* [v3 Patch 1/2] s2io: add dynamic LRO disable support
@ 2010-06-18 10:55 Amerigo Wang
2010-06-18 10:55 ` [v3 Patch 2/2] mlx4: " Amerigo Wang
0 siblings, 1 reply; 6+ messages in thread
From: Amerigo Wang @ 2010-06-18 10:55 UTC (permalink / raw)
To: netdev
Cc: nhorman, sgruszka, herbert.xu, Amerigo Wang, bhutchings,
Ramkrishna.Vepa, davem
This patch adds dynamic LRO diable support for s2io net driver.
I don't have s2io card, so only did compiling test. Anyone who wants
to test this is more than welcome.
This is based on Neil's initial work.
Signed-off-by: WANG Cong <amwang@redhat.com>
Signed-off-by: Neil Horman <nhorman@redhat.com>
Acked-by: Neil Horman <nhorman@redhat.com>
Reviewed-by: Stanislaw Gruszka <sgruszka@redhat.com>
Cc: Ramkrishna Vepa <Ramkrishna.Vepa@exar.com>
---
diff --git a/drivers/net/s2io.c b/drivers/net/s2io.c
index 668327c..e380f37 100644
--- a/drivers/net/s2io.c
+++ b/drivers/net/s2io.c
@@ -795,7 +795,6 @@ static int init_shared_mem(struct s2io_nic *nic)
ring->rx_curr_put_info.ring_len = rx_cfg->num_rxd - 1;
ring->nic = nic;
ring->ring_no = i;
- ring->lro = lro_enable;
blk_cnt = rx_cfg->num_rxd / (rxd_count[nic->rxd_mode] + 1);
/* Allocating all the Rx blocks */
@@ -6684,6 +6683,41 @@ static int s2io_ethtool_op_set_tso(struct net_device *dev, u32 data)
return 0;
}
+static int s2io_ethtool_set_flags(struct net_device *dev, u32 data)
+{
+ struct s2io_nic *sp = netdev_priv(dev);
+ int rc = 0;
+ int changed = 0;
+
+ if (data & (ETH_FLAG_NTUPLE | ETH_FLAG_RXHASH))
+ return -EOPNOTSUPP;
+
+ if (data & ETH_FLAG_LRO) {
+ if (lro_enable) {
+ if (!(dev->features & NETIF_F_LRO)) {
+ dev->features |= NETIF_F_LRO;
+ changed = 1;
+ }
+ } else
+ rc = -EINVAL;
+ } else if (dev->features & NETIF_F_LRO) {
+ dev->features &= ~NETIF_F_LRO;
+ changed = 1;
+ }
+
+ if (changed && netif_running(dev)) {
+ s2io_stop_all_tx_queue(sp);
+ s2io_card_down(sp);
+ sp->lro = dev->features & NETIF_F_LRO;
+ rc = s2io_card_up(sp);
+ if (rc)
+ s2io_reset(sp);
+ else
+ s2io_start_all_tx_queue(sp);
+ }
+
+ return rc;
+}
static const struct ethtool_ops netdev_ethtool_ops = {
.get_settings = s2io_ethtool_gset,
@@ -6701,6 +6735,8 @@ static const struct ethtool_ops netdev_ethtool_ops = {
.get_rx_csum = s2io_ethtool_get_rx_csum,
.set_rx_csum = s2io_ethtool_set_rx_csum,
.set_tx_csum = s2io_ethtool_op_set_tx_csum,
+ .set_flags = s2io_ethtool_set_flags,
+ .get_flags = ethtool_op_get_flags,
.set_sg = ethtool_op_set_sg,
.get_tso = s2io_ethtool_op_get_tso,
.set_tso = s2io_ethtool_op_set_tso,
@@ -7229,6 +7265,7 @@ static int s2io_card_up(struct s2io_nic *sp)
struct ring_info *ring = &mac_control->rings[i];
ring->mtu = dev->mtu;
+ ring->lro = sp->lro;
ret = fill_rx_buffers(sp, ring, 1);
if (ret) {
DBG_PRINT(ERR_DBG, "%s: Out of memory in Open\n",
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [v3 Patch 2/2] mlx4: add dynamic LRO disable support
2010-06-18 10:55 [v3 Patch 1/2] s2io: add dynamic LRO disable support Amerigo Wang
@ 2010-06-18 10:55 ` Amerigo Wang
2010-06-18 11:09 ` Stanislaw Gruszka
0 siblings, 1 reply; 6+ messages in thread
From: Amerigo Wang @ 2010-06-18 10:55 UTC (permalink / raw)
To: netdev
Cc: nhorman, sgruszka, herbert.xu, Amerigo Wang, bhutchings,
Ramkrishna.Vepa, davem
This patch adds dynamic LRO diable support for mlx4 net driver.
