings,
.get_ethtool_stats = qlcnic_get_ethtool_stats,
.get_sset_count = qlcnic_get_sset_count,
- .get_rx_csum = qlcnic_get_rx_csum,
- .set_rx_csum = qlcnic_set_rx_csum,
.get_coalesce = qlcnic_get_intr_coalesce,
.set_coalesce = qlcnic_set_intr_coalesce,
- .get_flags = ethtool_op_get_flags,
- .set_flags = qlcnic_set_flags,
.set_phys_id = qlcnic_set_led,
.set_msglevel = qlcnic_set_msglevel,
.get_msglevel = qlcnic_get_msglevel,
diff --git a/drivers/net/qlcnic/qlcnic_hw.c b/drivers/net/qlcnic/qlcnic_hw.c
index 498cca9..82613c8 100644
--- a/drivers/net/qlcnic/qlcnic_hw.c
+++ b/drivers/net/qlcnic/qlcnic_hw.c
@@ -758,6 +758,41 @@ int qlcnic_change_mtu(struct net_device *netdev, int mtu)
return rc;
}
+
+u32 qlcnic_fix_features(struct net_device *netdev, u32 features)
+{
+ if ((adapter->flags & QLCNIC_ESWITCH_ENABLED)) {
+ u32 changed = features ^ dev->features;
+ features ^= changed & (NETIF_F_ALL_CSUM | NETIF_F_RXCSUM);
+ }
+
+ if (!(features & NETIF_F_RXCSUM))
+ features &= ~NETIF_F_LRO;
+
+ return features;
+}
+
+
+int qlcnic_set_features(struct net_device *netdev, u32 features)
+{
+ struct qlcnic_adapter *adapter = netdev_priv(netdev);
+ u32 changed = dev->features ^ features;
+ int hw_lro = (features & NETIF_F_LRO) ? QLCNIC_LRO_ENABLED : 0;
+
+ if (!(changed & NETIF_F_LRO))
+ return 0;
+
+ dev->features = features;
+
+ if (qlcnic_config_hw_lro(adapter, hw_lro))
+ return -EIO;
+
+ if ((hw_lro == 0) && qlcnic_send_lro_cleanup(adapter))
+ return -EIO;
+
+ return 1;
+}
+
/*
* Changes the CRB window to the specified window.
*/
diff --git a/drivers/net/qlcnic/qlcnic_init.c b/drivers/net/qlcnic/qlcnic_init.c
index 4ec0eeb..d0f338b 100644
--- a/drivers/net/qlcnic/qlcnic_init.c
+++ b/drivers/net/qlcnic/qlcnic_init.c
@@ -1390,8 +1390,8 @@ static struct sk_buff *qlcnic_process_rxbuf(struct qlcnic_adapter *adapter,
skb = buffer->skb;
- if (likely(adapter->rx_csum && (cksum == STATUS_CKSUM_OK ||
- cksum == STATUS_CKSUM_LOOP))) {
+ if (likely((adapter->netdev->features & NETIF_F_RXCSUM) &&
+ (cksum == STATUS_CKSUM_OK || cksum == STATUS_CKSUM_LOOP))) {
adapter->stats.csummed++;
skb->ip_summed = CHECKSUM_UNNECESSARY;
} else {
diff --git a/drivers/net/qlcnic/qlcnic_main.c b/drivers/net/qlcnic/qlcnic_main.c
index e9e9ba6..06bb233 100644
--- a/drivers/net/qlcnic/qlcnic_main.c
+++ b/drivers/net/qlcnic/qlcnic_main.c
@@ -328,6 +328,8 @@ static const struct net_device_ops qlcnic_netdev_ops = {
.ndo_set_multicast_list = qlcnic_set_multi,
.ndo_set_mac_address = qlcnic_set_mac,
.ndo_change_mtu = qlcnic_change_mtu,
+ .ndo_fix_features = qlcnic_fix_features,
+ .ndo_set_features = qlcnic_set_features,
.ndo_tx_timeout = qlcnic_tx_timeout,
.ndo_vlan_rx_add_vid = qlcnic_vlan_rx_add,
.ndo_vlan_rx_kill_vid = qlcnic_vlan_rx_del,
@@ -764,7 +766,7 @@ qlcnic_set_netdev_features(struct qlcnic_adapter *adapter,
struct net_device *netdev = adapter->netdev;
unsigned long features, vlan_features;
- features = (NETIF_F_SG | NETIF_F_IP_CSUM |
+ features = (NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_RXCSUM |
NETIF_F_IPV6_CSUM | NETIF_F_GRO);
vlan_features = (NETIF_F_SG | NETIF_F_IP_CSUM |
NETIF_F_IPV6_CSUM | NETIF_F_HW_VLAN_FILTER);
@@ -779,14 +781,12 @@ qlcnic_set_netdev_features(struct qlcnic_adapter *adapter,
if (esw_cfg->offload_flags & BIT_0) {
netdev->features |= features;
- adapter->rx_csum = 1;
if (!(esw_cfg->offload_flags & BIT_1))
netdev->features &= ~NETIF_F_TSO;
if (!(esw_cfg->offload_flags & BIT_2))
netdev->features &= ~NETIF_F_TSO6;
} else {
netdev->features &= ~features;
- adapter->rx_csum = 0;
}
netdev->vlan_features = (features & vlan_features);
@@ -1436,7 +1436,6 @@ qlcnic_setup_netdev(struct qlcnic_adapter *adapter,
int err;
struct pci_dev *pdev = adapter->pdev;
- adapter->rx_csum = 1;
adapter->mc_enabled = 0;
adapter->max_mc_count = 38;
@@ -1447,26 +1446,24 @@ qlcnic_setup_netdev(struct qlcnic_adapter *adapter,
SET_ETHTOOL_OPS(netdev, &qlcnic_ethtool_ops);
- netdev->features |= (NETIF_F_SG | NETIF_F_IP_CSUM |
- NETIF_F_IPV6_CSUM | NETIF_F_GRO | NETIF_F_HW_VLAN_RX);
- netdev->vlan_features |= (NETIF_F_SG | NETIF_F_IP_CSUM |
- NETIF_F_IPV6_CSUM | NETIF_F_HW_VLAN_FILTER);
-
- if (adapter->capabilities & QLCNIC_FW_CAPABILITY_TSO) {
- netdev->features |= (NETIF_F_TSO | NETIF_F_TSO6);
- netdev->vlan_features |= (NETIF_F_TSO | NETIF_F_TSO6);
- }
+ netdev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM |
+ NETIF_F_IPV6_CSUM | NETIF_F_RXCSUM;
+ if (adapter->capabilities & QLCNIC_FW_CAPABILITY_TSO)
+ netdev->hw_features |= NETIF_F_TSO | NETIF_F_TSO6;
if (pci_using_dac) {
- netdev->features |= NETIF_F_HIGHDMA;
- netdev->vlan_features |= NETIF_F_HIGHDMA;
- }
+ netdev->hw_features |= NETIF_F_HIGHDMA;
+
+ netdev->vlan_features = netdev->hw_features;
if (adapter->capabilities & QLCNIC_FW_CAPABILITY_FVLANTX)
- netdev->features |= (NETIF_F_HW_VLAN_TX);
-
+ netdev->hw_features |= NETIF_F_HW_VLAN_TX;
if (adapter->capabilities & QLCNIC_FW_CAPABILITY_HW_LRO)
- netdev->features |= NETIF_F_LRO;
+ netdev->hw_features |= NETIF_F_LRO;
+
+ netdev->features |= netdev->hw_features |
+ NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_FILTER;
+
netdev->irq = adapter->msix_entries[0].vector;
netif_carrier_off(netdev);
--
1.7.2.5
^ permalink raw reply related
* Re: Add NAPI support to ll_temac driver
From: Michal Simek @ 2011-04-19 12:48 UTC (permalink / raw)
To: Ben Hutchings; +Cc: Eric Dumazet, netdev
In-Reply-To: <1303215925.3464.54.camel@localhost>
Ben Hutchings wrote:
> On Tue, 2011-04-19 at 12:43 +0200, Eric Dumazet wrote:
> [...]
>> One possible way to get better performance is to change driver to
>> allocate skbs only right before calling netif_rx(), so that you dont
>> have to access cold sk_buff data twice (once when allocating skb and put
>> it in ring buffer, a second time when receiving frame)
>>
>> drivers/net/niu.c is a good example for this (NAPI + netdev_alloc_skb()
>> just in time + pull in skbhead only first cache line of packet)
> [...]
