* Re: [linux-nics] [PATCH] e1000e: Fix 82574/82583 TimeSync errata handling for SYSTIM read
From: Jeff Kirsher @ 2015-01-15 0:36 UTC (permalink / raw)
To: Bhavesh Davda
Cc: linux.nics, netdev, pv-drivers, nithin, ninad, smurali, gyang
In-Reply-To: <1421278252-13622-1-git-send-email-bhavesh@vmware.com>
[-- Attachment #1: Type: text/plain, Size: 565 bytes --]
On Wed, 2015-01-14 at 15:30 -0800, Bhavesh Davda wrote:
> In emulated 82574 vNICs, TIMINCA might read as '0', so this change
> prevents a
> divide-by-zero error.
>
> Signed-off-by: Bhavesh Davda <bhavesh@vmware.com>
> Acked-by: Nithin Raju <nithin@vmware.com>
> Acked-by: Ninad Ghodke <ninad@vmware.com>
> Reviewed-by: Guolin Yang <gyang@vmware.com>
> Reviewed-by: Srividya Murali <smurali@vmware.com>
> ---
> drivers/net/ethernet/intel/e1000e/netdev.c | 3 +++
> 1 file changed, 3 insertions(+)
Thanks Bhavesh, I will add your patch to my queue.
[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 819 bytes --]
^ permalink raw reply
* [PATCHv2] ixgbe: Re-enable relaxed ordering as part of init/restart sequence for non-DCA config
From: Sowmini Varadhan @ 2015-01-15 1:03 UTC (permalink / raw)
To: sowmini.varadhan, jeffrey.t.kirsher, jesse.brandeburg,
bruce.w.allan, carolyn.wyborny, donald.c.skidmore, gregory.v.rose,
matthew.vick, john.ronciak, mitch.a.williams
Cc: linux.nics, e1000-devel, netdev, linux-kernel, sparclinux,
emil.s.tantilov
Relaxed ordering is disabled by default at driver initialization
and re-enabled when DCA is used. The reason it is disabled was
due to an issue on some chipsets (see comments in ixgbe_update_tx_dca()).
But when DCA is not used, RO needs to be re-enabled, else we have
a serialization bottleneck on platforms like SPARC.
This patch eliminates the bottleneck for ixgbe when DCA is not configured.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Cc: Emil Tantilov <emil.s.tantilov@intel.com>
---
v2: incoroporate comments from Emil Tantilov
drivers/net/ethernet/intel/ixgbe/ixgbe_82598.c | 23 +++++++++++++++++++++++
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c | 20 ++++++++++++++++++++
drivers/net/ethernet/intel/ixgbe/ixgbe_common.h | 1 +
drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 11 +++++++++++
drivers/net/ethernet/intel/ixgbe/ixgbe_type.h | 1 +
drivers/net/ethernet/intel/ixgbe/ixgbe_x540.c | 1 +
6 files changed, 57 insertions(+), 0 deletions(-)
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_82598.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_82598.c
index c5c97b4..0c78b5a 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_82598.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_82598.c
@@ -1154,6 +1154,28 @@ static void ixgbe_set_rxpba_82598(struct ixgbe_hw *hw, int num_pb,
IXGBE_WRITE_REG(hw, IXGBE_TXPBSIZE(i), IXGBE_TXPBSIZE_40KB);
}
+static void ixgbe_enable_relaxed_ordering_82598(struct ixgbe_hw *hw)
+{
+ u32 regval;
+ u32 i;
+
+ /* Enable relaxed ordering */
+ for (i = 0; ((i < hw->mac.max_tx_queues) &&
+ (i < IXGBE_DCA_MAX_QUEUES_82598)); i++) {
+ regval = IXGBE_READ_REG(hw, IXGBE_DCA_TXCTRL(i));
+ regval |= IXGBE_DCA_TXCTRL_DESC_WRO_EN;
+ IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL(i), regval);
+ }
+
+ for (i = 0; ((i < hw->mac.max_rx_queues) &&
+ (i < IXGBE_DCA_MAX_QUEUES_82598)); i++) {
+ regval = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
+ regval |= IXGBE_DCA_RXCTRL_DATA_WRO_EN |
+ IXGBE_DCA_RXCTRL_HEAD_WRO_EN;
+ IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(i), regval);
+ }
+}
+
static struct ixgbe_mac_operations mac_ops_82598 = {
.init_hw = &ixgbe_init_hw_generic,
.reset_hw = &ixgbe_reset_hw_82598,
@@ -1161,6 +1183,7 @@ static struct ixgbe_mac_operations mac_ops_82598 = {
.clear_hw_cntrs = &ixgbe_clear_hw_cntrs_generic,
.get_media_type = &ixgbe_get_media_type_82598,
.enable_rx_dma = &ixgbe_enable_rx_dma_generic,
+ .enable_relaxed_ordering = &ixgbe_enable_relaxed_ordering_82598,
.get_mac_addr = &ixgbe_get_mac_addr_generic,
.stop_adapter = &ixgbe_stop_adapter_generic,
.get_bus_info = &ixgbe_get_bus_info_generic,
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
index 9c66bab..4453d92 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
@@ -338,6 +338,26 @@ s32 ixgbe_start_hw_gen2(struct ixgbe_hw *hw)
return 0;
}
+void ixgbe_enable_relaxed_ordering(struct ixgbe_hw *hw)
+{
+ u32 i;
+ u32 regval;
+
+ /* Enable relaxed ordering */
+ for (i = 0; i < hw->mac.max_tx_queues; i++) {
+ regval = IXGBE_READ_REG(hw, IXGBE_DCA_TXCTRL_82599(i));
+ regval |= IXGBE_DCA_TXCTRL_DESC_WRO_EN;
+ IXGBE_WRITE_REG(hw, IXGBE_DCA_TXCTRL_82599(i), regval);
+ }
+
+ for (i = 0; i < hw->mac.max_rx_queues; i++) {
+ regval = IXGBE_READ_REG(hw, IXGBE_DCA_RXCTRL(i));
+ regval |= (IXGBE_DCA_RXCTRL_DATA_WRO_EN |
+ IXGBE_DCA_RXCTRL_HEAD_WRO_EN);
+ IXGBE_WRITE_REG(hw, IXGBE_DCA_RXCTRL(i), regval);
+ }
+}
+
/**
* ixgbe_init_hw_generic - Generic hardware initialization
* @hw: pointer to hardware structure
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_common.h b/drivers/net/ethernet/intel/ixgbe/ixgbe_common.h
index 8cfadcb..c399c18 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_common.h
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_common.h
@@ -37,6 +37,7 @@ s32 ixgbe_init_ops_generic(struct ixgbe_hw *hw);
s32 ixgbe_init_hw_generic(struct ixgbe_hw *hw);
s32 ixgbe_start_hw_generic(struct ixgbe_hw *hw);
s32 ixgbe_start_hw_gen2(struct ixgbe_hw *hw);
+void ixgbe_enable_relaxed_ordering(struct ixgbe_hw *hw);
s32 ixgbe_clear_hw_cntrs_generic(struct ixgbe_hw *hw);
s32 ixgbe_read_pba_string_generic(struct ixgbe_hw *hw, u8 *pba_num,
u32 pba_num_size);
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index 2ed2c7d..dec815b 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -4898,6 +4898,12 @@ void ixgbe_reset(struct ixgbe_adapter *adapter)
if (test_bit(__IXGBE_PTP_RUNNING, &adapter->state))
ixgbe_ptp_reset(adapter);
+
+#ifndef CONFIG_IXGBE_DCA
+ /* ixgbe_update_tx_dca() will re-enable RO when DCA is enabled */
+ if (hw->mac.ops.enable_relaxed_ordering)
+ hw->mac.ops.enable_relaxed_ordering(hw);
+#endif
}
/**
@@ -8470,6 +8476,11 @@ skip_sriov:
"representative who provided you with this "
"hardware.\n");
}
+#ifndef CONFIG_IXGBE_DCA
+ /* ixgbe_update_tx_dca() will re-enable RO when DCA is enabled */
+ if (hw->mac.ops.enable_relaxed_ordering)
+ hw->mac.ops.enable_relaxed_ordering(hw);
+#endif
strcpy(netdev->name, "eth%d");
err = register_netdev(netdev);
if (err)
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_type.h b/drivers/net/ethernet/intel/ixgbe/ixgbe_type.h
index d101b25..b967241 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_type.h
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_type.h
@@ -3013,6 +3013,7 @@ struct ixgbe_mac_operations {
void (*release_swfw_sync)(struct ixgbe_hw *, u32);
s32 (*prot_autoc_read)(struct ixgbe_hw *, bool *, u32 *);
s32 (*prot_autoc_write)(struct ixgbe_hw *, u32, bool);
+ void (*enable_relaxed_ordering)(struct ixgbe_hw *);
/* Link */
void (*disable_tx_laser)(struct ixgbe_hw *);
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_x540.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_x540.c
index ba54ff0..88adad2 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_x540.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_x540.c
@@ -781,6 +781,7 @@ static struct ixgbe_mac_operations mac_ops_X540 = {
.clear_hw_cntrs = &ixgbe_clear_hw_cntrs_generic,
.get_media_type = &ixgbe_get_media_type_X540,
.enable_rx_dma = &ixgbe_enable_rx_dma_generic,
+ .enable_relaxed_ordering = &ixgbe_enable_relaxed_ordering,
.get_mac_addr = &ixgbe_get_mac_addr_generic,
.get_san_mac_addr = &ixgbe_get_san_mac_addr_generic,
.get_device_caps = &ixgbe_get_device_caps_generic,
--
1.7.1
^ permalink raw reply related
* Re: [PATCH 1/5] vxlan: Group Policy extension
From: Tom Herbert @ 2015-01-15 1:08 UTC (permalink / raw)
To: Thomas Graf
Cc: David Miller, Jesse Gross, Stephen Hemminger, Pravin B Shelar,
Alexei Starovoitov, Nicolas Dichtel, Linux Netdev List,
dev@openvswitch.org
In-Reply-To: <20150115002323.GH2105@casper.infradead.org>
On Wed, Jan 14, 2015 at 4:23 PM, Thomas Graf <tgraf@suug.ch> wrote:
> On 01/14/15 at 04:18pm, Tom Herbert wrote:
>> > diff --git a/drivers/net/vxlan.c b/drivers/net/vxlan.c
>> > index 99df0d7..06f7196 100644
>> > --- a/drivers/net/vxlan.c
>> > +++ b/drivers/net/vxlan.c
>> > @@ -126,6 +126,7 @@ struct vxlan_dev {
>> > __u8 tos; /* TOS override */
>> > __u8 ttl;
>> > u32 flags; /* VXLAN_F_* in vxlan.h */
>> > + u32 exts; /* Enabled extensions */
>> >
>>
>> Thomas, why not just make a VXAM_F_GPB flag? Then this setting can be
>> saved in the flags for vxlan_dev and vxlan_sock so no exts field.