It also fixes a bug of mlx4, which checks NETIF_F_LRO flag in rx
path without rtnl lock.
I don't have mlx4 card, so only did compiling test. Anyone who wants
to test this is more than welcome.
This is based on Neil's initial work too.
Signed-off-by: WANG Cong <amwang@redhat.com>
Signed-off-by: Neil Horman <nhorman@redhat.com>
Acked-by: Neil Horman <nhorman@redhat.com>
Reviewed-by: Stanislaw Gruszka <sgruszka@redhat.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
---
diff --git a/drivers/net/mlx4/en_ethtool.c b/drivers/net/mlx4/en_ethtool.c
index d5afd03..d68b796 100644
--- a/drivers/net/mlx4/en_ethtool.c
+++ b/drivers/net/mlx4/en_ethtool.c
@@ -387,6 +387,41 @@ static void mlx4_en_get_ringparam(struct net_device *dev,
param->tx_pending = mdev->profile.prof[priv->port].tx_ring_size;
}
+static int mlx4_ethtool_op_set_flags(struct net_device *dev, u32 data)
+{
+ struct mlx4_en_priv *priv = netdev_priv(dev);
+ struct mlx4_en_dev *mdev = priv->mdev;
+ int rc = 0;
+ int changed = 0;
+
+ if (data & (ETH_FLAG_NTUPLE | ETH_FLAG_RXHASH))
+ return -EOPNOTSUPP;
+
+ if (data & ETH_FLAG_LRO) {
+ if (!(dev->features & NETIF_F_LRO))
+ changed = 1;
+ } else if (dev->features & NETIF_F_LRO) {
+ changed = 1;
+ mdev->profile.num_lro = 0;
+ }
+
+ if (changed) {
+ if (netif_running(dev)) {
+ mutex_lock(&mdev->state_lock);
+ mlx4_en_stop_port(dev);
+ }
+ dev->features ^= NETIF_F_LRO;
+ if (netif_running(dev)) {
+ rc = mlx4_en_start_port(dev);
+ if (rc)
+ en_err(priv, "Failed to restart port\n");
+ mutex_unlock(&mdev->state_lock);
+ }
+ }
+
+ return rc;
+}
+
const struct ethtool_ops mlx4_en_ethtool_ops = {
.get_drvinfo = mlx4_en_get_drvinfo,
.get_settings = mlx4_en_get_settings,
@@ -415,7 +450,7 @@ const struct ethtool_ops mlx4_en_ethtool_ops = {
.get_ringparam = mlx4_en_get_ringparam,
.set_ringparam = mlx4_en_set_ringparam,
.get_flags = ethtool_op_get_flags,
- .set_flags = ethtool_op_set_flags,
+ .set_flags = mlx4_ethtool_op_set_flags,
};
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [v3 Patch 2/2] mlx4: add dynamic LRO disable support
2010-06-18 10:55 ` [v3 Patch 2/2] mlx4: " Amerigo Wang
@ 2010-06-18 11:09 ` Stanislaw Gruszka
2010-06-21 2:23 ` Cong Wang
0 siblings, 1 reply; 6+ messages in thread
From: Stanislaw Gruszka @ 2010-06-18 11:09 UTC (permalink / raw)
To: Amerigo Wang
Cc: netdev, nhorman, herbert.xu, bhutchings, Ramkrishna.Vepa, davem
On Fri, Jun 18, 2010 at 06:55:38AM -0400, Amerigo Wang wrote:
> +static int mlx4_ethtool_op_set_flags(struct net_device *dev, u32 data)
> +{
> + struct mlx4_en_priv *priv = netdev_priv(dev);
> + struct mlx4_en_dev *mdev = priv->mdev;
> + int rc = 0;
> + int changed = 0;
> +
> + if (data & (ETH_FLAG_NTUPLE | ETH_FLAG_RXHASH))
> + return -EOPNOTSUPP;
> +
> + if (data & ETH_FLAG_LRO) {
> + if (!(dev->features & NETIF_F_LRO))
> + changed = 1;
> + } else if (dev->features & NETIF_F_LRO) {
> + changed = 1;
> + mdev->profile.num_lro = 0;
Everything fine except that, what for you zero num_lro value?