>
> If the hardware can do RX checksumming (it's not clear) then the driver
> should pass the paged buffers into GRO and that will take care of skb
> allocation as necessary.
Hardware supports RX and TX partial checksumming. I can enable it. The driver
has also this option and from my tests there is of course some performance
improvemetn.
Just for sure - here are links on documentation.
http://www.xilinx.com/support/documentation/ip_documentation/xps_ll_temac.pdf
or
http://www.xilinx.com/support/documentation/ip_documentation/axi_ethernet/v2_01_a/ds759_axi_ethernet.pdf
About SKB allocation. I fixed our non mainline driver to allocate skb based on
current mtu size. Mainline driver allocate max mtu (9k). This has also impact on
performance because Microblaze works with smaller SKBs.
Can you please be more specific about passing the paged buffers into GRO?
Or point me to any documentation or code which can help me to understand what
that means.
Thanks,
Michal
--
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian
^ permalink raw reply
* [PATCH v2] net: qlcnic: convert to hw_features
From: Michał Mirosław @ 2011-04-19 13:03 UTC (permalink / raw)
To: netdev; +Cc: Amit Kumar Salecha, Anirban Chakraborty, linux-driver
In-Reply-To: <20110419124238.AECB513909@rere.qmqm.pl>
Bit more than minimal conversion. There might be some issues because
of qlcnic_set_netdev_features() if it's called after netdev init.
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
---
v2: fix compilation issues (I forgot to 'stg refresh' before sending v1)
drivers/net/qlcnic/qlcnic.h | 3 +-
drivers/net/qlcnic/qlcnic_ethtool.c | 120 -----------------------------------
drivers/net/qlcnic/qlcnic_hw.c | 37 +++++++++++
drivers/net/qlcnic/qlcnic_init.c | 4 +-
drivers/net/qlcnic/qlcnic_main.c | 35 +++++------
5 files changed, 57 insertions(+), 142 deletions(-)
diff --git a/drivers/net/qlcnic/qlcnic.h b/drivers/net/qlcnic/qlcnic.h
index e5d3053..fa5b15c 100644
--- a/drivers/net/qlcnic/qlcnic.h
+++ b/drivers/net/qlcnic/qlcnic.h
@@ -926,7 +926,6 @@ struct qlcnic_adapter {
u8 max_rds_rings;
u8 max_sds_rings;
u8 msix_supported;
- u8 rx_csum;
u8 portnum;
u8 physical_port;
u8 reset_context;
@@ -1247,6 +1246,8 @@ void qlcnic_advert_link_change(struct qlcnic_adapter *adapter, int linkup);
int qlcnic_fw_cmd_set_mtu(struct qlcnic_adapter *adapter, int mtu);
int qlcnic_change_mtu(struct net_device *netdev, int new_mtu);
+u32 qlcnic_fix_features(struct net_device *netdev, u32 features);
+int qlcnic_set_features(struct net_device *netdev, u32 features);
int qlcnic_config_hw_lro(struct qlcnic_adapter *adapter, int enable);
int qlcnic_config_bridged_mode(struct qlcnic_adapter *adapter, u32 enable);
int qlcnic_send_lro_cleanup(struct qlcnic_adapter *adapter);
diff --git a/drivers/net/qlcnic/qlcnic_ethtool.c b/drivers/net/qlcnic/qlcnic_ethtool.c
index 3cd8a16..615a5ab 100644
--- a/drivers/net/qlcnic/qlcnic_ethtool.c
+++ b/drivers/net/qlcnic/qlcnic_ethtool.c
@@ -764,73 +764,6 @@ qlcnic_get_ethtool_stats(struct net_device *dev,
qlcnic_fill_device_stats(&index, data, &port_stats.tx);
}
-static int qlcnic_set_tx_csum(struct net_device *dev, u32 data)
-{
- struct qlcnic_adapter *adapter = netdev_priv(dev);
-
- if ((adapter->flags & QLCNIC_ESWITCH_ENABLED))
- return -EOPNOTSUPP;
- if (data)
- dev->features |= (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM);
- else
- dev->features &= ~(NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM);
-
- return 0;
-
-}
-static u32 qlcnic_get_tx_csum(struct net_device *dev)
-{
- return dev->features & NETIF_F_IP_CSUM;
-}
-
-static u32 qlcnic_get_rx_csum(struct net_device *dev)
-{
- struct qlcnic_adapter *adapter = netdev_priv(dev);
- return adapter->rx_csum;
-}
-
-static int qlcnic_set_rx_csum(struct net_device *dev, u32 data)
-{
- struct qlcnic_adapter *adapter = netdev_priv(dev);
-
- if ((adapter->flags & QLCNIC_ESWITCH_ENABLED))
- return -EOPNOTSUPP;
- if (!!data) {
- adapter->rx_csum = !!data;
- return 0;
- }
-
- if (dev->features & NETIF_F_LRO) {
- if (qlcnic_config_hw_lro(adapter, QLCNIC_LRO_DISABLED))
- return -EIO;
-
- dev->features &= ~NETIF_F_LRO;
- qlcnic_send_lro_cleanup(adapter);
- dev_info(&adapter->pdev->dev,
- "disabling LRO as rx_csum is off\n");
- }
- adapter->rx_csum = !!data;
- return 0;
-}
-
-static u32 qlcnic_get_tso(struct net_device *dev)
-{
- return (dev->features & (NETIF_F_TSO | NETIF_F_TSO6)) != 0;
-}
-
-static int qlcnic_set_tso(struct net_device *dev, u32 data)
-{
- struct qlcnic_adapter *adapter = netdev_priv(dev);
- if (!(adapter->capabilities & QLCNIC_FW_CAPABILITY_TSO))
- return -EOPNOTSUPP;
- if (data)
- dev->features |= (NETIF_F_TSO | NETIF_F_TSO6);
- else
- dev->features &= ~(NETIF_F_TSO | NETIF_F_TSO6);
-
- return 0;
-}
-
static int qlcnic_set_led(struct net_device *dev,
enum ethtool_phys_id_state state)
{
@@ -993,50 +926,6 @@ static int qlcnic_get_intr_coalesce(struct net_device *netdev,
return 0;
}
-static int qlcnic_set_flags(struct net_device *netdev, u32 data)
-{
- struct qlcnic_adapter *adapter = netdev_priv(netdev);
- int hw_lro;
-
- if (ethtool_invalid_flags(netdev, data, ETH_FLAG_LRO))
- return -EINVAL;
-
- if (data & ETH_FLAG_LRO) {
-
- if (netdev->features & NETIF_F_LRO)
- return 0;
-
- if (!(adapter->capabilities & QLCNIC_FW_CAPABILITY_HW_LRO))
- return -EINVAL;
-
- if (!adapter->rx_csum) {
- dev_info(&adapter->pdev->dev, "rx csum is off, "
- "cannot toggle lro\n");
- return -EINVAL;
- }
-
- hw_lro = QLCNIC_LRO_ENABLED;
- netdev->features |= NETIF_F_LRO;
-
- } else {
-
- if (!(netdev->features & NETIF_F_LRO))
- return 0;
-
- hw_lro = 0;
- netdev->features &= ~NETIF_F_LRO;
- }
-
- if (qlcnic_config_hw_lro(adapter, hw_lro))
- return -EIO;
-
- if ((hw_lro == 0) && qlcnic_send_lro_cleanup(adapter))
- return -EIO;
-
-
- return 0;
-}
-
static u32 qlcnic_get_msglevel(struct net_device *netdev)
{
struct qlcnic_adapter *adapter = netdev_priv(netdev);
@@ -1064,23 +953,14 @@ const struct ethtool_ops qlcnic_ethtool_ops = {
.set_ringparam = qlcnic_set_ringparam,
.get_pauseparam = qlcnic_get_pauseparam,
.set_pauseparam = qlcnic_set_pauseparam,
- .get_tx_csum = qlcnic_get_tx_csum,
- .set_tx_csum = qlcnic_set_tx_csum,
- .set_sg = ethtool_op_set_sg,
- .get_tso = qlcnic_get_tso,
- .set_tso = qlcnic_set_tso,
.get_wol = qlcnic_get_wol,
.set_wol = qlcnic_set_wol,
.self_test = qlcnic_diag_test,
.get_strings = qlcnic_get_strings,
.get_ethtool_stats = qlcnic_get_ethtool_stats,
.get_sset_count = qlcnic_get_sset_count,
- .get_rx_csum = qlcnic_get_rx_csum,
- .set_rx_csum = qlcnic_set_rx_csum,
.get_coalesce = qlcnic_get_intr_coalesce,
.set_coalesce = qlcnic_set_intr_coalesce,
- .get_flags = ethtool_op_get_flags,
- .set_flags = qlcnic_set_flags,
.set_phys_id = qlcnic_set_led,
.set_msglevel = qlcnic_set_msglevel,
.get_msglevel = qlcnic_get_msglevel,
diff --git a/drivers/net/qlcnic/qlcnic_hw.c b/drivers/net/qlcnic/qlcnic_hw.c
index 498cca9..cbb27f2 100644
--- a/drivers/net/qlcnic/qlcnic_hw.c
+++ b/drivers/net/qlcnic/qlcnic_hw.c
@@ -758,6 +758,43 @@ int qlcnic_change_mtu(struct net_device *netdev, int mtu)
return rc;
}
+
+u32 qlcnic_fix_features(struct net_device *netdev, u32 features)
+{
+ struct qlcnic_adapter *adapter = netdev_priv(netdev);
+
+ if ((adapter->flags & QLCNIC_ESWITCH_ENABLED)) {
+ u32 changed = features ^ netdev->features;
+ features ^= changed & (NETIF_F_ALL_CSUM | NETIF_F_RXCSUM);
+ }
+
+ if (!(features & NETIF_F_RXCSUM))
+ features &= ~NETIF_F_LRO;
+
+ return features;
+}
+
+
+int qlcnic_set_features(struct net_device *netdev, u32 features)
+{
+ struct qlcnic_adapter *adapter = netdev_priv(netdev);
+ u32 changed = netdev->features ^ features;
+ int hw_lro = (features & NETIF_F_LRO) ? QLCNIC_LRO_ENABLED : 0;
+
+ if (!(changed & NETIF_F_LRO))
+ return 0;
+
+ netdev->features = features ^ NETIF_F_LRO;
+
+ if (qlcnic_config_hw_lro(adapter, hw_lro))
+ return -EIO;
+
+ if ((hw_lro == 0) && qlcnic_send_lro_cleanup(adapter))
+ return -EIO;
+
+ return 0;
+}
+
/*
* Changes the CRB window to the specified window.