>
> Because we need to compare enabled extensions in vxlan_find_sock() to
> make sure we are not sharing a VXLAN socket with extensions enabled
> with a user which does not have the same extensions enabled.
>
> However, we do not want vxlan_find_sock() to compare all flags.
>
> So we need a bitmap that is ignored during the share check (flags) and
> a bitmap that must match to allow sharing (exts).
>
> The RCO extension is currently suffering from this bug which is causing
> a compatibility issue. I explained in the thread of your patch. I was
> under the imrpession that you would either send a v2 or fix it in a
> follow-up.
As I mentioned, we would also need to match receive checksum settings
which is not appropriately called an extension. A mask of interesting
flags could be used to do the comparison in vxlan_find_sock.
^ permalink raw reply
* Re: [PATCH 1/5] vxlan: Group Policy extension
From: Thomas Graf @ 2015-01-15 1:28 UTC (permalink / raw)
To: Tom Herbert
Cc: David Miller, Jesse Gross, Stephen Hemminger, Pravin B Shelar,
Alexei Starovoitov, Nicolas Dichtel, Linux Netdev List,
dev@openvswitch.org
In-Reply-To: <CA+mtBx9=coY1YwPn2ds66PD-J7FDaN_3Yf-_nB_vUEeZEDyJdg@mail.gmail.com>
On 01/14/15 at 05:08pm, Tom Herbert wrote:
> On Wed, Jan 14, 2015 at 4:23 PM, Thomas Graf <tgraf@suug.ch> wrote:
> > Because we need to compare enabled extensions in vxlan_find_sock() to
> > make sure we are not sharing a VXLAN socket with extensions enabled
> > with a user which does not have the same extensions enabled.
> >
> > However, we do not want vxlan_find_sock() to compare all flags.
> >
> > So we need a bitmap that is ignored during the share check (flags) and
> > a bitmap that must match to allow sharing (exts).
> >
> > The RCO extension is currently suffering from this bug which is causing
> > a compatibility issue. I explained in the thread of your patch. I was
> > under the imrpession that you would either send a v2 or fix it in a
> > follow-up.
>
> As I mentioned, we would also need to match receive checksum settings
> which is not appropriately called an extension. A mask of interesting
> flags could be used to do the comparison in vxlan_find_sock.
What exactly is the problem of having a distinct bitmap used by
extensions? It is the least error prone method because it's clear that
all extensions must match and we don't have to maintain an additional
bitmask which can be forgotten to be updated.
If you need to compare additional receive checksum settings for RCO
then that should be separate because as you say it's not an extension.
^ permalink raw reply
* [PATCH] net: sxgbe: Fix NULL dereferece when using DT
From: Kukjin Kim @ 2015-01-15 1:41 UTC (permalink / raw)
To: netdev; +Cc: Dave Miller, Girish K.S, Byungho An, Kukjin Kim
From: Girish K.S <ks.giri@samsung.com>
When the MAC address is provided in the device tree file, the
condition is true and kernel crashes due to NULL dereference.
Signed-off-by: Girish K.S <ks.giri@samsung.com>
Signed-off-by: Byungho An <bh74.an@samsung.com>
Signed-off-by: Kukjin Kim <kgene@kernel.org>
---
drivers/net/ethernet/samsung/sxgbe/sxgbe_platform.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/samsung/sxgbe/sxgbe_platform.c b/drivers/net/ethernet/samsung/sxgbe/sxgbe_platform.c
index 866560e..b02eed1 100644
--- a/drivers/net/ethernet/samsung/sxgbe/sxgbe_platform.c
+++ b/drivers/net/ethernet/samsung/sxgbe/sxgbe_platform.c
@@ -108,10 +108,6 @@ static int sxgbe_platform_probe(struct platform_device *pdev)
}
}
- /* Get MAC address if available (DT) */
- if (mac)
- ether_addr_copy(priv->dev->dev_addr, mac);
-
priv = sxgbe_drv_probe(&(pdev->dev), plat_dat, addr);
if (!priv) {
pr_err("%s: main driver probe failed\n", __func__);
@@ -125,6 +121,10 @@ static int sxgbe_platform_probe(struct platform_device *pdev)
goto err_drv_remove;
}
+ /* Get MAC address if available (DT) */
+ if (mac)
+ ether_addr_copy(priv->dev->dev_addr, mac);
+
/* Get the TX/RX IRQ numbers */
for (i = 0, chan = 1; i < SXGBE_TX_QUEUES; i++) {
priv->txq[i]->irq_no = irq_of_parse_and_map(node, chan++);
--
2.0.0
^ permalink raw reply related
* [PATCH] [PATCH] net: sxgbe: Fix waring for double kfree()
From: Kukjin Kim @ 2015-01-15 1:43 UTC (permalink / raw)
To: netdev; +Cc: Dave Miller, Dan Carpenter, Byungho An, Kukjin Kim
From: Byungho An <bh74.an@samsung.com>
This patch fixes double kfree() calls at init_rx_ring() because
it causes static checker warning.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Byungho An <bh74.an@samsung.com>
Signed-off-by: Kukjin Kim <kgene@kernel.org>
---
drivers/net/ethernet/samsung/sxgbe/sxgbe_main.c | 21 ++++++++++-----------
1 file changed, 10 insertions(+), 11 deletions(-)
diff --git a/drivers/net/ethernet/samsung/sxgbe/sxgbe_main.c b/drivers/net/ethernet/samsung/sxgbe/sxgbe_main.c
index 6984944..fe7538d 100644
--- a/drivers/net/ethernet/samsung/sxgbe/sxgbe_main.c
+++ b/drivers/net/ethernet/samsung/sxgbe/sxgbe_main.c
@@ -474,13 +474,19 @@ static int init_rx_ring(struct net_device *dev, u8 queue_no,
/* allocate memory for RX skbuff array */
rx_ring->rx_skbuff_dma = kmalloc_array(rx_rsize,
sizeof(dma_addr_t), GFP_KERNEL);
- if (rx_ring->rx_skbuff_dma == NULL)
- goto dmamem_err;
+ if (!rx_ring->rx_skbuff_dma) {
+ dma_free_coherent(priv->device,
+ rx_rsize * sizeof(struct sxgbe_rx_norm_desc),
+ rx_ring->dma_rx, rx_ring->dma_rx_phy);
+ goto error;
+ }
rx_ring->rx_skbuff = kmalloc_array(rx_rsize,
sizeof(struct sk_buff *), GFP_KERNEL);
- if (rx_ring->rx_skbuff == NULL)
- goto rxbuff_err;
+ if (!rx_ring->rx_skbuff) {
+ kfree(rx_ring->rx_skbuff_dma);
+ goto error;
+ }
/* initialise the buffers */
for (desc_index = 0; desc_index < rx_rsize; desc_index++) {
@@ -502,13 +508,6 @@ static int init_rx_ring(struct net_device *dev, u8 queue_no,
err_init_rx_buffers:
while (--desc_index >= 0)
free_rx_ring(priv->device, rx_ring, desc_index);
- kfree(rx_ring->rx_skbuff);
-rxbuff_err:
- kfree(rx_ring->rx_skbuff_dma);
-dmamem_err:
- dma_free_coherent(priv->device,
- rx_rsize * sizeof(struct sxgbe_rx_norm_desc),
- rx_ring->dma_rx, rx_ring->dma_rx_phy);
error:
return -ENOMEM;
}
--
2.0.0
^ permalink raw reply related
* Re: [ 2375.793397] WARNING: CPU: 0 PID: 1149 at net/netlink/genetlink.c:1037 genl_unbind+0xc0/0xd0()
From: Jeff Layton @ 2015-01-15 2:20 UTC (permalink / raw)
To: Johannes Berg; +Cc: Jeff Layton, netdev
In-Reply-To: <1421277946.1950.38.camel@sipsolutions.net>
On Thu, 15 Jan 2015 00:25:46 +0100
Johannes Berg <johannes@sipsolutions.net> wrote:
> On Wed, 2015-01-14 at 23:48 +0100, Johannes Berg wrote:
>
> > > [ 2375.793396] ------------[ cut here ]------------
> > > [ 2375.793397] WARNING: CPU: 0 PID: 1149 at net/netlink/genetlink.c:1037 genl_unbind+0xc0/0xd0()
> >
> > This warning is supposed to happen only when you somehow manage to
> > unsubscribe from a generic netlink group that doesn't actually exist, or
> > so.
>
> Ok - after long deliberation I found a way to trigger it. It requires
> that you leave a multicast group (likely by destroying a socket) at the
> same time as the kernel unregisters the generic netlink group. I have no
> idea what generic netlink group you might be using here, but I could
> reproduce it with a strategically placed delay in the netlink code and
> the nl80211 genl group by opening a socket, closing the socket, and
> removing the cfg80211 module (to unregister the nl80211 genl group)
> while the socket was still being closed.
>
> I'll think about a fix tomorrow - it doesn't seem trivial due to
> possible locking concerns.
>
> On the bright side, I cannot see a way to reproduce this without
> removing the genl family at the same time - which is good because it
> means that I've just again audited the case I was worried about (the
> bind/unbind not being symmetric) - it is asymmetric but only in the case
> of genl family removal which seems reasonable (but I should document
> it.)
>
> johannes
>
Cool.
FWIW, it popped around a dozen times or so?
Unfortunately, I didn't save the logs from the run. I'll try to
reproduce it again tomorrow (and save the logs this time), but I don't
see it every time.
--
Jeff Layton <jlayton@primarydata.com>
^ permalink raw reply
* Re: [linux-nics] [PATCHv4 net] i40e: Implement ndo_gso_check()
From: Jesse Gross @ 2015-01-15 2:24 UTC (permalink / raw)
To: Jeff Kirsher
Cc: Joe Stringer, netdev, linux.nics, Linux Kernel Mailing List,
Tom Herbert
In-Reply-To: <CAEP_g=8mS=P2U_vU2ZMEBUkU8jWo5t2P0h9kxF_jbzJhExAZ9g@mail.gmail.com>
On Fri, Dec 26, 2014 at 3:58 PM, Jesse Gross <jesse@nicira.com> wrote:
> On Fri, Dec 5, 2014 at 2:12 PM, Jeff Kirsher
> <jeffrey.t.kirsher@intel.com> wrote:
>> On Fri, 2014-12-05 at 10:41 -0800, Joe Stringer wrote:
>>> ndo_gso_check() was recently introduced to allow NICs to report the
>>> offloading support that they have on a per-skb basis. Add an
>>> implementation for this driver which checks for IPIP, GRE, UDP
>>> tunnels.
>>>
>>> Signed-off-by: Joe Stringer <joestringer@nicira.com>
>>> ---
>>> v4: Simplify the check to just do tunnel header length.
>>> Fix #define style issue.
>>> v3: Drop IPIP and GRE (no driver support even though hw supports it).
>>> Check for UDP outer protocol for UDP tunnels.
>>> v2: Expand to include IP in IP and IPv4/IPv6 inside GRE/UDP tunnels.