If we set it to zero it will stay zero and we will not create
proper number of lro descriptors in mlx4_en_create_rx_ring()
(called from mlx4_en_set_ringparam() -> mlx4_en_alloc_resources())
when someone enable LRO again on.
Stanislaw
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [v3 Patch 2/2] mlx4: add dynamic LRO disable support
2010-06-18 11:09 ` Stanislaw Gruszka
@ 2010-06-21 2:23 ` Cong Wang
2010-06-21 9:02 ` Stanislaw Gruszka
0 siblings, 1 reply; 6+ messages in thread
From: Cong Wang @ 2010-06-21 2:23 UTC (permalink / raw)
To: Stanislaw Gruszka
Cc: netdev, nhorman, herbert.xu, bhutchings, Ramkrishna.Vepa, davem
On 06/18/10 19:09, Stanislaw Gruszka wrote:
> On Fri, Jun 18, 2010 at 06:55:38AM -0400, Amerigo Wang wrote:
>> +static int mlx4_ethtool_op_set_flags(struct net_device *dev, u32 data)
>> +{
>> + struct mlx4_en_priv *priv = netdev_priv(dev);
>> + struct mlx4_en_dev *mdev = priv->mdev;
>> + int rc = 0;
>> + int changed = 0;
>> +
>> + if (data& (ETH_FLAG_NTUPLE | ETH_FLAG_RXHASH))
>> + return -EOPNOTSUPP;
>> +
>> + if (data& ETH_FLAG_LRO) {
>> + if (!(dev->features& NETIF_F_LRO))
>> + changed = 1;
>> + } else if (dev->features& NETIF_F_LRO) {
>> + changed = 1;
>> + mdev->profile.num_lro = 0;
>
> Everything fine except that, what for you zero num_lro value?
>
> If we set it to zero it will stay zero and we will not create
> proper number of lro descriptors in mlx4_en_create_rx_ring()
> (called from mlx4_en_set_ringparam() -> mlx4_en_alloc_resources())
> when someone enable LRO again on.
>
Huh? Isn't ->num_lro which controls LRO of mlx4 driver?
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [v3 Patch 2/2] mlx4: add dynamic LRO disable support
2010-06-21 2:23 ` Cong Wang
@ 2010-06-21 9:02 ` Stanislaw Gruszka
2010-06-22 8:42 ` Cong Wang
0 siblings, 1 reply; 6+ messages in thread
From: Stanislaw Gruszka @ 2010-06-21 9:02 UTC (permalink / raw)
To: Cong Wang; +Cc: netdev, nhorman, herbert.xu, bhutchings, Ramkrishna.Vepa, davem
On Mon, Jun 21, 2010 at 10:23:00AM +0800, Cong Wang wrote:
> On 06/18/10 19:09, Stanislaw Gruszka wrote:
>> On Fri, Jun 18, 2010 at 06:55:38AM -0400, Amerigo Wang wrote:
>>> +static int mlx4_ethtool_op_set_flags(struct net_device *dev, u32 data)
>>> +{
>>> + struct mlx4_en_priv *priv = netdev_priv(dev);
>>> + struct mlx4_en_dev *mdev = priv->mdev;
>>> + int rc = 0;
>>> + int changed = 0;
>>> +
>>> + if (data& (ETH_FLAG_NTUPLE | ETH_FLAG_RXHASH))
>>> + return -EOPNOTSUPP;
As we are here, better would be
if (data & ~ETH_FLAG_LRO)
return -EOPNOTSUPP;
code will persist correct when someone add new flags.