*/
diff --git a/drivers/net/qlcnic/qlcnic_init.c b/drivers/net/qlcnic/qlcnic_init.c
index 4ec0eeb..d0f338b 100644
--- a/drivers/net/qlcnic/qlcnic_init.c
+++ b/drivers/net/qlcnic/qlcnic_init.c
@@ -1390,8 +1390,8 @@ static struct sk_buff *qlcnic_process_rxbuf(struct qlcnic_adapter *adapter,
skb = buffer->skb;
- if (likely(adapter->rx_csum && (cksum == STATUS_CKSUM_OK ||
- cksum == STATUS_CKSUM_LOOP))) {
+ if (likely((adapter->netdev->features & NETIF_F_RXCSUM) &&
+ (cksum == STATUS_CKSUM_OK || cksum == STATUS_CKSUM_LOOP))) {
adapter->stats.csummed++;
skb->ip_summed = CHECKSUM_UNNECESSARY;
} else {
diff --git a/drivers/net/qlcnic/qlcnic_main.c b/drivers/net/qlcnic/qlcnic_main.c
index e9e9ba6..6e61951 100644
--- a/drivers/net/qlcnic/qlcnic_main.c
+++ b/drivers/net/qlcnic/qlcnic_main.c
@@ -328,6 +328,8 @@ static const struct net_device_ops qlcnic_netdev_ops = {
.ndo_set_multicast_list = qlcnic_set_multi,
.ndo_set_mac_address = qlcnic_set_mac,
.ndo_change_mtu = qlcnic_change_mtu,
+ .ndo_fix_features = qlcnic_fix_features,
+ .ndo_set_features = qlcnic_set_features,
.ndo_tx_timeout = qlcnic_tx_timeout,
.ndo_vlan_rx_add_vid = qlcnic_vlan_rx_add,
.ndo_vlan_rx_kill_vid = qlcnic_vlan_rx_del,
@@ -764,7 +766,7 @@ qlcnic_set_netdev_features(struct qlcnic_adapter *adapter,
struct net_device *netdev = adapter->netdev;
unsigned long features, vlan_features;
- features = (NETIF_F_SG | NETIF_F_IP_CSUM |
+ features = (NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_RXCSUM |
NETIF_F_IPV6_CSUM | NETIF_F_GRO);
vlan_features = (NETIF_F_SG | NETIF_F_IP_CSUM |
NETIF_F_IPV6_CSUM | NETIF_F_HW_VLAN_FILTER);
@@ -779,14 +781,12 @@ qlcnic_set_netdev_features(struct qlcnic_adapter *adapter,
if (esw_cfg->offload_flags & BIT_0) {
netdev->features |= features;
- adapter->rx_csum = 1;
if (!(esw_cfg->offload_flags & BIT_1))
netdev->features &= ~NETIF_F_TSO;
if (!(esw_cfg->offload_flags & BIT_2))
netdev->features &= ~NETIF_F_TSO6;
} else {
netdev->features &= ~features;
- adapter->rx_csum = 0;
}
netdev->vlan_features = (features & vlan_features);
@@ -1436,7 +1436,6 @@ qlcnic_setup_netdev(struct qlcnic_adapter *adapter,
int err;
struct pci_dev *pdev = adapter->pdev;
- adapter->rx_csum = 1;
adapter->mc_enabled = 0;
adapter->max_mc_count = 38;
@@ -1447,26 +1446,24 @@ qlcnic_setup_netdev(struct qlcnic_adapter *adapter,
SET_ETHTOOL_OPS(netdev, &qlcnic_ethtool_ops);
- netdev->features |= (NETIF_F_SG | NETIF_F_IP_CSUM |
- NETIF_F_IPV6_CSUM | NETIF_F_GRO | NETIF_F_HW_VLAN_RX);
- netdev->vlan_features |= (NETIF_F_SG | NETIF_F_IP_CSUM |
- NETIF_F_IPV6_CSUM | NETIF_F_HW_VLAN_FILTER);
+ netdev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM |
+ NETIF_F_IPV6_CSUM | NETIF_F_RXCSUM;
- if (adapter->capabilities & QLCNIC_FW_CAPABILITY_TSO) {
- netdev->features |= (NETIF_F_TSO | NETIF_F_TSO6);
- netdev->vlan_features |= (NETIF_F_TSO | NETIF_F_TSO6);
- }
+ if (adapter->capabilities & QLCNIC_FW_CAPABILITY_TSO)
+ netdev->hw_features |= NETIF_F_TSO | NETIF_F_TSO6;
+ if (pci_using_dac)
+ netdev->hw_features |= NETIF_F_HIGHDMA;
- if (pci_using_dac) {
- netdev->features |= NETIF_F_HIGHDMA;
- netdev->vlan_features |= NETIF_F_HIGHDMA;
- }
+ netdev->vlan_features = netdev->hw_features;
if (adapter->capabilities & QLCNIC_FW_CAPABILITY_FVLANTX)
- netdev->features |= (NETIF_F_HW_VLAN_TX);
-
+ netdev->hw_features |= NETIF_F_HW_VLAN_TX;
if (adapter->capabilities & QLCNIC_FW_CAPABILITY_HW_LRO)
- netdev->features |= NETIF_F_LRO;
+ netdev->hw_features |= NETIF_F_LRO;
+
+ netdev->features |= netdev->hw_features |
+ NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_FILTER;
+
netdev->irq = adapter->msix_entries[0].vector;
netif_carrier_off(netdev);
--
1.7.2.5
^ permalink raw reply related
* Re: Add NAPI support to ll_temac driver
From: Eric Dumazet @ 2011-04-19 13:13 UTC (permalink / raw)
To: monstr; +Cc: Ben Hutchings, netdev
In-Reply-To: <4DAD84B1.9020405@monstr.eu>
Le mardi 19 avril 2011 à 14:48 +0200, Michal Simek a écrit :
> Can you please be more specific about passing the paged buffers into GRO?
> Or point me to any documentation or code which can help me to understand what
> that means.