>>> Add MAX_INNER_LENGTH (as 80).
>>> ---
>>> drivers/net/ethernet/intel/i40e/i40e_main.c | 12 ++++++++++++
>>> 1 file changed, 12 insertions(+)
>>
>> Thanks Joe, I will update the patch in my queue with your latest
>> version.
>
> Jeff, as a heads-up, this patch and the corresponding one for fm10k
> will no longer apply now that the ndo has changed. This was changed by
> 5f35227e ("net: Generalize ndo_gso_check to ndo_features_check"). The
> update needed is trivial and is just due to the change in the function
> signature but I'm not sure where you are in the testing process with
> this.
Jeff - just wanted to check if you had seen this.
^ permalink raw reply
* Re: [PATCH v2 net] be2net: Allow GRE to work concurrently while a VxLAN tunnel is configured
From: Tom Herbert @ 2015-01-15 2:29 UTC (permalink / raw)
To: Sriharsha Basavapatna; +Cc: Linux Netdev List
In-Reply-To: <1421318323-2547-1-git-send-email-sriharsha.basavapatna@emulex.com>
On Thu, Jan 15, 2015 at 2:38 AM, Sriharsha Basavapatna
<sriharsha.basavapatna@emulex.com> wrote:
> Other tunnels like GRE break while VxLAN offloads are enabled in Skyhawk-R. To
> avoid this, we should restrict offload features on a per-packet basis in such
> conditions.
>
> Signed-off-by: Sriharsha Basavapatna <sriharsha.basavapatna@emulex.com>
> ---
> v2 changes: fixed minor nits pointed out by Sergei Shtylyov
> ---
> drivers/net/ethernet/emulex/benet/be_main.c | 41 +++++++++++++++++++++++++--
> 1 file changed, 38 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/net/ethernet/emulex/benet/be_main.c b/drivers/net/ethernet/emulex/benet/be_main.c
> index 41a0a54..d48806b 100644
> --- a/drivers/net/ethernet/emulex/benet/be_main.c
> +++ b/drivers/net/ethernet/emulex/benet/be_main.c
> @@ -4383,8 +4383,9 @@ static int be_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
> * distinguish various types of transports (VxLAN, GRE, NVGRE ..). So, offload
> * is expected to work across all types of IP tunnels once exported. Skyhawk
> * supports offloads for either VxLAN or NVGRE, exclusively. So we export VxLAN
> - * offloads in hw_enc_features only when a VxLAN port is added. Note this only
> - * ensures that other tunnels work fine while VxLAN offloads are not enabled.
> + * offloads in hw_enc_features only when a VxLAN port is added. If other (non
> + * VxLAN) tunnels are configured while VxLAN offloads are enabled, offloads for
> + * those other tunnels are unexported on the fly through ndo_features_check().
> *
> * Skyhawk supports VxLAN offloads only for one UDP dport. So, if the stack
> * adds more than one port, disable offloads and don't re-enable them again
> @@ -4463,7 +4464,41 @@ static netdev_features_t be_features_check(struct sk_buff *skb,
> struct net_device *dev,
> netdev_features_t features)
> {
> - return vxlan_features_check(skb, features);
> + struct be_adapter *adapter = netdev_priv(dev);
> + u8 l4_hdr = 0;
> +
> + /* The code below restricts offload features for some tunneled packets.
> + * Offload features for normal (non tunnel) packets are unchanged.
> + */
> + if (!skb->encapsulation ||
> + !(adapter->flags & BE_FLAGS_VXLAN_OFFLOADS))
> + return features;
> +
> + /* It's an encapsulated packet and VxLAN offloads are enabled. We
> + * should disable tunnel offload features if it's not a VxLAN packet,
> + * as tunnel offloads have been enabled only for VxLAN. This is done to
> + * allow other tunneled traffic like GRE work fine while VxLAN
> + * offloads are configured in Skyhawk-R.
> + */
> + switch (vlan_get_protocol(skb)) {
> + case htons(ETH_P_IP):
> + l4_hdr = ip_hdr(skb)->protocol;
> + break;
> + case htons(ETH_P_IPV6):
> + l4_hdr = ipv6_hdr(skb)->nexthdr;
> + break;
> + default:
> + return features;
> + }
> +
> + if (l4_hdr != IPPROTO_UDP ||
I don't understand why this is needed. The only GSO type with with
encapsulation allowed by device features SKB_GSO_UDP_TUNNEL. This
should not be GRE for instance. Do you see cases where protocol is not
UDP at this point?
> + skb->inner_protocol_type != ENCAP_TYPE_ETHER ||
> + skb->inner_protocol != htons(ETH_P_TEB) ||
> + skb_inner_mac_header(skb) - skb_transport_header(skb) !=
> + sizeof(struct udphdr) + sizeof(struct vxlanhdr))
> + return features & ~(NETIF_F_ALL_CSUM | NETIF_F_GSO_MASK);
> +
> + return features;
> }
> #endif
>
> --
> 1.7.9.5
>
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
^ permalink raw reply
* Re: [patch net-next v4 1/2] tc: add BPF based action
From: Gao feng @ 2015-01-15 2:42 UTC (permalink / raw)
To: Jiri Pirko, netdev; +Cc: davem, jhs, dborkman, ast, hannes
In-Reply-To: <1421257404-25452-1-git-send-email-jiri@resnulli.us>
On 01/15/2015 01:43 AM, Jiri Pirko wrote:
>
> +config NET_ACT_BPF
> + tristate "BPF based action"
> + depends on NET_CLS_ACT
> + ---help---
> + Say Y here to execute BPF code on packets. The BPF code will decide
> + if the packet should be dropped of not.
dropped or not ?
^ permalink raw reply
* linux-next: manual merge of the net-next tree with the net tree
From: Stephen Rothwell @ 2015-01-15 2:47 UTC (permalink / raw)
To: David Miller, netdev; +Cc: linux-next, linux-kernel, David Vrabel
[-- Attachment #1: Type: text/plain, Size: 1485 bytes --]
Hi all,
Today's linux-next merge of the net-next tree got a conflict in
drivers/net/xen-netfront.c between commit 900e183301b5 ("xen-netfront:
use different locks for Rx and Tx stats") from the net tree and commit
a55e8bb8fb89 ("xen-netfront: refactor making Tx requests") from the
net-next tree.
I fixed it up (see below) and can carry the fix as necessary (no action
is required).
--
Cheers,
Stephen Rothwell sfr@canb.auug.org.au
diff --cc drivers/net/xen-netfront.c
index d8c10764f130,01a4350eb313..000000000000
--- a/drivers/net/xen-netfront.c
+++ b/drivers/net/xen-netfront.c
@@@ -562,18 -518,15 +517,15 @@@ static u16 xennet_select_queue(struct n
static int xennet_start_xmit(struct sk_buff *skb, struct net_device *dev)
{
- unsigned short id;
struct netfront_info *np = netdev_priv(dev);
- struct netfront_stats *stats = this_cpu_ptr(np->stats);
+ struct netfront_stats *tx_stats = this_cpu_ptr(np->tx_stats);
- struct xen_netif_tx_request *tx;
- char *data = skb->data;
- RING_IDX i;
- grant_ref_t ref;
- unsigned long mfn;
+ struct xen_netif_tx_request *tx, *first_tx;
+ unsigned int i;
int notify;
int slots;
- unsigned int offset = offset_in_page(data);
- unsigned int len = skb_headlen(skb);
+ struct page *page;
+ unsigned int offset;
+ unsigned int len;
unsigned long flags;
struct netfront_queue *queue = NULL;
unsigned int num_queues = dev->real_num_tx_queues;
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]
^ permalink raw reply
* Re: [PATCH iproute2] ss: Filter inet dgram sockets with established state by default
From: Stephen Hemminger @ 2015-01-15 2:48 UTC (permalink / raw)
To: Vadim Kochan; +Cc: netdev
In-Reply-To: <20150114224347.GA14983@angus-think.lan>
On Thu, 15 Jan 2015 00:43:47 +0200
Vadim Kochan <vadim4j@gmail.com> wrote:
> On Wed, Jan 14, 2015 at 02:41:20PM -0800, Stephen Hemminger wrote:
> > On Wed, 14 Jan 2015 08:49:44 +0200
> > Vadim Kochan <vadim4j@gmail.com> wrote:
> >
> > > On Tue, Jan 13, 2015 at 05:31:50PM -0800, Stephen Hemminger wrote:
> > > > On Thu, 8 Jan 2015 19:32:22 +0200
> > > > Vadim Kochan <vadim4j@gmail.com> wrote:
> > > >
> > > > > From: Vadim Kochan <vadim4j@gmail.com>
> > > > >
> > > > > As inet dgram sockets (udp, raw) can call connect(...) - they
> > > > > might be set in ESTABLISHED state. So keep the original behaviour of
> > > > > 'ss' which filtered them by ESTABLISHED state by default. So:
> > > > >
> > > > > $ ss -u
> > > > >
> > > > > or
> > > > >
> > > > > $ ss -w
> > > > >
> > > > > Will show only ESTABLISHED UDP sockets by default.
> > > > >
> > > > > Signed-off-by: Vadim Kochan <vadim4j@gmail.com>
> > > > > ---
> > > > > misc/ss.c | 4 ++--
> > > > > 1 file changed, 2 insertions(+), 2 deletions(-)
> > > > >
> > > > > diff --git a/misc/ss.c b/misc/ss.c
> > > > > index 08d210a..015d829 100644
> > > > > --- a/misc/ss.c
> > > > > +++ b/misc/ss.c
> > > > > @@ -170,11 +170,11 @@ static const struct filter default_dbs[MAX_DB] = {
> > > > > .families = (1 << AF_INET) | (1 << AF_INET6),
> > > > > },
> > > > > [UDP_DB] = {
> > > > > - .states = (1 << SS_CLOSE),
> > > > > + .states = (1 << SS_ESTABLISHED),
> > > > > .families = (1 << AF_INET) | (1 << AF_INET6),
> > > > > },
> > > > > [RAW_DB] = {
> > > > > - .states = (1 << SS_CLOSE),
> > > > > + .states = (1 << SS_ESTABLISHED),
> > > > > .families = (1 << AF_INET) | (1 << AF_INET6),
> > > > > },
> > > > > [UNIX_DG_DB] = {
> > > >
> > > > This is a change likely to break somebody using 'ss -u' now and the bound
> > > > sockets will disappear from the output.
> > > >
> > >
> > > But thats was as original behaviour before I added table-driven code
> > > (about few commits ago), so thats a rather fix (sorry I did not noticed
> > > about it) to keep the previous behaviour for dgram sockets - show
> > > established states by default.
> > >
> > > Regards,
> >
> > Ok, I will merge it and update the comments.
> Even with this PATCH I am still confused what is preferred behaviour -
> show established dgram sockets (as it was all the way) or closed + established by default.
>
> What do you think ?
>
> Thanks,
Make it work like earliest releases (like 3 months ago).