>>> +
>>> + if (data& ETH_FLAG_LRO) {
>>> + if (!(dev->features& NETIF_F_LRO))
>>> + changed = 1;
>>> + } else if (dev->features& NETIF_F_LRO) {
>>> + changed = 1;
>>> + mdev->profile.num_lro = 0;
>>
>> Everything fine except that, what for you zero num_lro value?
>>
>> If we set it to zero it will stay zero and we will not create
>> proper number of lro descriptors in mlx4_en_create_rx_ring()
>> (called from mlx4_en_set_ringparam() -> mlx4_en_alloc_resources())
>> when someone enable LRO again on.
>>
>
> Huh? Isn't ->num_lro which controls LRO of mlx4 driver?
It is, but only in mlx4_en_add() just before register_netdev(),
when we setup default dev->features. Otherwise dev->features
tells if LRO is enabled or disabled.
This realize me, that we should not dev->features |= NETIF_F_LRO
if mdev->profile.num_lro == 0 .
Stanislaw
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [v3 Patch 2/2] mlx4: add dynamic LRO disable support
2010-06-21 9:02 ` Stanislaw Gruszka
@ 2010-06-22 8:42 ` Cong Wang
0 siblings, 0 replies; 6+ messages in thread
From: Cong Wang @ 2010-06-22 8:42 UTC (permalink / raw)
To: Stanislaw Gruszka
Cc: netdev, nhorman, herbert.xu, bhutchings, Ramkrishna.Vepa, davem
On 06/21/10 17:02, Stanislaw Gruszka wrote:
> On Mon, Jun 21, 2010 at 10:23:00AM +0800, Cong Wang wrote:
>> On 06/18/10 19:09, Stanislaw Gruszka wrote:
>>> On Fri, Jun 18, 2010 at 06:55:38AM -0400, Amerigo Wang wrote:
>>>> +static int mlx4_ethtool_op_set_flags(struct net_device *dev, u32 data)
>>>> +{
>>>> + struct mlx4_en_priv *priv = netdev_priv(dev);
>>>> + struct mlx4_en_dev *mdev = priv->mdev;
>>>> + int rc = 0;
>>>> + int changed = 0;
>>>> +
>>>> + if (data& (ETH_FLAG_NTUPLE | ETH_FLAG_RXHASH))
>>>> + return -EOPNOTSUPP;
>
> As we are here, better would be
> if (data& ~ETH_FLAG_LRO)
> return -EOPNOTSUPP;
> code will persist correct when someone add new flags.
Yeah, better.
>
>>>> +
>>>> + if (data& ETH_FLAG_LRO) {
>>>> + if (!(dev->features& NETIF_F_LRO))
>>>> + changed = 1;
>>>> + } else if (dev->features& NETIF_F_LRO) {
>>>> + changed = 1;
>>>> + mdev->profile.num_lro = 0;
>>>
>>> Everything fine except that, what for you zero num_lro value?
>>>
>>> If we set it to zero it will stay zero and we will not create
>>> proper number of lro descriptors in mlx4_en_create_rx_ring()
>>> (called from mlx4_en_set_ringparam() -> mlx4_en_alloc_resources())
>>> when someone enable LRO again on.
>>>
>>
>> Huh? Isn't ->num_lro which controls LRO of mlx4 driver?
>
> It is, but only in mlx4_en_add() just before register_netdev(),
> when we setup default dev->features. Otherwise dev->features
> tells if LRO is enabled or disabled.
>
> This realize me, that we should not dev->features |= NETIF_F_LRO
> if mdev->profile.num_lro == 0 .
>
Got it! I misunderstood ->num_lro. :( Setting NETIF_F_LRO should be enough
for mlx4 driver...
I will send the updated version.
Thanks!
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2010-06-22 8:38 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2010-06-18 10:55 [v3 Patch 1/2] s2io: add dynamic LRO disable support Amerigo Wang
2010-06-18 10:55 ` [v3 Patch 2/2] mlx4: " Amerigo Wang
2010-06-18 11:09 ` Stanislaw Gruszka
2010-06-21 2:23 ` Cong Wang
2010-06-21 9:02 ` Stanislaw Gruszka
2010-06-22 8:42 ` Cong Wang
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).