Search for napi_get_frags() :
drivers/net/mlx4/en_rx.c:597: struct sk_buff *gro_skb = napi_get_frags(&cq->napi);
drivers/net/cxgb3/sge.c:2091: skb = napi_get_frags(&qs->napi);
drivers/net/cxgb4/sge.c:1517: skb = napi_get_frags(&rxq->rspq.napi);
drivers/net/qlge/qlge_main.c:1484: skb = napi_get_frags(napi);
drivers/net/sfc/rx.c:471: skb = napi_get_frags(napi);
drivers/net/benet/be_main.c:1039: skb = napi_get_frags(&eq_obj->napi);
drivers/net/cxgb4vf/sge.c:1479: skb = napi_get_frags(&rxq->rspq.napi);
^ permalink raw reply
* Re: Add NAPI support to ll_temac driver
From: Ben Hutchings @ 2011-04-19 13:14 UTC (permalink / raw)
To: monstr; +Cc: Eric Dumazet, netdev
In-Reply-To: <4DAD84B1.9020405@monstr.eu>
On Tue, 2011-04-19 at 14:48 +0200, Michal Simek wrote:
> Ben Hutchings wrote:
> > On Tue, 2011-04-19 at 12:43 +0200, Eric Dumazet wrote:
> > [...]
> >> One possible way to get better performance is to change driver to
> >> allocate skbs only right before calling netif_rx(), so that you dont
> >> have to access cold sk_buff data twice (once when allocating skb and put
> >> it in ring buffer, a second time when receiving frame)
> >>
> >> drivers/net/niu.c is a good example for this (NAPI + netdev_alloc_skb()
> >> just in time + pull in skbhead only first cache line of packet)
> > [...]
> >
> > If the hardware can do RX checksumming (it's not clear) then the driver
> > should pass the paged buffers into GRO and that will take care of skb
> > allocation as necessary.
>
> Hardware supports RX and TX partial checksumming. I can enable it. The driver
> has also this option and from my tests there is of course some performance
> improvemetn.
>
> Just for sure - here are links on documentation.
> http://www.xilinx.com/support/documentation/ip_documentation/xps_ll_temac.pdf
> or
> http://www.xilinx.com/support/documentation/ip_documentation/axi_ethernet/v2_01_a/ds759_axi_ethernet.pdf
I'm not going to read those. Just providing brief advice.
> About SKB allocation. I fixed our non mainline driver to allocate skb based on
> current mtu size. Mainline driver allocate max mtu (9k). This has also impact on
> performance because Microblaze works with smaller SKBs.
>
> Can you please be more specific about passing the paged buffers into GRO?
> Or point me to any documentation or code which can help me to understand what
> that means.
You would use napi_get_frags() to get a new or recycled skb, fill in
skb->frags, then call napi_gro_frags() to pass it into GRO. The benet,
cxgb3 and sfc drivers do this.
Ben.
--
Ben Hutchings, Senior Software Engineer, Solarflare
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
* Re: [PATCH] net: xen-netback: convert to hw_features
From: Ian Campbell @ 2011-04-19 13:17 UTC (permalink / raw)
To: Michał Mirosław
Cc: netdev@vger.kernel.org, xen-devel@lists.xensource.com
In-Reply-To: <20110419115612.C4ACA13909@rere.qmqm.pl>
On Tue, 2011-04-19 at 12:56 +0100, Michał Mirosław wrote:
> Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Thanks for beating me to this! However the prototype for
xenvif_fix_features is wrong (needs to take a net_device not a xenvif).
I fixed it with the following, I also moved the !can_sg MTU clamping
into a set_features hook (like we do with netfront). Am I right that
this pattern copes with changes to SG via ethtool etc better? I think
it's more future proof in any case.
NB: I'm having some issues with my test hardware at the moment so this
is reviewed by eye and compile tested only...
I'm also happy for this to be folded into the original with my
"Signed-off-/Acked-by Ian Campbell <ian.campbell@citrix.com>" if that is
preferable.
Cheers,
Ian.
8<-----------------
net: xen-netback: correct prototype of xenvif_fix_features.
Also check MTU vs NETIF_F_SG in ndo_set_features hook to allow for
dynamic modification.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
diff --git a/drivers/net/xen-netback/interface.c b/drivers/net/xen-netback/interface.c
index fe25308..61757bd 100644
--- a/drivers/net/xen-netback/interface.c
+++ b/drivers/net/xen-netback/interface.c
@@ -165,9 +165,9 @@ static int xenvif_change_mtu(struct net_device *dev, int mtu)
return 0;
}
-static u32 xenvif_fix_features(struct xenvif *vif, u32 features)
+static u32 xenvif_fix_features(struct net_device *dev, u32 features)
{
- struct net_device *dev = vif->dev;
+ struct xenvif *vif = netdev_priv(dev);
if (!vif->can_sg)
features &= ~NETIF_F_SG;
@@ -179,6 +179,16 @@ static u32 xenvif_fix_features(struct xenvif *vif, u32 features)
return features;
}
+static int xenvif_set_features(struct net_device *dev, u32 features)
+{
+ if (!(features & NETIF_F_SG) && dev->mtu > ETH_DATA_LEN) {
+ netdev_info(dev, "Reducing MTU because no SG offload");
+ dev->mtu = ETH_DATA_LEN;
+ }
+
+ return 0;
+}
+
static const struct xenvif_stat {
char name[ETH_GSTRING_LEN];
u16 offset;
@@ -237,6 +247,7 @@ static struct net_device_ops xenvif_netdev_ops = {
.ndo_stop = xenvif_close,
.ndo_change_mtu = xenvif_change_mtu,
.ndo_fix_features = xenvif_fix_features,
+ .ndo_set_features = xenvif_set_features,
};
struct xenvif *xenvif_alloc(struct device *parent, domid_t domid,
@@ -329,8 +340,6 @@ int xenvif_connect(struct xenvif *vif, unsigned long tx_ring_ref,
rtnl_lock();
if (netif_running(vif->dev))
xenvif_up(vif);
- if (!vif->can_sg && vif->dev->mtu > ETH_DATA_LEN)
- dev_set_mtu(vif->dev, ETH_DATA_LEN);
netdev_update_features(vif->dev);
netif_carrier_on(vif->dev);
rtnl_unlock();
^ permalink raw reply related
* Re: Add NAPI support to ll_temac driver
From: Michal Simek @ 2011-04-19 13:18 UTC (permalink / raw)
To: Ben Hutchings; +Cc: Eric Dumazet, netdev
In-Reply-To: <1303218898.3464.59.camel@localhost>
Ben Hutchings wrote:
> On Tue, 2011-04-19 at 14:48 +0200, Michal Simek wrote:
>> Ben Hutchings wrote:
>>> On Tue, 2011-04-19 at 12:43 +0200, Eric Dumazet wrote:
>>> [...]
>>>> One possible way to get better performance is to change driver to
>>>> allocate skbs only right before calling netif_rx(), so that you dont
>>>> have to access cold sk_buff data twice (once when allocating skb and put
>>>> it in ring buffer, a second time when receiving frame)
>>>>
>>>> drivers/net/niu.c is a good example for this (NAPI + netdev_alloc_skb()
>>>> just in time + pull in skbhead only first cache line of packet)
>>> [...]
>>>
>>> If the hardware can do RX checksumming (it's not clear) then the driver
>>> should pass the paged buffers into GRO and that will take care of skb
>>> allocation as necessary.
>> Hardware supports RX and TX partial checksumming. I can enable it. The driver
>> has also this option and from my tests there is of course some performance
>> improvemetn.
>>
>> Just for sure - here are links on documentation.
>> http://www.xilinx.com/support/documentation/ip_documentation/xps_ll_temac.pdf
>> or
>> http://www.xilinx.com/support/documentation/ip_documentation/axi_ethernet/v2_01_a/ds759_axi_ethernet.pdf
>
> I'm not going to read those. Just providing brief advice.
>
>> About SKB allocation. I fixed our non mainline driver to allocate skb based on
>> current mtu size. Mainline driver allocate max mtu (9k). This has also impact on
>> performance because Microblaze works with smaller SKBs.
>>
>> Can you please be more specific about passing the paged buffers into GRO?
>> Or point me to any documentation or code which can help me to understand what
>> that means.
>
> You would use napi_get_frags() to get a new or recycled skb, fill in
> skb->frags, then call napi_gro_frags() to pass it into GRO. The benet,
> cxgb3 and sfc drivers do this.
ok. Will look.
Thanks,
Michal
--
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian
^ permalink raw reply
* Re: Add NAPI support to ll_temac driver
From: Eric Dumazet @ 2011-04-19 13:21 UTC (permalink / raw)
To: monstr; +Cc: Ben Hutchings, netdev
In-Reply-To: <4DAD84B1.9020405@monstr.eu>
Le mardi 19 avril 2011 à 14:48 +0200, Michal Simek a écrit :
> About SKB allocation. I fixed our non mainline driver to allocate skb based on
> current mtu size. Mainline driver allocate max mtu (9k). This has also impact on
> performance because Microblaze works with smaller SKBs.