^ permalink raw reply
* Re: iproute2 ss: Some thoughts about additional info output layout
From: Stephen Hemminger @ 2015-01-14 22:55 UTC (permalink / raw)
To: Vadim Kochan; +Cc: netdev
In-Reply-To: <20150108221240.GA23636@angus-think.lan>
On Fri, 9 Jan 2015 00:12:40 +0200
Vadim Kochan <vadim4j@gmail.com> wrote:
> Hi,
>
> I think that current output of ss utility visually looks little weird
> when additional info options were specified, so I feel that a lot of
> yours will say that it will break existing scripts of ss output parsing
> but I will try, so thats how I think ss output would looks better in
> case if additionally info options were specified (I am not sure how it
> would looks in the email) :
The current output is pretty much unparseable for options with ss.
For a parseable output generating JSON would be much better, I did that for
a couple of utilities.
I see no problem with having a better human format.
^ permalink raw reply
* [PATCH 4/5] openvswitch: Allow for any level of nesting in flow attributes
From: Thomas Graf @ 2015-01-15 2:53 UTC (permalink / raw)
To: davem-fT/PcQaiUtIeIZ0/mPfg9Q, jesse-l0M0P4e3n4LQT0dZR+AlfA,
stephen-OTpzqLSitTUnbdJkjeBofR2eb7JE58TQ,
pshelar-l0M0P4e3n4LQT0dZR+AlfA, therbert-hpIqsD4AKlfQT0dZR+AlfA,
alexei.starovoitov-Re5JQEeQqe8AvxtiuMwx3w,
nicolas.dichtel-pdR9zngts4EAvxtiuMwx3w
Cc: dev-yBygre7rU0TnMu66kgdUjQ, netdev-u79uwXL29TY76Z2rM5mHXA
In-Reply-To: <cover.1421290198.git.tgraf-G/eBtMaohhA@public.gmane.org>
nlattr_set() is currently hardcoded to two levels of nesting. This change
introduces struct ovs_len_tbl to define minimal length requirements plus
next level nesting tables to traverse the key attributes to arbitrary depth.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
---
v5->v6:
- No change
v4->v5:
- No change
v3->v4:
- No change. The spotted bug is unrelatd to this series and will be fixed
in a separate patch
v2->v3:
- No change
v1->v2:
- New patch to allow nested Netlink attributes inside
OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS
net/openvswitch/flow_netlink.c | 106 ++++++++++++++++++++++-------------------
1 file changed, 56 insertions(+), 50 deletions(-)
diff --git a/net/openvswitch/flow_netlink.c b/net/openvswitch/flow_netlink.c
index 2e8a9cd..518941c 100644
--- a/net/openvswitch/flow_netlink.c
+++ b/net/openvswitch/flow_netlink.c
@@ -50,6 +50,13 @@
#include "flow_netlink.h"
+struct ovs_len_tbl {
+ int len;
+ const struct ovs_len_tbl *next;
+};
+
+#define OVS_ATTR_NESTED -1
+
static void update_range(struct sw_flow_match *match,
size_t offset, size_t size, bool is_mask)
{
@@ -289,29 +296,44 @@ size_t ovs_key_attr_size(void)
+ nla_total_size(28); /* OVS_KEY_ATTR_ND */
}
+static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
+ [OVS_TUNNEL_KEY_ATTR_ID] = { .len = sizeof(u64) },
+ [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = { .len = sizeof(u32) },
+ [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = { .len = sizeof(u32) },
+ [OVS_TUNNEL_KEY_ATTR_TOS] = { .len = 1 },
+ [OVS_TUNNEL_KEY_ATTR_TTL] = { .len = 1 },
+ [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = { .len = 0 },
+ [OVS_TUNNEL_KEY_ATTR_CSUM] = { .len = 0 },
+ [OVS_TUNNEL_KEY_ATTR_TP_SRC] = { .len = sizeof(u16) },
+ [OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
+ [OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
+ [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_NESTED },
+};
+
/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
-static const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
- [OVS_KEY_ATTR_ENCAP] = -1,
- [OVS_KEY_ATTR_PRIORITY] = sizeof(u32),
- [OVS_KEY_ATTR_IN_PORT] = sizeof(u32),
- [OVS_KEY_ATTR_SKB_MARK] = sizeof(u32),
- [OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet),
- [OVS_KEY_ATTR_VLAN] = sizeof(__be16),
- [OVS_KEY_ATTR_ETHERTYPE] = sizeof(__be16),
- [OVS_KEY_ATTR_IPV4] = sizeof(struct ovs_key_ipv4),
- [OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6),
- [OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp),
- [OVS_KEY_ATTR_TCP_FLAGS] = sizeof(__be16),
- [OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp),
- [OVS_KEY_ATTR_SCTP] = sizeof(struct ovs_key_sctp),
- [OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp),
- [OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6),
- [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp),
- [OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd),
- [OVS_KEY_ATTR_RECIRC_ID] = sizeof(u32),
- [OVS_KEY_ATTR_DP_HASH] = sizeof(u32),
- [OVS_KEY_ATTR_TUNNEL] = -1,
- [OVS_KEY_ATTR_MPLS] = sizeof(struct ovs_key_mpls),
+static const struct ovs_len_tbl ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
+ [OVS_KEY_ATTR_ENCAP] = { .len = OVS_ATTR_NESTED },
+ [OVS_KEY_ATTR_PRIORITY] = { .len = sizeof(u32) },
+ [OVS_KEY_ATTR_IN_PORT] = { .len = sizeof(u32) },
+ [OVS_KEY_ATTR_SKB_MARK] = { .len = sizeof(u32) },
+ [OVS_KEY_ATTR_ETHERNET] = { .len = sizeof(struct ovs_key_ethernet) },
+ [OVS_KEY_ATTR_VLAN] = { .len = sizeof(__be16) },
+ [OVS_KEY_ATTR_ETHERTYPE] = { .len = sizeof(__be16) },
+ [OVS_KEY_ATTR_IPV4] = { .len = sizeof(struct ovs_key_ipv4) },
+ [OVS_KEY_ATTR_IPV6] = { .len = sizeof(struct ovs_key_ipv6) },
+ [OVS_KEY_ATTR_TCP] = { .len = sizeof(struct ovs_key_tcp) },
+ [OVS_KEY_ATTR_TCP_FLAGS] = { .len = sizeof(__be16) },
+ [OVS_KEY_ATTR_UDP] = { .len = sizeof(struct ovs_key_udp) },
+ [OVS_KEY_ATTR_SCTP] = { .len = sizeof(struct ovs_key_sctp) },
+ [OVS_KEY_ATTR_ICMP] = { .len = sizeof(struct ovs_key_icmp) },
+ [OVS_KEY_ATTR_ICMPV6] = { .len = sizeof(struct ovs_key_icmpv6) },
+ [OVS_KEY_ATTR_ARP] = { .len = sizeof(struct ovs_key_arp) },
+ [OVS_KEY_ATTR_ND] = { .len = sizeof(struct ovs_key_nd) },
+ [OVS_KEY_ATTR_RECIRC_ID] = { .len = sizeof(u32) },
+ [OVS_KEY_ATTR_DP_HASH] = { .len = sizeof(u32) },
+ [OVS_KEY_ATTR_TUNNEL] = { .len = OVS_ATTR_NESTED,
+ .next = ovs_tunnel_key_lens, },
+ [OVS_KEY_ATTR_MPLS] = { .len = sizeof(struct ovs_key_mpls) },
};
static bool is_all_zero(const u8 *fp, size_t size)
@@ -352,8 +374,8 @@ static int __parse_flow_nlattrs(const struct nlattr *attr,
return -EINVAL;
}
- expected_len = ovs_key_lens[type];
- if (nla_len(nla) != expected_len && expected_len != -1) {
+ expected_len = ovs_key_lens[type].len;
+ if (nla_len(nla) != expected_len && expected_len != OVS_ATTR_NESTED) {
OVS_NLERR(log, "Key %d has unexpected len %d expected %d",
type, nla_len(nla), expected_len);
return -EINVAL;
@@ -451,30 +473,16 @@ static int ipv4_tun_from_nlattr(const struct nlattr *attr,
int type = nla_type(a);
int err;
- static const u32 ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
- [OVS_TUNNEL_KEY_ATTR_ID] = sizeof(u64),
- [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = sizeof(u32),
- [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = sizeof(u32),
- [OVS_TUNNEL_KEY_ATTR_TOS] = 1,
- [OVS_TUNNEL_KEY_ATTR_TTL] = 1,
- [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = 0,
- [OVS_TUNNEL_KEY_ATTR_CSUM] = 0,
- [OVS_TUNNEL_KEY_ATTR_TP_SRC] = sizeof(u16),
- [OVS_TUNNEL_KEY_ATTR_TP_DST] = sizeof(u16),
- [OVS_TUNNEL_KEY_ATTR_OAM] = 0,
- [OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = -1,
- };
-
if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
OVS_NLERR(log, "Tunnel attr %d out of range max %d",
type, OVS_TUNNEL_KEY_ATTR_MAX);
return -EINVAL;
}
- if (ovs_tunnel_key_lens[type] != nla_len(a) &&
- ovs_tunnel_key_lens[type] != -1) {
+ if (ovs_tunnel_key_lens[type].len != nla_len(a) &&
+ ovs_tunnel_key_lens[type].len != OVS_ATTR_NESTED) {
OVS_NLERR(log, "Tunnel attr %d has unexpected len %d expected %d",
- type, nla_len(a), ovs_tunnel_key_lens[type]);
+ type, nla_len(a), ovs_tunnel_key_lens[type].len);
return -EINVAL;
}
@@ -912,18 +920,16 @@ static int ovs_key_from_nlattrs(struct sw_flow_match *match, u64 attrs,
return 0;
}
-static void nlattr_set(struct nlattr *attr, u8 val, bool is_attr_mask_key)
+static void nlattr_set(struct nlattr *attr, u8 val,
+ const struct ovs_len_tbl *tbl)
{
struct nlattr *nla;
int rem;
/* The nlattr stream should already have been validated */
nla_for_each_nested(nla, attr, rem) {
- /* We assume that ovs_key_lens[type] == -1 means that type is a
- * nested attribute
- */
- if (is_attr_mask_key && ovs_key_lens[nla_type(nla)] == -1)
- nlattr_set(nla, val, false);
+ if (tbl && tbl[nla_type(nla)].len == OVS_ATTR_NESTED)
+ nlattr_set(nla, val, tbl[nla_type(nla)].next);
else
memset(nla_data(nla), val, nla_len(nla));
}
@@ -931,7 +937,7 @@ static void nlattr_set(struct nlattr *attr, u8 val, bool is_attr_mask_key)
static void mask_set_nlattr(struct nlattr *attr, u8 val)
{
- nlattr_set(attr, val, true);
+ nlattr_set(attr, val, ovs_key_lens);
}
/**
@@ -1628,8 +1634,8 @@ static int validate_set(const struct nlattr *a,
return -EINVAL;
if (key_type > OVS_KEY_ATTR_MAX ||
- (ovs_key_lens[key_type] != nla_len(ovs_key) &&
- ovs_key_lens[key_type] != -1))
+ (ovs_key_lens[key_type].len != nla_len(ovs_key) &&
+ ovs_key_lens[key_type].len != OVS_ATTR_NESTED))
return -EINVAL;
switch (key_type) {
--
1.9.3
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
^ permalink raw reply related
* [PATCH 1/5] vxlan: Group Policy extension
From: Thomas Graf @ 2015-01-15 2:53 UTC (permalink / raw)
To: davem, jesse, stephen, pshelar, therbert, alexei.starovoitov,
nicolas.dichtel
Cc: netdev, dev
In-Reply-To: <cover.1421290198.git.tgraf@suug.ch>
Implements supports for the Group Policy VXLAN extension [0] to provide
a lightweight and simple security label mechanism across network peers
based on VXLAN. The security context and associated metadata is mapped
to/from skb->mark. This allows further mapping to a SELinux context
using SECMARK, to implement ACLs directly with nftables, iptables, OVS,
tc, etc.