>
Make sure it wont crash/freeze if some evil guy sends a big frame
^ permalink raw reply
* Re: [PATCH] net: xen-netback: convert to hw_features
From: Michał Mirosław @ 2011-04-19 13:30 UTC (permalink / raw)
To: Ian Campbell; +Cc: netdev@vger.kernel.org, xen-devel@lists.xensource.com
In-Reply-To: <1303219073.5997.191.camel@zakaz.uk.xensource.com>
On Tue, Apr 19, 2011 at 02:17:53PM +0100, Ian Campbell wrote:
> On Tue, 2011-04-19 at 12:56 +0100, Michał Mirosław wrote:
> > Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
> Thanks for beating me to this! However the prototype for
> xenvif_fix_features is wrong (needs to take a net_device not a xenvif).
I'll resend v2 with this fix.
> I fixed it with the following, I also moved the !can_sg MTU clamping
> into a set_features hook (like we do with netfront). Am I right that
> this pattern copes with changes to SG via ethtool etc better? I think
> it's more future proof in any case.
This looks wrong. Even if SG is turned on, you might get big skbs which
are linearized. There is a difference in SG capability and SG offload
status and as I see it the capability is what you need to test for MTU.
> NB: I'm having some issues with my test hardware at the moment so this
> is reviewed by eye and compile tested only...
>
> I'm also happy for this to be folded into the original with my
> "Signed-off-/Acked-by Ian Campbell <ian.campbell@citrix.com>" if that is
> preferable.
Thanks.
Best Regards,
Michał Mirosław
^ permalink raw reply
* [PATCH v2] net: xen-netback: convert to hw_features
From: Michał Mirosław @ 2011-04-19 13:35 UTC (permalink / raw)
To: netdev; +Cc: Ian Campbell, xen-devel
In-Reply-To: <20110419115612.C4ACA13909@rere.qmqm.pl>
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
---
v2: fix xenvif_fix_features() arguments
drivers/net/xen-netback/common.h | 3 -
drivers/net/xen-netback/interface.c | 84 ++++++----------------------------
2 files changed, 15 insertions(+), 72 deletions(-)
diff --git a/drivers/net/xen-netback/common.h b/drivers/net/xen-netback/common.h
index 5d7bbf2..8753e6d 100644
--- a/drivers/net/xen-netback/common.h
+++ b/drivers/net/xen-netback/common.h
@@ -73,9 +73,6 @@ struct xenvif {
struct vm_struct *tx_comms_area;
struct vm_struct *rx_comms_area;
- /* Flags that must not be set in dev->features */
- u32 features_disabled;
-
/* Frontend feature information. */
u8 can_sg:1;
u8 gso:1;
diff --git a/drivers/net/xen-netback/interface.c b/drivers/net/xen-netback/interface.c
index de569cc..0ca86f9 100644
--- a/drivers/net/xen-netback/interface.c
+++ b/drivers/net/xen-netback/interface.c
@@ -165,69 +165,18 @@ static int xenvif_change_mtu(struct net_device *dev, int mtu)
return 0;
}
-static void xenvif_set_features(struct xenvif *vif)
-{
- struct net_device *dev = vif->dev;
- u32 features = dev->features;
-
- if (vif->can_sg)
- features |= NETIF_F_SG;
- if (vif->gso || vif->gso_prefix)
- features |= NETIF_F_TSO;
- if (vif->csum)
- features |= NETIF_F_IP_CSUM;
-
- features &= ~(vif->features_disabled);
-
- if (!(features & NETIF_F_SG) && dev->mtu > ETH_DATA_LEN)
- dev->mtu = ETH_DATA_LEN;
-
- dev->features = features;
-}
-
-static int xenvif_set_tx_csum(struct net_device *dev, u32 data)
+static u32 xenvif_fix_features(struct net_device *dev, u32 features)
{
struct xenvif *vif = netdev_priv(dev);
- if (data) {
- if (!vif->csum)
- return -EOPNOTSUPP;
- vif->features_disabled &= ~NETIF_F_IP_CSUM;
- } else {
- vif->features_disabled |= NETIF_F_IP_CSUM;
- }
-
- xenvif_set_features(vif);
- return 0;
-}
-static int xenvif_set_sg(struct net_device *dev, u32 data)
-{
- struct xenvif *vif = netdev_priv(dev);
- if (data) {
- if (!vif->can_sg)
- return -EOPNOTSUPP;
- vif->features_disabled &= ~NETIF_F_SG;
- } else {
- vif->features_disabled |= NETIF_F_SG;
- }
-
- xenvif_set_features(vif);
- return 0;
-}
-
-static int xenvif_set_tso(struct net_device *dev, u32 data)
-{
- struct xenvif *vif = netdev_priv(dev);
- if (data) {
- if (!vif->gso && !vif->gso_prefix)
- return -EOPNOTSUPP;
- vif->features_disabled &= ~NETIF_F_TSO;
- } else {
- vif->features_disabled |= NETIF_F_TSO;
- }
+ if (!vif->can_sg)
+ features &= ~NETIF_F_SG;
+ if (!vif->gso && !vif->gso_prefix)
+ features &= ~NETIF_F_TSO;
+ if (!vif->csum)
+ features &= ~NETIF_F_IP_CSUM;
- xenvif_set_features(vif);
- return 0;
+ return features;
}
static const struct xenvif_stat {
@@ -274,12 +223,6 @@ static void xenvif_get_strings(struct net_device *dev, u32 stringset, u8 * data)
}
static struct ethtool_ops xenvif_ethtool_ops = {
- .get_tx_csum = ethtool_op_get_tx_csum,
- .set_tx_csum = xenvif_set_tx_csum,
- .get_sg = ethtool_op_get_sg,
- .set_sg = xenvif_set_sg,
- .get_tso = ethtool_op_get_tso,
- .set_tso = xenvif_set_tso,
.get_link = ethtool_op_get_link,
.get_sset_count = xenvif_get_sset_count,
@@ -293,6 +236,7 @@ static struct net_device_ops xenvif_netdev_ops = {
.ndo_open = xenvif_open,
.ndo_stop = xenvif_close,
.ndo_change_mtu = xenvif_change_mtu,
+ .ndo_fix_features = xenvif_fix_features,
};
struct xenvif *xenvif_alloc(struct device *parent, domid_t domid,
@@ -331,7 +275,8 @@ struct xenvif *xenvif_alloc(struct device *parent, domid_t domid,
vif->credit_timeout.expires = jiffies;
dev->netdev_ops = &xenvif_netdev_ops;
- xenvif_set_features(vif);
+ dev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_TSO;
+ dev->features = dev->hw_features;
SET_ETHTOOL_OPS(dev, &xenvif_ethtool_ops);
dev->tx_queue_len = XENVIF_QUEUE_LENGTH;
@@ -367,8 +312,6 @@ int xenvif_connect(struct xenvif *vif, unsigned long tx_ring_ref,
if (vif->irq)
return 0;
- xenvif_set_features(vif);
-
err = xen_netbk_map_frontend_rings(vif, tx_ring_ref, rx_ring_ref);
if (err < 0)
goto err;
@@ -384,9 +327,12 @@ int xenvif_connect(struct xenvif *vif, unsigned long tx_ring_ref,
xenvif_get(vif);
rtnl_lock();
- netif_carrier_on(vif->dev);
if (netif_running(vif->dev))
xenvif_up(vif);
+ if (!vif->can_sg && vif->dev->mtu > ETH_DATA_LEN)
+ dev_set_mtu(vif->dev, ETH_DATA_LEN);
+ netdev_update_features(vif->dev);
+ netif_carrier_on(vif->dev);
rtnl_unlock();
return 0;
--
1.7.2.5
^ permalink raw reply related
* Re: [PATCH] net: xen-netback: convert to hw_features
From: Ian Campbell @ 2011-04-19 13:39 UTC (permalink / raw)
To: Michał Mirosław
Cc: netdev@vger.kernel.org, xen-devel@lists.xensource.com
In-Reply-To: <20110419133019.GA3973@rere.qmqm.pl>
On Tue, 2011-04-19 at 14:30 +0100, Michał Mirosław wrote:
> On Tue, Apr 19, 2011 at 02:17:53PM +0100, Ian Campbell wrote:
> > On Tue, 2011-04-19 at 12:56 +0100, Michał Mirosław wrote:
> > > Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
> > Thanks for beating me to this! However the prototype for
> > xenvif_fix_features is wrong (needs to take a net_device not a xenvif).
>
> I'll resend v2 with this fix.
>
> > I fixed it with the following, I also moved the !can_sg MTU clamping
> > into a set_features hook (like we do with netfront). Am I right that
> > this pattern copes with changes to SG via ethtool etc better? I think
> > it's more future proof in any case.
>
> This looks wrong. Even if SG is turned on, you might get big skbs which
> are linearized. There is a difference in SG capability and SG offload
> status and as I see it the capability is what you need to test for MTU.
So the existing stuff in drivers/net/xen-netfront.c is wrong too?
> > NB: I'm having some issues with my test hardware at the moment so this
> > is reviewed by eye and compile tested only...
> >
> > I'm also happy for this to be folded into the original with my
> > "Signed-off-/Acked-by Ian Campbell <ian.campbell@citrix.com>" if that is
> > preferable.
>
> Thanks.
>
> Best Regards,
> Michał Mirosław
^ permalink raw reply
* Re: [PATCH v2] net: xen-netback: convert to hw_features
From: Michał Mirosław @ 2011-04-19 13:39 UTC (permalink / raw)
To: David S. Miller; +Cc: netdev, Ian Campbell, xen-devel
In-Reply-To: <20110419133506.D332413909@rere.qmqm.pl>
On Tue, Apr 19, 2011 at 03:35:06PM +0200, Michał Mirosław wrote:
> Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
> Acked-by: Ian Campbell <ian.campbell@citrix.com>
David, I was too quick with this, so please wait with this patch until Ian
acks it again.
Best Regards,
Michał Mirosław
^ permalink raw reply
* Re: [PATCH] net: xen-netback: convert to hw_features
From: Michał Mirosław @ 2011-04-19 13:43 UTC (permalink / raw)
To: Ian Campbell; +Cc: netdev@vger.kernel.org, xen-devel@lists.xensource.com
In-Reply-To: <1303220340.5997.197.camel@zakaz.uk.xensource.com>
On Tue, Apr 19, 2011 at 02:39:00PM +0100, Ian Campbell wrote:
> On Tue, 2011-04-19 at 14:30 +0100, Michał Mirosław wrote:
> > On Tue, Apr 19, 2011 at 02:17:53PM +0100, Ian Campbell wrote:
> > > I fixed it with the following, I also moved the !can_sg MTU clamping
> > > into a set_features hook (like we do with netfront). Am I right that
> > > this pattern copes with changes to SG via ethtool etc better? I think
> > > it's more future proof in any case.
> > This looks wrong. Even if SG is turned on, you might get big skbs which
> > are linearized. There is a difference in SG capability and SG offload
> > status and as I see it the capability is what you need to test for MTU.
> So the existing stuff in drivers/net/xen-netfront.c is wrong too?
Looks like it. But I don't really know what are the real constraints for MTU.
What I know is that SG even if turned on needs not be used (and currently
it's not e.g. if checksum offload is disabled). So MTU setting should not
depend on SG offload state but on some capability.
Best Regards,
Michał Mirosław
^ permalink raw reply
* [patch net-next-2.6] bonding: move processing of recv handlers into handle_frame()
From: Jiri Pirko @ 2011-04-19 13:48 UTC (permalink / raw)
To: netdev
Cc: davem, shemminger, kaber, fubar, eric.dumazet, nicolas.2p.debian,
andy
Since now when bonding uses rx_handler, all traffic going into bond
device goes thru bond_handle_frame. So there's no need to go back into
bonding code later via ptype handlers. This patch converts
original ptype handlers into "bonding receive probes". These functions
are called from bond_handle_frame and they are registered per-mode.
Note that vlan packets are also handled because they are always untagged
thanks to vlan_untag()
Note that this also allows arpmon for eth-bond-bridge-vlan topology.
Signed-off-by: Jiri Pirko <jpirko@redhat.com>
---
drivers/net/bonding/bond_3ad.c | 29 ++--------
drivers/net/bonding/bond_3ad.h | 4 +-
drivers/net/bonding/bond_alb.c | 46 ++++-----------
drivers/net/bonding/bond_alb.h | 1 -
drivers/net/bonding/bond_main.c | 121 +++++++------------------------------
drivers/net/bonding/bond_sysfs.c | 6 --
drivers/net/bonding/bonding.h | 4 +-
net/core/dev.c | 21 -------
8 files changed, 43 insertions(+), 189 deletions(-)
diff --git a/drivers/net/bonding/bond_3ad.c b/drivers/net/bonding/bond_3ad.c
index 123dd60..d0981c2 100644
--- a/drivers/net/bonding/bond_3ad.c
+++ b/drivers/net/bonding/bond_3ad.c
@@ -2465,35 +2465,16 @@ out:
return NETDEV_TX_OK;
}
-int bond_3ad_lacpdu_recv(struct sk_buff *skb, struct net_device *dev, struct packet_type* ptype, struct net_device *orig_dev)
+void bond_3ad_lacpdu_recv(struct sk_buff *skb, struct bonding *bond,
+ struct slave *slave)
{
- struct bonding *bond = netdev_priv(dev);
- struct slave *slave = NULL;
- int ret = NET_RX_DROP;
-
- if (!(dev->flags & IFF_MASTER))
- goto out;
-
- skb = skb_share_check(skb, GFP_ATOMIC);
- if (!skb)
- goto out;
+ if (skb->protocol != PKT_TYPE_LACPDU)
+ return;
if (!pskb_may_pull(skb, sizeof(struct lacpdu)))
- goto out;
+ return;
read_lock(&bond->lock);
- slave = bond_get_slave_by_dev(netdev_priv(dev), orig_dev);
- if (!slave)
- goto out_unlock;
-
bond_3ad_rx_indication((struct lacpdu *) skb->data, slave, skb->len);
-
- ret = NET_RX_SUCCESS;
-
-out_unlock:
read_unlock(&bond->lock);
-out:
- dev_kfree_skb(skb);
-
- return ret;
}
diff --git a/drivers/net/bonding/bond_3ad.h b/drivers/net/bonding/bond_3ad.h
index b28baff..291dbd4 100644
--- a/drivers/net/bonding/bond_3ad.h
+++ b/drivers/net/bonding/bond_3ad.h
@@ -258,7 +258,6 @@ struct ad_bond_info {
* requested
*/
struct timer_list ad_timer;
- struct packet_type ad_pkt_type;
};
struct ad_slave_info {
@@ -280,7 +279,8 @@ void bond_3ad_adapter_duplex_changed(struct slave *slave);
void bond_3ad_handle_link_change(struct slave *slave, char link);
int bond_3ad_get_active_agg_info(struct bonding *bond, struct ad_info *ad_info);
int bond_3ad_xmit_xor(struct sk_buff *skb, struct net_device *dev);
-int bond_3ad_lacpdu_recv(struct sk_buff *skb, struct net_device *dev, struct packet_type* ptype, struct net_device *orig_dev);
+void bond_3ad_lacpdu_recv(struct sk_buff *skb, struct bonding *bond,
+ struct slave *slave);
int bond_3ad_set_carrier(struct bonding *bond);
#endif //__BOND_3AD_H__
diff --git a/drivers/net/bonding/bond_alb.c b/drivers/net/bonding/bond_alb.c
index ba71582..3b7b040 100644
--- a/drivers/net/bonding/bond_alb.c
+++ b/drivers/net/bonding/bond_alb.c
@@ -308,49 +308,33 @@ static void rlb_update_entry_from_arp(struct bonding *bond, struct arp_pkt *arp)
_unlock_rx_hashtbl(bond);
}
-static int rlb_arp_recv(struct sk_buff *skb, struct net_device *bond_dev, struct packet_type *ptype, struct net_device *orig_dev)
+static void rlb_arp_recv(struct sk_buff *skb, struct bonding *bond,
+ struct slave *slave)
{
- struct bonding *bond;
- struct arp_pkt *arp = (struct arp_pkt *)skb->data;
- int res = NET_RX_DROP;
+ struct arp_pkt *arp;
- while (bond_dev->priv_flags & IFF_802_1Q_VLAN)
- bond_dev = vlan_dev_real_dev(bond_dev);
-
- if (!(bond_dev->priv_flags & IFF_BONDING) ||
- !(bond_dev->flags & IFF_MASTER))
- goto out;
+ if (skb->protocol != cpu_to_be16(ETH_P_ARP))
+ return;
+ arp = (struct arp_pkt *) skb->data;
if (!arp) {
pr_debug("Packet has no ARP data\n");
- goto out;
+ return;
}
- skb = skb_share_check(skb, GFP_ATOMIC);
- if (!skb)