The group membership is defined by the lower 16 bits of skb->mark, the
upper 16 bits are used for flags.
SELinux allows to manage label to secure local resources. However,
distributed applications require ACLs to implemented across hosts. This
is typically achieved by matching on L2-L4 fields to identify the
original sending host and process on the receiver. On top of that,
netlabel and specifically CIPSO [1] allow to map security contexts to
universal labels. However, netlabel and CIPSO are relatively complex.
This patch provides a lightweight alternative for overlay network
environments with a trusted underlay. No additional control protocol
is required.
Host 1: Host 2:
Group A Group B Group B Group A
+-----+ +-------------+ +-------+ +-----+
| lxc | | SELinux CTX | | httpd | | VM |
+--+--+ +--+----------+ +---+---+ +--+--+
\---+---/ \----+---/
| |
+---+---+ +---+---+
| vxlan | | vxlan |
+---+---+ +---+---+
+------------------------------+
Backwards compatibility:
A VXLAN-GBP socket can receive standard VXLAN frames and will assign
the default group 0x0000 to such frames. A Linux VXLAN socket will
drop VXLAN-GBP frames. The extension is therefore disabled by default
and needs to be specifically enabled:
ip link add [...] type vxlan [...] gbp
In a mixed environment with VXLAN and VXLAN-GBP sockets, the GBP socket
must run on a separate port number.
Examples:
iptables:
host1# iptables -I OUTPUT -m owner --uid-owner 101 -j MARK --set-mark 0x200
host2# iptables -I INPUT -m mark --mark 0x200 -j DROP
OVS:
# ovs-ofctl add-flow br0 'in_port=1,actions=load:0x200->NXM_NX_TUN_GBP_ID[],NORMAL'
# ovs-ofctl add-flow br0 'in_port=2,tun_gbp_id=0x200,actions=drop'
[0] https://tools.ietf.org/html/draft-smith-vxlan-group-policy
[1] http://lwn.net/Articles/204905/
Signed-off-by: Thomas Graf <tgraf@suug.ch>
---
v5->v6:
- Use flags instead of exts member to store enablement of GBP as suggested
by Tom
v4->v5:
- Rebased on top of Tom's RCO work
- Dropped IFLA_VXLAN_EXTENSION container attribute and embedded IFLA_VXLAN_GBP
as top level VXLAN attribute like RCO for consistency.
v3->v4:
- Patch 1 was no longer needed due to Tom Herbert's 3bf394 ("vxlan: Improve
support for header flags"). Moved remaining header description to this patch.
- Zero out vxlan_metadata in vxlan_tnl_send() as suggested by Jesse.
- Reported enabled extensions to user space as requested by Nicolas.
- Use VXLAN_HF_GBP instead of bitfield to be in line with Tom's work.
v2->v3:
- Removed empty struct vxlan_gbp as spotted by Alexei
v1->v2:
- split GBP header definition into separate struct vxlanhdr_gbp as requested
by Alexei
drivers/net/vxlan.c | 84 ++++++++++++++++++++++++++++++++++++-------
include/net/vxlan.h | 79 +++++++++++++++++++++++++++++++++++++---
include/uapi/linux/if_link.h | 1 +
net/openvswitch/vport-vxlan.c | 9 +++--
4 files changed, 152 insertions(+), 21 deletions(-)
diff --git a/drivers/net/vxlan.c b/drivers/net/vxlan.c
index 99df0d7..6dbf8e0 100644
--- a/drivers/net/vxlan.c
+++ b/drivers/net/vxlan.c
@@ -620,7 +620,8 @@ static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
continue;
vh2 = (struct vxlanhdr *)(p->data + off_vx);
- if (vh->vx_vni != vh2->vx_vni) {
+ if (vh->vx_flags != vh2->vx_flags ||
+ vh->vx_vni != vh2->vx_vni) {
NAPI_GRO_CB(p)->same_flow = 0;
continue;
}
@@ -1183,6 +1184,7 @@ static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
struct vxlan_sock *vs;
struct vxlanhdr *vxh;
u32 flags, vni;
+ struct vxlan_metadata md = {0};
/* Need Vxlan and inner Ethernet header to be present */
if (!pskb_may_pull(skb, VXLAN_HLEN))
@@ -1216,6 +1218,24 @@ static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
vni &= VXLAN_VID_MASK;
}
+ /* For backwards compatibility, only allow reserved fields to be
+ * used by VXLAN extensions if explicitly requested.
+ */
+ if ((flags & VXLAN_HF_GBP) && (vs->flags & VXLAN_F_GBP)) {
+ struct vxlanhdr_gbp *gbp;
+
+ gbp = (struct vxlanhdr_gbp *)vxh;
+ md.gbp = ntohs(gbp->policy_id);
+
+ if (gbp->dont_learn)
+ md.gbp |= VXLAN_GBP_DONT_LEARN;
+
+ if (gbp->policy_applied)
+ md.gbp |= VXLAN_GBP_POLICY_APPLIED;
+
+ flags &= ~VXLAN_GBP_USED_BITS;
+ }
+
if (flags || (vni & ~VXLAN_VID_MASK)) {
/* If there are any unprocessed flags remaining treat
* this as a malformed packet. This behavior diverges from
@@ -1229,7 +1249,8 @@ static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
goto bad_flags;
}
- vs->rcv(vs, skb, vxh->vx_vni);
+ md.vni = vxh->vx_vni;
+ vs->rcv(vs, skb, &md);
return 0;
drop:
@@ -1246,8 +1267,8 @@ error:
return 1;
}
-static void vxlan_rcv(struct vxlan_sock *vs,
- struct sk_buff *skb, __be32 vx_vni)
+static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
+ struct vxlan_metadata *md)
{
struct iphdr *oip = NULL;
struct ipv6hdr *oip6 = NULL;
@@ -1258,7 +1279,7 @@ static void vxlan_rcv(struct vxlan_sock *vs,
int err = 0;
union vxlan_addr *remote_ip;
- vni = ntohl(vx_vni) >> 8;
+ vni = ntohl(md->vni) >> 8;
/* Is this VNI defined? */
vxlan = vxlan_vs_find_vni(vs, vni);
if (!vxlan)
@@ -1292,6 +1313,7 @@ static void vxlan_rcv(struct vxlan_sock *vs,
goto drop;
skb_reset_network_header(skb);
+ skb->mark = md->gbp;
if (oip6)
err = IP6_ECN_decapsulate(oip6, skb);
@@ -1641,13 +1663,30 @@ static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
return false;
}
+static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, struct vxlan_sock *vs,
+ struct vxlan_metadata *md)
+{
+ struct vxlanhdr_gbp *gbp;
+
+ gbp = (struct vxlanhdr_gbp *)vxh;
+ vxh->vx_flags |= htonl(VXLAN_HF_GBP);
+
+ if (md->gbp & VXLAN_GBP_DONT_LEARN)
+ gbp->dont_learn = 1;
+
+ if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
+ gbp->policy_applied = 1;
+
+ gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
+}
+
#if IS_ENABLED(CONFIG_IPV6)
static int vxlan6_xmit_skb(struct vxlan_sock *vs,
struct dst_entry *dst, struct sk_buff *skb,
struct net_device *dev, struct in6_addr *saddr,
struct in6_addr *daddr, __u8 prio, __u8 ttl,
- __be16 src_port, __be16 dst_port, __be32 vni,
- bool xnet)
+ __be16 src_port, __be16 dst_port,
+ struct vxlan_metadata *md, bool xnet)
{
struct vxlanhdr *vxh;
int min_headroom;
@@ -1696,7 +1735,7 @@ static int vxlan6_xmit_skb(struct vxlan_sock *vs,
vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
vxh->vx_flags = htonl(VXLAN_HF_VNI);
- vxh->vx_vni = vni;
+ vxh->vx_vni = md->vni;
if (type & SKB_GSO_TUNNEL_REMCSUM) {
u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
@@ -1714,6 +1753,9 @@ static int vxlan6_xmit_skb(struct vxlan_sock *vs,
}
}
+ if (vs->flags & VXLAN_F_GBP)
+ vxlan_build_gbp_hdr(vxh, vs, md);
+
skb_set_inner_protocol(skb, htons(ETH_P_TEB));
udp_tunnel6_xmit_skb(vs->sock, dst, skb, dev, saddr, daddr, prio,
@@ -1728,7 +1770,8 @@ err:
int vxlan_xmit_skb(struct vxlan_sock *vs,
struct rtable *rt, struct sk_buff *skb,
__be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
- __be16 src_port, __be16 dst_port, __be32 vni, bool xnet)
+ __be16 src_port, __be16 dst_port,
+ struct vxlan_metadata *md, bool xnet)
{
struct vxlanhdr *vxh;
int min_headroom;
@@ -1771,7 +1814,7 @@ int vxlan_xmit_skb(struct vxlan_sock *vs,
vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
vxh->vx_flags = htonl(VXLAN_HF_VNI);
- vxh->vx_vni = vni;
+ vxh->vx_vni = md->vni;
if (type & SKB_GSO_TUNNEL_REMCSUM) {
u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
@@ -1789,6 +1832,9 @@ int vxlan_xmit_skb(struct vxlan_sock *vs,
}
}
+ if (vs->flags & VXLAN_F_GBP)
+ vxlan_build_gbp_hdr(vxh, vs, md);
+
skb_set_inner_protocol(skb, htons(ETH_P_TEB));
return udp_tunnel_xmit_skb(vs->sock, rt, skb, src, dst, tos,
@@ -1849,6 +1895,7 @@ static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
const struct iphdr *old_iph;
struct flowi4 fl4;
union vxlan_addr *dst;
+ struct vxlan_metadata md;
__be16 src_port = 0, dst_port;
u32 vni;
__be16 df = 0;
@@ -1919,11 +1966,12 @@ static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
+ md.vni = htonl(vni << 8);
+ md.gbp = skb->mark;
err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb,
fl4.saddr, dst->sin.sin_addr.s_addr,
- tos, ttl, df, src_port, dst_port,
- htonl(vni << 8),
+ tos, ttl, df, src_port, dst_port, &md,
!net_eq(vxlan->net, dev_net(vxlan->dev)));
if (err < 0) {
/* skb is already freed. */
@@ -1976,10 +2024,12 @@ static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
}
ttl = ttl ? : ip6_dst_hoplimit(ndst);
+ md.vni = htonl(vni << 8);
+ md.gbp = skb->mark;
err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb,
dev, &fl6.saddr, &fl6.daddr, 0, ttl,
- src_port, dst_port, htonl(vni << 8),
+ src_port, dst_port, &md,
!net_eq(vxlan->net, dev_net(vxlan->dev)));
#endif
}
@@ -2382,6 +2432,7 @@ static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
[IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
[IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
+ [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
};
static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
@@ -2706,6 +2757,9 @@ static int vxlan_newlink(struct net *net, struct net_device *dev,
nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
vxlan->flags |= VXLAN_F_REMCSUM_RX;
+ if (data[IFLA_VXLAN_GBP])
+ vxlan->flags |= VXLAN_F_GBP;
+
if (vxlan_find_vni(net, vni, use_ipv6 ? AF_INET6 : AF_INET,
vxlan->dst_port)) {
pr_info("duplicate VNI %u\n", vni);
@@ -2851,6 +2905,10 @@ static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
goto nla_put_failure;
+ if (vxlan->flags & VXLAN_F_GBP &&
+ nla_put_flag(skb, IFLA_VXLAN_GBP))
+ goto nla_put_failure;
+
return 0;
nla_put_failure:
diff --git a/include/net/vxlan.h b/include/net/vxlan.h
index 0a7443b..f4a3583 100644
--- a/include/net/vxlan.h
+++ b/include/net/vxlan.h
@@ -11,15 +11,76 @@
#define VNI_HASH_BITS 10
#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
-/* VXLAN protocol header */
+/*
+ * VXLAN Group Based Policy Extension:
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * |1|-|-|-|1|-|-|-|R|D|R|R|A|R|R|R| Group Policy ID |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * | VXLAN Network Identifier (VNI) | Reserved |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *
+ * D = Don't Learn bit. When set, this bit indicates that the egress
+ * VTEP MUST NOT learn the source address of the encapsulated frame.