- goto out;
-
- if (!pskb_may_pull(skb, arp_hdr_len(bond_dev)))
- goto out;
+ if (!pskb_may_pull(skb, arp_hdr_len(bond->dev)))
+ return;
if (skb->len < sizeof(struct arp_pkt)) {
pr_debug("Packet is too small to be an ARP\n");
- goto out;
+ return;
}
if (arp->op_code == htons(ARPOP_REPLY)) {
/* update rx hash table for this ARP */
- bond = netdev_priv(bond_dev);
rlb_update_entry_from_arp(bond, arp);
pr_debug("Server received an ARP Reply from client\n");
}
-
- res = NET_RX_SUCCESS;
-
-out:
- dev_kfree_skb(skb);
-
- return res;
}
/* Caller must hold bond lock for read */
@@ -759,7 +743,6 @@ static void rlb_init_table_entry(struct rlb_client_info *entry)
static int rlb_initialize(struct bonding *bond)
{
struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond));
- struct packet_type *pk_type = &(BOND_ALB_INFO(bond).rlb_pkt_type);
struct rlb_client_info *new_hashtbl;
int size = RLB_HASH_TABLE_SIZE * sizeof(struct rlb_client_info);
int i;
@@ -784,13 +767,8 @@ static int rlb_initialize(struct bonding *bond)
_unlock_rx_hashtbl(bond);
- /*initialize packet type*/
- pk_type->type = cpu_to_be16(ETH_P_ARP);
- pk_type->dev = bond->dev;
- pk_type->func = rlb_arp_recv;
-
/* register to receive ARPs */
- dev_add_pack(pk_type);
+ bond->recv_probe = rlb_arp_recv;
return 0;
}
@@ -799,8 +777,6 @@ static void rlb_deinitialize(struct bonding *bond)
{
struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond));
- dev_remove_pack(&(bond_info->rlb_pkt_type));
-
_lock_rx_hashtbl(bond);
kfree(bond_info->rx_hashtbl);
diff --git a/drivers/net/bonding/bond_alb.h b/drivers/net/bonding/bond_alb.h
index 8ca7158..90f140a 100644
--- a/drivers/net/bonding/bond_alb.h
+++ b/drivers/net/bonding/bond_alb.h
@@ -129,7 +129,6 @@ struct alb_bond_info {
int lp_counter;
/* -------- rlb parameters -------- */
int rlb_enabled;
- struct packet_type rlb_pkt_type;
struct rlb_client_info *rx_hashtbl; /* Receive hash table */
spinlock_t rx_hashtbl_lock;
u32 rx_hashtbl_head;
diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index 4ce14bd..66d9dc6 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -1439,27 +1439,17 @@ static void bond_setup_by_slave(struct net_device *bond_dev,
}
/* On bonding slaves other than the currently active slave, suppress
- * duplicates except for 802.3ad ETH_P_SLOW, alb non-mcast/bcast, and
- * ARP on active-backup slaves with arp_validate enabled.
+ * duplicates except for alb non-mcast/bcast.
*/
static bool bond_should_deliver_exact_match(struct sk_buff *skb,
struct slave *slave,
struct bonding *bond)
{
if (bond_is_slave_inactive(slave)) {
- if (slave_do_arp_validate(bond, slave) &&
- skb->protocol == __cpu_to_be16(ETH_P_ARP))
- return false;
-
if (bond->params.mode == BOND_MODE_ALB &&
skb->pkt_type != PACKET_BROADCAST &&
skb->pkt_type != PACKET_MULTICAST)
- return false;
-
- if (bond->params.mode == BOND_MODE_8023AD &&
- skb->protocol == __cpu_to_be16(ETH_P_SLOW))
return false;
-
return true;
}
return false;
@@ -1483,6 +1473,15 @@ static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb)
if (bond->params.arp_interval)
slave->dev->last_rx = jiffies;
+ if (bond->recv_probe) {
+ struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
+
+ if (likely(nskb)) {
+ bond->recv_probe(nskb, bond, slave);
+ dev_kfree_skb(nskb);
+ }
+ }
+
if (bond_should_deliver_exact_match(skb, slave, bond)) {
return RX_HANDLER_EXACT;
}
@@ -2743,48 +2742,26 @@ static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32
}
}
-static int bond_arp_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
+static void bond_arp_rcv(struct sk_buff *skb, struct bonding *bond,
+ struct slave *slave)
{
struct arphdr *arp;
- struct slave *slave;
- struct bonding *bond;
unsigned char *arp_ptr;
__be32 sip, tip;
- if (dev->priv_flags & IFF_802_1Q_VLAN) {
- /*
- * When using VLANS and bonding, dev and oriv_dev may be
- * incorrect if the physical interface supports VLAN
- * acceleration. With this change ARP validation now
- * works for hosts only reachable on the VLAN interface.
- */
- dev = vlan_dev_real_dev(dev);
- orig_dev = dev_get_by_index_rcu(dev_net(skb->dev),skb->skb_iif);
- }
-
- if (!(dev->priv_flags & IFF_BONDING) || !(dev->flags & IFF_MASTER))
- goto out;
+ if (skb->protocol != __cpu_to_be16(ETH_P_ARP))
+ return;
- bond = netdev_priv(dev);
read_lock(&bond->lock);
- pr_debug("bond_arp_rcv: bond %s skb->dev %s orig_dev %s\n",
- bond->dev->name, skb->dev ? skb->dev->name : "NULL",
- orig_dev ? orig_dev->name : "NULL");
+ pr_debug("bond_arp_rcv: bond %s skb->dev %s\n",
+ bond->dev->name, skb->dev->name);
- slave = bond_get_slave_by_dev(bond, orig_dev);
- if (!slave || !slave_do_arp_validate(bond, slave))
- goto out_unlock;
-
- skb = skb_share_check(skb, GFP_ATOMIC);
- if (!skb)
- goto out_unlock;
-
- if (!pskb_may_pull(skb, arp_hdr_len(dev)))
+ if (!pskb_may_pull(skb, arp_hdr_len(bond->dev)))
goto out_unlock;
arp = arp_hdr(skb);
- if (arp->ar_hln != dev->addr_len ||
+ if (arp->ar_hln != bond->dev->addr_len ||
skb->pkt_type == PACKET_OTHERHOST ||
skb->pkt_type == PACKET_LOOPBACK ||
arp->ar_hrd != htons(ARPHRD_ETHER) ||
@@ -2793,9 +2770,9 @@ static int bond_arp_rcv(struct sk_buff *skb, struct net_device *dev, struct pack
goto out_unlock;
arp_ptr = (unsigned char *)(arp + 1);
- arp_ptr += dev->addr_len;
+ arp_ptr += bond->dev->addr_len;
memcpy(&sip, arp_ptr, 4);
- arp_ptr += 4 + dev->addr_len;
+ arp_ptr += 4 + bond->dev->addr_len;
memcpy(&tip, arp_ptr, 4);
pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n",
@@ -2818,9 +2795,6 @@ static int bond_arp_rcv(struct sk_buff *skb, struct net_device *dev, struct pack
out_unlock:
read_unlock(&bond->lock);
-out:
- dev_kfree_skb(skb);
- return NET_RX_SUCCESS;
}
/*
@@ -3407,48 +3381,6 @@ static struct notifier_block bond_inetaddr_notifier = {
.notifier_call = bond_inetaddr_event,
};
-/*-------------------------- Packet type handling ---------------------------*/
-
-/* register to receive lacpdus on a bond */
-static void bond_register_lacpdu(struct bonding *bond)
-{
- struct packet_type *pk_type = &(BOND_AD_INFO(bond).ad_pkt_type);
-
- /* initialize packet type */
- pk_type->type = PKT_TYPE_LACPDU;
- pk_type->dev = bond->dev;
- pk_type->func = bond_3ad_lacpdu_recv;
-
- dev_add_pack(pk_type);
-}
-
-/* unregister to receive lacpdus on a bond */
-static void bond_unregister_lacpdu(struct bonding *bond)
-{
- dev_remove_pack(&(BOND_AD_INFO(bond).ad_pkt_type));
-}
-
-void bond_register_arp(struct bonding *bond)
-{
- struct packet_type *pt = &bond->arp_mon_pt;
-
- if (pt->type)
- return;
-
- pt->type = htons(ETH_P_ARP);
- pt->dev = bond->dev;
- pt->func = bond_arp_rcv;
- dev_add_pack(pt);
-}
-
-void bond_unregister_arp(struct bonding *bond)
-{
- struct packet_type *pt = &bond->arp_mon_pt;
-
- dev_remove_pack(pt);
- pt->type = 0;
-}
-
/*---------------------------- Hashing Policies -----------------------------*/
/*
@@ -3542,14 +3474,14 @@ static int bond_open(struct net_device *bond_dev)
queue_delayed_work(bond->wq, &bond->arp_work, 0);
if (bond->params.arp_validate)
- bond_register_arp(bond);
+ bond->recv_probe = bond_arp_rcv;
}
if (bond->params.mode == BOND_MODE_8023AD) {
INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler);
queue_delayed_work(bond->wq, &bond->ad_work, 0);
/* register to receive LACPDUs */
- bond_register_lacpdu(bond);
+ bond->recv_probe = bond_3ad_lacpdu_recv;
bond_3ad_initiate_agg_selection(bond, 1);
}
@@ -3560,14 +3492,6 @@ static int bond_close(struct net_device *bond_dev)
{
struct bonding *bond = netdev_priv(bond_dev);
- if (bond->params.mode == BOND_MODE_8023AD) {
- /* Unregister the receive of LACPDUs */
- bond_unregister_lacpdu(bond);
- }
-
- if (bond->params.arp_validate)
- bond_unregister_arp(bond);
-
write_lock_bh(&bond->lock);
/* signal timers not to re-arm */
@@ -3604,6 +3528,7 @@ static int bond_close(struct net_device *bond_dev)
*/
bond_alb_deinitialize(bond);
}
+ bond->recv_probe = NULL;
return 0;
}
diff --git a/drivers/net/bonding/bond_sysfs.c b/drivers/net/bonding/bond_sysfs.c
index 259ff32..935406a 100644
--- a/drivers/net/bonding/bond_sysfs.c
+++ b/drivers/net/bonding/bond_sysfs.c
@@ -422,11 +422,6 @@ static ssize_t bonding_store_arp_validate(struct device *d,
bond->dev->name, arp_validate_tbl[new_value].modename,
new_value);
- if (!bond->params.arp_validate && new_value)
- bond_register_arp(bond);
- else if (bond->params.arp_validate && !new_value)
- bond_unregister_arp(bond);
-
bond->params.arp_validate = new_value;
return count;
@@ -923,7 +918,6 @@ static ssize_t bonding_store_miimon(struct device *d,
bond->dev->name);
bond->params.arp_interval = 0;
if (bond->params.arp_validate) {
- bond_unregister_arp(bond);
bond->params.arp_validate =
BOND_ARP_VALIDATE_NONE;
}
diff --git a/drivers/net/bonding/bonding.h b/drivers/net/bonding/bonding.h
index 6126c6a..85fb822 100644
--- a/drivers/net/bonding/bonding.h
+++ b/drivers/net/bonding/bonding.h
@@ -226,6 +226,8 @@ struct bonding {
struct slave *primary_slave;
bool force_primary;
s32 slave_cnt; /* never change this value outside the attach/detach wrappers */
+ void (*recv_probe)(struct sk_buff *, struct bonding *,
+ struct slave *);
rwlock_t lock;
rwlock_t curr_slave_lock;
s8 kill_timers;
@@ -399,8 +401,6 @@ void bond_set_mode_ops(struct bonding *bond, int mode);
int bond_parse_parm(const char *mode_arg, const struct bond_parm_tbl *tbl);
void bond_select_active_slave(struct bonding *bond);
void bond_change_active_slave(struct bonding *bond, struct slave *new_active);
-void bond_register_arp(struct bonding *);
-void bond_unregister_arp(struct bonding *);
void bond_create_debugfs(void);
void bond_destroy_debugfs(void);
void bond_debug_register(struct bonding *bond);
diff --git a/net/core/dev.c b/net/core/dev.c
index 3871bf6..f3aed72 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -3076,25 +3076,6 @@ void netdev_rx_handler_unregister(struct net_device *dev)
}
EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister);
-static void vlan_on_bond_hook(struct sk_buff *skb)
-{
- /*
- * Make sure ARP frames received on VLAN interfaces stacked on
- * bonding interfaces still make their way to any base bonding
- * device that may have registered for a specific ptype.
- */
- if (skb->dev->priv_flags & IFF_802_1Q_VLAN &&
- vlan_dev_real_dev(skb->dev)->priv_flags & IFF_BONDING &&
- skb->protocol == htons(ETH_P_ARP)) {
- struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
-
- if (!skb2)
- return;
- skb2->dev = vlan_dev_real_dev(skb->dev);
- netif_rx(skb2);
- }
-}
-
static int __netif_receive_skb(struct sk_buff *skb)
{
struct packet_type *ptype, *pt_prev;
@@ -3190,8 +3171,6 @@ ncls:
goto out;
}
- vlan_on_bond_hook(skb);
-
/* deliver only exact match when indicated */
null_or_dev = deliver_exact ? skb->dev : NULL;
--
1.7.4.2
^ permalink raw reply related
* Re: [patch net-next-2.6] bonding: move processing of recv handlers into handle_frame()
From: Eric Dumazet @ 2011-04-19 14:02 UTC (permalink / raw)
To: Jiri Pirko
Cc: netdev, davem, shemminger, kaber, fubar, nicolas.2p.debian, andy
In-Reply-To: <1303220896-9092-1-git-send-email-jpirko@redhat.com>
Le mardi 19 avril 2011 à 15:48 +0200, Jiri Pirko a écrit :
> Since now when bonding uses rx_handler, all traffic going into bond
> device goes thru bond_handle_frame. So there's no need to go back into
> bonding code later via ptype handlers. This patch converts
> original ptype handlers into "bonding receive probes". These functions
> are called from bond_handle_frame and they are registered per-mode.
>
> Note that vlan packets are also handled because they are always untagged
> thanks to vlan_untag()
>
> Note that this also allows arpmon for eth-bond-bridge-vlan topology.
>
> Signed-off-by: Jiri Pirko <jpirko@redhat.com>
> }
> diff --git a/drivers/net/bonding/bonding.h b/drivers/net/bonding/bonding.h
> index 6126c6a..85fb822 100644
> --- a/drivers/net/bonding/bonding.h
> +++ b/drivers/net/bonding/bonding.h
> @@ -226,6 +226,8 @@ struct bonding {
> struct slave *primary_slave;
> bool force_primary;
> s32 slave_cnt; /* never change this value outside the attach/detach wrappers */
> + void (*recv_probe)(struct sk_buff *, struct bonding *,
> + struct slave *);
> rwlock_t lock;
> rwlock_t curr_slave_lock;
> s8 kill_timers;
Any performance numbers ?
I am asking because recv_probe sits in a often dirtied cache line...
It would be nice to separate rx & tx path needs
^ permalink raw reply
* Re: [PATCH] net: s390: convert to hw_features
From: Frank Blaschka @ 2011-04-19 14:11 UTC (permalink / raw)
To: Michał Mirosław; +Cc: netdev, linux-s390
In-Reply-To: <20110419104320.314AB13A66@rere.qmqm.pl>
On Tue, Apr 19, 2011 at 12:43:20PM +0200, Michał Mirosław wrote:
> options.large_send was easy to get rid of. options.checksum_type has deeper
> roots so is left for later cleanup.
>
> Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
> ---
Hi Micha,
thx for the patch. There are some small issues I will fix up.
I will add the patch to my queue. It might be a good time to
get rid of the large_send and checksum sysfs attribute as well.
I will work on that too, thx again ...
Frank
^ permalink raw reply
* RE: [PATCH net-next 4/5] be2net: pass domain id to be_cmd_link_status_query
From: Ajit.Khaparde @ 2011-04-19 14:23 UTC (permalink / raw)
To: davem; +Cc: netdev
In-Reply-To: <20110418.225659.112612060.davem@davemloft.net>
> From: David Miller [davem@davemloft.net]
> Sent: Tuesday, April 19, 2011 12:56 AM
> To: Khaparde, Ajit
> Cc: netdev@vger.kernel.org
> Subject: Re: [PATCH net-next 4/5] be2net: pass domain id to be_cmd_link_status_query
> From: Ajit Khaparde <ajit.khaparde@emulex.com>
> Date: Mon, 18 Apr 2011 17:30:33 -0500
>
> Signed-off-by: Ajit Khaparde <ajit.khaparde@emulex.com>
> That's not all this patch does:
Oh.. sorry about that. I did not realize this. I had worked on these patches over a couple of days
and on one of those days, I seem to have mixed the patches. Will respin these again.
> It seems to be also converting the driver over to the new VLAN
> interfaces.
> Please seperate this part into a seperate patch and resubmit your
> patch series.
> Thanks.
^ permalink raw reply
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