+ *
+ * A = Indicates that the group policy has already been applied to
+ * this packet. Policies MUST NOT be applied by devices when the
+ * A bit is set.
+ *
+ * [0] https://tools.ietf.org/html/draft-smith-vxlan-group-policy
+ */
+struct vxlanhdr_gbp {
+ __u8 vx_flags;
+#ifdef __LITTLE_ENDIAN_BITFIELD
+ __u8 reserved_flags1:3,
+ policy_applied:1,
+ reserved_flags2:2,
+ dont_learn:1,
+ reserved_flags3:1;
+#elif defined(__BIG_ENDIAN_BITFIELD)
+ __u8 reserved_flags1:1,
+ dont_learn:1,
+ reserved_flags2:2,
+ policy_applied:1,
+ reserved_flags3:3;
+#else
+#error "Please fix <asm/byteorder.h>"
+#endif
+ __be16 policy_id;
+ __be32 vx_vni;
+};
+
+#define VXLAN_GBP_USED_BITS (VXLAN_HF_GBP | 0xFFFFFF)
+
+/* skb->mark mapping
+ *
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * |R|R|R|R|R|R|R|R|R|D|R|R|A|R|R|R| Group Policy ID |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ */
+#define VXLAN_GBP_DONT_LEARN (BIT(6) << 16)
+#define VXLAN_GBP_POLICY_APPLIED (BIT(3) << 16)
+#define VXLAN_GBP_ID_MASK (0xFFFF)
+
+/* VXLAN protocol header:
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * |G|R|R|R|I|R|R|C| Reserved |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * | VXLAN Network Identifier (VNI) | Reserved |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ *
+ * G = 1 Group Policy (VXLAN-GBP)
+ * I = 1 VXLAN Network Identifier (VNI) present
+ * C = 1 Remote checksum offload (RCO)
+ */
struct vxlanhdr {
__be32 vx_flags;
__be32 vx_vni;
};
/* VXLAN header flags. */
-#define VXLAN_HF_VNI 0x08000000
-#define VXLAN_HF_RCO 0x00200000
+#define VXLAN_HF_RCO BIT(24)
+#define VXLAN_HF_VNI BIT(27)
+#define VXLAN_HF_GBP BIT(31)
/* Remote checksum offload header option */
#define VXLAN_RCO_MASK 0x7f /* Last byte of vni field */
@@ -32,8 +93,14 @@ struct vxlanhdr {
#define VXLAN_VID_MASK (VXLAN_N_VID - 1)
#define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
+struct vxlan_metadata {
+ __be32 vni;
+ u32 gbp;
+};
+
struct vxlan_sock;
-typedef void (vxlan_rcv_t)(struct vxlan_sock *vh, struct sk_buff *skb, __be32 key);
+typedef void (vxlan_rcv_t)(struct vxlan_sock *vh, struct sk_buff *skb,
+ struct vxlan_metadata *md);
/* per UDP socket information */
struct vxlan_sock {
@@ -60,6 +127,7 @@ struct vxlan_sock {
#define VXLAN_F_UDP_ZERO_CSUM6_RX 0x100
#define VXLAN_F_REMCSUM_TX 0x200
#define VXLAN_F_REMCSUM_RX 0x400
+#define VXLAN_F_GBP 0x800
struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
vxlan_rcv_t *rcv, void *data,
@@ -70,7 +138,8 @@ void vxlan_sock_release(struct vxlan_sock *vs);
int vxlan_xmit_skb(struct vxlan_sock *vs,
struct rtable *rt, struct sk_buff *skb,
__be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
- __be16 src_port, __be16 dst_port, __be32 vni, bool xnet);
+ __be16 src_port, __be16 dst_port, struct vxlan_metadata *md,
+ bool xnet);
static inline netdev_features_t vxlan_features_check(struct sk_buff *skb,
netdev_features_t features)
diff --git a/include/uapi/linux/if_link.h b/include/uapi/linux/if_link.h
index b2723f6..2a8380e 100644
--- a/include/uapi/linux/if_link.h
+++ b/include/uapi/linux/if_link.h
@@ -372,6 +372,7 @@ enum {
IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
IFLA_VXLAN_REMCSUM_TX,
IFLA_VXLAN_REMCSUM_RX,
+ IFLA_VXLAN_GBP,
__IFLA_VXLAN_MAX
};
#define IFLA_VXLAN_MAX (__IFLA_VXLAN_MAX - 1)
diff --git a/net/openvswitch/vport-vxlan.c b/net/openvswitch/vport-vxlan.c
index d7c46b3..deed9e3 100644
--- a/net/openvswitch/vport-vxlan.c
+++ b/net/openvswitch/vport-vxlan.c
@@ -59,7 +59,8 @@ static inline struct vxlan_port *vxlan_vport(const struct vport *vport)
}
/* Called with rcu_read_lock and BH disabled. */
-static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb, __be32 vx_vni)
+static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
+ struct vxlan_metadata *md)
{
struct ovs_tunnel_info tun_info;
struct vport *vport = vs->data;
@@ -68,7 +69,7 @@ static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb, __be32 vx_vni)
/* Save outer tunnel values */
iph = ip_hdr(skb);
- key = cpu_to_be64(ntohl(vx_vni) >> 8);
+ key = cpu_to_be64(ntohl(md->vni) >> 8);
ovs_flow_tun_info_init(&tun_info, iph,
udp_hdr(skb)->source, udp_hdr(skb)->dest,
key, TUNNEL_KEY, NULL, 0);
@@ -146,6 +147,7 @@ static int vxlan_tnl_send(struct vport *vport, struct sk_buff *skb)
struct vxlan_port *vxlan_port = vxlan_vport(vport);
__be16 dst_port = inet_sk(vxlan_port->vs->sock->sk)->inet_sport;
struct ovs_key_ipv4_tunnel *tun_key;
+ struct vxlan_metadata md = {0};
struct rtable *rt;
struct flowi4 fl;
__be16 src_port;
@@ -178,12 +180,13 @@ static int vxlan_tnl_send(struct vport *vport, struct sk_buff *skb)
skb->ignore_df = 1;
src_port = udp_flow_src_port(net, skb, 0, 0, true);
+ md.vni = htonl(be64_to_cpu(tun_key->tun_id) << 8);
err = vxlan_xmit_skb(vxlan_port->vs, rt, skb,
fl.saddr, tun_key->ipv4_dst,
tun_key->ipv4_tos, tun_key->ipv4_ttl, df,
src_port, dst_port,
- htonl(be64_to_cpu(tun_key->tun_id) << 8),
+ &md,
false);
if (err < 0)
ip_rt_put(rt);
--
1.9.3
^ permalink raw reply related
* [PATCH 3/5] openvswitch: Rename GENEVE_TUN_OPTS() to TUN_METADATA_OPTS()
From: Thomas Graf @ 2015-01-15 2:53 UTC (permalink / raw)
To: davem, jesse, stephen, pshelar, therbert, alexei.starovoitov,
nicolas.dichtel
Cc: netdev, dev
In-Reply-To: <cover.1421290198.git.tgraf@suug.ch>
Also factors out Geneve validation code into a new separate function
validate_and_copy_geneve_opts().
A subsequent patch will introduce VXLAN options. Rename the existing
GENEVE_TUN_OPTS() to reflect its extended purpose of carrying generic
tunnel metadata options.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
---
v5->v6:
- No change
v4->v5:
- No change
v3->v4:
- Renamed validate_and_copy_geneve_opts() to validate_geneve_opts() as
suggested by Jesse
v2->v3:
- No change
v1->v2:
- Don't rename genev_tun_opt_from_nlattr() and keep it Geneve specific,
pointed out by Jesse.
- Factor out Geneve specific validation code into separate function as
requested by Jesse.
net/openvswitch/flow.c | 2 +-
net/openvswitch/flow.h | 14 ++++----
net/openvswitch/flow_netlink.c | 72 +++++++++++++++++++++++-------------------
3 files changed, 47 insertions(+), 41 deletions(-)
diff --git a/net/openvswitch/flow.c b/net/openvswitch/flow.c
index df334fe..e2c348b 100644
--- a/net/openvswitch/flow.c
+++ b/net/openvswitch/flow.c
@@ -691,7 +691,7 @@ int ovs_flow_key_extract(const struct ovs_tunnel_info *tun_info,
BUILD_BUG_ON((1 << (sizeof(tun_info->options_len) *
8)) - 1
> sizeof(key->tun_opts));
- memcpy(GENEVE_OPTS(key, tun_info->options_len),
+ memcpy(TUN_METADATA_OPTS(key, tun_info->options_len),
tun_info->options, tun_info->options_len);
key->tun_opts_len = tun_info->options_len;
} else {
diff --git a/net/openvswitch/flow.h b/net/openvswitch/flow.h
index a8b30f3..d3d0a40 100644
--- a/net/openvswitch/flow.h
+++ b/net/openvswitch/flow.h
@@ -53,7 +53,7 @@ struct ovs_key_ipv4_tunnel {
struct ovs_tunnel_info {
struct ovs_key_ipv4_tunnel tunnel;
- const struct geneve_opt *options;
+ const void *options;
u8 options_len;
};
@@ -61,10 +61,10 @@ struct ovs_tunnel_info {
* maximum size. This allows us to get the benefits of variable length
* matching for small options.
*/
-#define GENEVE_OPTS(flow_key, opt_len) \
- ((struct geneve_opt *)((flow_key)->tun_opts + \
- FIELD_SIZEOF(struct sw_flow_key, tun_opts) - \
- opt_len))
+#define TUN_METADATA_OFFSET(opt_len) \
+ (FIELD_SIZEOF(struct sw_flow_key, tun_opts) - opt_len)
+#define TUN_METADATA_OPTS(flow_key, opt_len) \
+ ((void *)((flow_key)->tun_opts + TUN_METADATA_OFFSET(opt_len)))
static inline void __ovs_flow_tun_info_init(struct ovs_tunnel_info *tun_info,
__be32 saddr, __be32 daddr,
@@ -73,7 +73,7 @@ static inline void __ovs_flow_tun_info_init(struct ovs_tunnel_info *tun_info,
__be16 tp_dst,
__be64 tun_id,
__be16 tun_flags,
- const struct geneve_opt *opts,
+ const void *opts,
u8 opts_len)
{
tun_info->tunnel.tun_id = tun_id;
@@ -105,7 +105,7 @@ static inline void ovs_flow_tun_info_init(struct ovs_tunnel_info *tun_info,
__be16 tp_dst,
__be64 tun_id,
__be16 tun_flags,
- const struct geneve_opt *opts,
+ const void *opts,
u8 opts_len)
{
__ovs_flow_tun_info_init(tun_info, iph->saddr, iph->daddr,
diff --git a/net/openvswitch/flow_netlink.c b/net/openvswitch/flow_netlink.c
index d1eecf7..2e8a9cd 100644
--- a/net/openvswitch/flow_netlink.c
+++ b/net/openvswitch/flow_netlink.c
@@ -432,8 +432,7 @@ static int genev_tun_opt_from_nlattr(const struct nlattr *a,
SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
}
- opt_key_offset = (unsigned long)GENEVE_OPTS((struct sw_flow_key *)0,
- nla_len(a));
+ opt_key_offset = TUN_METADATA_OFFSET(nla_len(a));
SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, nla_data(a),
nla_len(a), is_mask);
return 0;
@@ -558,8 +557,7 @@ static int ipv4_tun_from_nlattr(const struct nlattr *attr,
static int __ipv4_tun_to_nlattr(struct sk_buff *skb,
const struct ovs_key_ipv4_tunnel *output,
- const struct geneve_opt *tun_opts,
- int swkey_tun_opts_len)
+ const void *tun_opts, int swkey_tun_opts_len)
{
if (output->tun_flags & TUNNEL_KEY &&
nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
@@ -600,8 +598,7 @@ static int __ipv4_tun_to_nlattr(struct sk_buff *skb,
static int ipv4_tun_to_nlattr(struct sk_buff *skb,
const struct ovs_key_ipv4_tunnel *output,
- const struct geneve_opt *tun_opts,
- int swkey_tun_opts_len)
+ const void *tun_opts, int swkey_tun_opts_len)
{
struct nlattr *nla;
int err;
@@ -1148,10 +1145,10 @@ int ovs_nla_put_flow(const struct sw_flow_key *swkey,
goto nla_put_failure;
if ((swkey->tun_key.ipv4_dst || is_mask)) {
- const struct geneve_opt *opts = NULL;
+ const void *opts = NULL;
if (output->tun_key.tun_flags & TUNNEL_OPTIONS_PRESENT)
- opts = GENEVE_OPTS(output, swkey->tun_opts_len);
+ opts = TUN_METADATA_OPTS(output, swkey->tun_opts_len);
if (ipv4_tun_to_nlattr(skb, &output->tun_key, opts,
swkey->tun_opts_len))
@@ -1540,6 +1537,34 @@ void ovs_match_init(struct sw_flow_match *match,
}
}
+static int validate_geneve_opts(struct sw_flow_key *key)
+{
+ struct geneve_opt *option;
+ int opts_len = key->tun_opts_len;
+ bool crit_opt = false;
+
+ option = (struct geneve_opt *)TUN_METADATA_OPTS(key, key->tun_opts_len);
+ while (opts_len > 0) {
+ int len;
+
+ if (opts_len < sizeof(*option))
+ return -EINVAL;
+
+ len = sizeof(*option) + option->length * 4;
+ if (len > opts_len)
+ return -EINVAL;
+
+ crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
+
+ option = (struct geneve_opt *)((u8 *)option + len);
+ opts_len -= len;
+ };
+
+ key->tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
+
+ return 0;
+}
+
static int validate_and_copy_set_tun(const struct nlattr *attr,
struct sw_flow_actions **sfa, bool log)
{
@@ -1555,28 +1580,9 @@ static int validate_and_copy_set_tun(const struct nlattr *attr,
return err;
if (key.tun_opts_len) {
- struct geneve_opt *option = GENEVE_OPTS(&key,
- key.tun_opts_len);
- int opts_len = key.tun_opts_len;
- bool crit_opt = false;
-
- while (opts_len > 0) {
- int len;
-
- if (opts_len < sizeof(*option))
- return -EINVAL;
-
- len = sizeof(*option) + option->length * 4;
- if (len > opts_len)
- return -EINVAL;
-
- crit_opt |= !!(option->type & GENEVE_CRIT_OPT_TYPE);
-
- option = (struct geneve_opt *)((u8 *)option + len);
- opts_len -= len;
- };
-
- key.tun_key.tun_flags |= crit_opt ? TUNNEL_CRIT_OPT : 0;
+ err = validate_geneve_opts(&key);
+ if (err < 0)
+ return err;
};
start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
@@ -1597,9 +1603,9 @@ static int validate_and_copy_set_tun(const struct nlattr *attr,
* everything else will go away after flow setup. We can append
* it to tun_info and then point there.
*/
- memcpy((tun_info + 1), GENEVE_OPTS(&key, key.tun_opts_len),
- key.tun_opts_len);
- tun_info->options = (struct geneve_opt *)(tun_info + 1);
+ memcpy((tun_info + 1),
+ TUN_METADATA_OPTS(&key, key.tun_opts_len), key.tun_opts_len);
+ tun_info->options = (tun_info + 1);
} else {
tun_info->options = NULL;
}
--
1.9.3
^ permalink raw reply related
* [PATCH 0/5 net-next v6] VXLAN Group Policy Extension
From: Thomas Graf @ 2015-01-15 2:53 UTC (permalink / raw)
To: davem, jesse, stephen, pshelar, therbert, alexei.starovoitov,
nicolas.dichtel
Cc: netdev, dev
Implements supports for the Group Policy VXLAN extension [0] to provide
a lightweight and simple security label mechanism across network peers
based on VXLAN. The security context and associated metadata is mapped
to/from skb->mark. This allows further mapping to a SELinux context
using SECMARK, to implement ACLs directly with nftables, iptables, OVS,
tc, etc.
The extension is disabled by default and should be run on a distinct
port in mixed Linux VXLAN VTEP environments. Liberal VXLAN VTEPs
which ignore unknown reserved bits will be able to receive VXLAN-GBP
frames.
Simple usage example:
10.1.1.1:
# ip link add vxlan0 type vxlan id 10 remote 10.1.1.2 gbp
# iptables -I OUTPUT -m owner --uid-owner 101 -j MARK --set-mark 0x200
10.1.1.2:
# ip link add vxlan0 type vxlan id 10 remote 10.1.1.1 gbp
# iptables -I INPUT -m mark --mark 0x200 -j DROP
iproute2 [1] and OVS [2] support will be provided in separate patches.
[0] https://tools.ietf.org/html/draft-smith-vxlan-group-policy
[1] https://github.com/tgraf/iproute2/tree/vxlan-gbp
[2] https://github.com/tgraf/ovs/tree/vxlan-gbp
Thomas Graf (5):
vxlan: Group Policy extension
vxlan: Only bind to sockets with compatible flags enabled
openvswitch: Rename GENEVE_TUN_OPTS() to TUN_METADATA_OPTS()
openvswitch: Allow for any level of nesting in flow attributes
openvswitch: Support VXLAN Group Policy extension
drivers/net/vxlan.c | 113 ++++++++++++----
include/net/ip_tunnels.h | 5 +-
include/net/vxlan.h | 82 ++++++++++-
include/uapi/linux/if_link.h | 1 +
include/uapi/linux/openvswitch.h | 11 ++
net/openvswitch/flow.c | 2 +-
net/openvswitch/flow.h | 14 +-
net/openvswitch/flow_netlink.c | 286 ++++++++++++++++++++++++++-------------
net/openvswitch/vport-geneve.c | 15 +-
net/openvswitch/vport-vxlan.c | 91 ++++++++++++-
net/openvswitch/vport-vxlan.h | 11 ++
11 files changed, 491 insertions(+), 140 deletions(-)
create mode 100644 net/openvswitch/vport-vxlan.h
--
1.9.3
^ permalink raw reply
* [PATCH 2/5] vxlan: Only bind to sockets with compatible flags enabled
From: Thomas Graf @ 2015-01-15 2:53 UTC (permalink / raw)
To: davem, jesse, stephen, pshelar, therbert, alexei.starovoitov,
nicolas.dichtel
Cc: netdev, dev
In-Reply-To: <cover.1421290198.git.tgraf@suug.ch>
A VXLAN net_device looking for an appropriate socket may only consider
a socket which has a matching set of flags/extensions enabled. If
incompatible flags are enabled, return a conflict to have the caller
create a distinct socket with distinct port.
The OVS VXLAN port is kept unaware of extensions at this point.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
---
v5->v6:
- Keep sharing logic but base it off unsharable flags instead of exts
member as suggested by Tom
v4->v5:
- No change
v3->v4:
- No change
v2->v3:
- No change
v1->v2:
- Improved commit message, reported by Jesse
drivers/net/vxlan.c | 29 ++++++++++++++++++-----------
include/net/vxlan.h | 3 +++
2 files changed, 21 insertions(+), 11 deletions(-)
diff --git a/drivers/net/vxlan.c b/drivers/net/vxlan.c
index 6dbf8e0..6b6b456 100644
--- a/drivers/net/vxlan.c
+++ b/drivers/net/vxlan.c
@@ -263,15 +263,19 @@ static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
}
-/* Find VXLAN socket based on network namespace, address family and UDP port */
-static struct vxlan_sock *vxlan_find_sock(struct net *net,
- sa_family_t family, __be16 port)
+/* Find VXLAN socket based on network namespace, address family and UDP port
+ * and enabled unshareable flags.
+ */
+static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
+ __be16 port, u32 flags)
{
struct vxlan_sock *vs;
+ u32 match_flags = flags & VXLAN_F_UNSHAREABLE;
hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
if (inet_sk(vs->sock->sk)->inet_sport == port &&
- inet_sk(vs->sock->sk)->sk.sk_family == family)
+ inet_sk(vs->sock->sk)->sk.sk_family == family &&
+ (vs->flags & VXLAN_F_UNSHAREABLE) == match_flags)
return vs;
}
return NULL;
@@ -291,11 +295,12 @@ static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
/* Look up VNI in a per net namespace table */
static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id,
- sa_family_t family, __be16 port)
+ sa_family_t family, __be16 port,
+ u32 flags)
{
struct vxlan_sock *vs;
- vs = vxlan_find_sock(net, family, port);
+ vs = vxlan_find_sock(net, family, port, flags);
if (!vs)
return NULL;
@@ -1957,7 +1962,8 @@ static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
ip_rt_put(rt);
dst_vxlan = vxlan_find_vni(vxlan->net, vni,
- dst->sa.sa_family, dst_port);
+ dst->sa.sa_family, dst_port,
+ vxlan->flags);
if (!dst_vxlan)
goto tx_error;
vxlan_encap_bypass(skb, vxlan, dst_vxlan);
@@ -2016,7 +2022,8 @@ static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
dst_release(ndst);
dst_vxlan = vxlan_find_vni(vxlan->net, vni,
- dst->sa.sa_family, dst_port);
+ dst->sa.sa_family, dst_port,
+ vxlan->flags);
if (!dst_vxlan)
goto tx_error;
vxlan_encap_bypass(skb, vxlan, dst_vxlan);
@@ -2186,7 +2193,7 @@ static int vxlan_init(struct net_device *dev)
spin_lock(&vn->sock_lock);
vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
- vxlan->dst_port);
+ vxlan->dst_port, vxlan->flags);
if (vs && atomic_add_unless(&vs->refcnt, 1, 0)) {
/* If we have a socket with same port already, reuse it */
vxlan_vs_add_dev(vs, vxlan);
@@ -2593,7 +2600,7 @@ struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
return vs;
spin_lock(&vn->sock_lock);
- vs = vxlan_find_sock(net, ipv6 ? AF_INET6 : AF_INET, port);
+ vs = vxlan_find_sock(net, ipv6 ? AF_INET6 : AF_INET, port, flags);
if (vs && ((vs->rcv != rcv) ||
!atomic_add_unless(&vs->refcnt, 1, 0)))
vs = ERR_PTR(-EBUSY);
@@ -2761,7 +2768,7 @@ static int vxlan_newlink(struct net *net, struct net_device *dev,
vxlan->flags |= VXLAN_F_GBP;
if (vxlan_find_vni(net, vni, use_ipv6 ? AF_INET6 : AF_INET,
- vxlan->dst_port)) {
+ vxlan->dst_port, vxlan->flags)) {
pr_info("duplicate VNI %u\n", vni);
return -EEXIST;
}
diff --git a/include/net/vxlan.h b/include/net/vxlan.h
index f4a3583..7be8c34 100644
--- a/include/net/vxlan.h
+++ b/include/net/vxlan.h
@@ -129,6 +129,9 @@ struct vxlan_sock {
#define VXLAN_F_REMCSUM_RX 0x400
#define VXLAN_F_GBP 0x800
+/* These flags must match in order for a socket to be shareable */
+#define VXLAN_F_UNSHAREABLE VXLAN_F_GBP
+
struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
vxlan_rcv_t *rcv, void *data,
bool no_share, u32 flags);
--
1.9.3
^ permalink raw reply related
* [PATCH 5/5] openvswitch: Support VXLAN Group Policy extension
From: Thomas Graf @ 2015-01-15 2:53 UTC (permalink / raw)
To: davem, jesse, stephen, pshelar, therbert, alexei.starovoitov,
nicolas.dichtel
Cc: netdev, dev
In-Reply-To: <cover.1421290198.git.tgraf@suug.ch>
Introduces support for the group policy extension to the VXLAN virtual
port. The extension is disabled by default and only enabled if the user
has provided the respective configuration.
ovs-vsctl add-port br0 vxlan0 -- \
set Interface vxlan0 type=vxlan options:exts=gbp
The configuration interface to enable the extension is based on a new
attribute OVS_VXLAN_EXT_GBP nested inside OVS_TUNNEL_ATTR_EXTENSION
which can carry additional extensions as needed in the future.
The group policy metadata is stored as binary blob (struct ovs_vxlan_opts)
internally just like Geneve options but transported as nested Netlink
attributes to user space.
Renames the existing TUNNEL_OPTIONS_PRESENT to TUNNEL_GENEVE_OPT with the
binary value kept intact, a new flag TUNNEL_VXLAN_OPT is introduced.
The attributes OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS and existing
OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS are implemented mutually exclusive.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
---
v5->v6:
- No change
v4->v5:
- No change
v3->v4:
- Fixed OVS_VXLAN_EXT_MAX->OVS_VXLAN_EXT_GBP typo as spotted by Jesse
- Only applied tunnel options if they are of the right type as
suggested by Jesse
v2->v3:
- No change
v1->v2:
- Addressed Jesse's request to transport VXLAN options as Netlink
attributes instead of a binary blob. Allows a partial transport of
VXLAN extensions. Internally, the datapath continues to use a binary
blob (defined in vport-vxlan.h) for performance reasons.
- Added new TUNNEL_GENEVE_OPT and TUNNEL_VXLAN_OPT flags to mark
tunnel option flavour
- Correctly report VXLAN options to user space
include/net/ip_tunnels.h | 5 +-
include/uapi/linux/openvswitch.h | 11 ++++
net/openvswitch/flow_netlink.c | 114 ++++++++++++++++++++++++++++++++++-----
net/openvswitch/vport-geneve.c | 15 ++++--
net/openvswitch/vport-vxlan.c | 82 +++++++++++++++++++++++++++-
net/openvswitch/vport-vxlan.h | 11 ++++
6 files changed, 218 insertions(+), 20 deletions(-)
create mode 100644 net/openvswitch/vport-vxlan.h
diff --git a/include/net/ip_tunnels.h b/include/net/ip_tunnels.h
index 25a59eb..ce4db3c 100644
--- a/include/net/ip_tunnels.h
+++ b/include/net/ip_tunnels.h
@@ -97,7 +97,10 @@ struct ip_tunnel {
#define TUNNEL_DONT_FRAGMENT __cpu_to_be16(0x0100)
#define TUNNEL_OAM __cpu_to_be16(0x0200)
#define TUNNEL_CRIT_OPT __cpu_to_be16(0x0400)
-#define TUNNEL_OPTIONS_PRESENT __cpu_to_be16(0x0800)
+#define TUNNEL_GENEVE_OPT __cpu_to_be16(0x0800)
+#define TUNNEL_VXLAN_OPT __cpu_to_be16(0x1000)
+
+#define TUNNEL_OPTIONS_PRESENT (TUNNEL_GENEVE_OPT | TUNNEL_VXLAN_OPT)
struct tnl_ptk_info {
__be16 flags;
diff --git a/include/uapi/linux/openvswitch.h b/include/uapi/linux/openvswitch.h
index 3a6dcaa..e474c95 100644
--- a/include/uapi/linux/openvswitch.h
+++ b/include/uapi/linux/openvswitch.h
@@ -248,11 +248,21 @@ enum ovs_vport_attr {
#define OVS_VPORT_ATTR_MAX (__OVS_VPORT_ATTR_MAX - 1)
+enum {
+ OVS_VXLAN_EXT_UNSPEC,
+ OVS_VXLAN_EXT_GBP, /* Flag or __u32 */
+ __OVS_VXLAN_EXT_MAX,
+};
+
+#define OVS_VXLAN_EXT_MAX (__OVS_VXLAN_EXT_MAX - 1)
+
+
/* OVS_VPORT_ATTR_OPTIONS attributes for tunnels.
*/
enum {
OVS_TUNNEL_ATTR_UNSPEC,
OVS_TUNNEL_ATTR_DST_PORT, /* 16-bit UDP port, used by L4 tunnels. */
+ OVS_TUNNEL_ATTR_EXTENSION,
__OVS_TUNNEL_ATTR_MAX
};
@@ -324,6 +334,7 @@ enum ovs_tunnel_key_attr {
OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS, /* Array of Geneve options. */
OVS_TUNNEL_KEY_ATTR_TP_SRC, /* be16 src Transport Port. */
OVS_TUNNEL_KEY_ATTR_TP_DST, /* be16 dst Transport Port. */
+ OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS, /* Nested OVS_VXLAN_EXT_* */
__OVS_TUNNEL_KEY_ATTR_MAX
};
diff --git a/net/openvswitch/flow_netlink.c b/net/openvswitch/flow_netlink.c
index 518941c..d210d1b 100644
--- a/net/openvswitch/flow_netlink.c
+++ b/net/openvswitch/flow_netlink.c
@@ -49,6 +49,7 @@
#include <net/mpls.h>
#include "flow_netlink.h"
+#include "vport-vxlan.h"
struct ovs_len_tbl {
int len;
@@ -268,6 +269,9 @@ size_t ovs_tun_key_attr_size(void)
+ nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_CSUM */
+ nla_total_size(0) /* OVS_TUNNEL_KEY_ATTR_OAM */
+ nla_total_size(256) /* OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS */
+ /* OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS is mutually exclusive with
+ * OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS and covered by it.
+ */
+ nla_total_size(2) /* OVS_TUNNEL_KEY_ATTR_TP_SRC */
+ nla_total_size(2); /* OVS_TUNNEL_KEY_ATTR_TP_DST */
}
@@ -308,6 +312,7 @@ static const struct ovs_len_tbl ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1]
[OVS_TUNNEL_KEY_ATTR_TP_DST] = { .len = sizeof(u16) },
[OVS_TUNNEL_KEY_ATTR_OAM] = { .len = 0 },
[OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS] = { .len = OVS_ATTR_NESTED },
+ [OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS] = { .len = OVS_ATTR_NESTED },
};
/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
@@ -460,6 +465,41 @@ static int genev_tun_opt_from_nlattr(const struct nlattr *a,
return 0;
}
+static const struct nla_policy vxlan_opt_policy[OVS_VXLAN_EXT_MAX + 1] = {
+ [OVS_VXLAN_EXT_GBP] = { .type = NLA_U32 },
+};
+
+static int vxlan_tun_opt_from_nlattr(const struct nlattr *a,
+ struct sw_flow_match *match, bool is_mask,
+ bool log)
+{
+ struct nlattr *tb[OVS_VXLAN_EXT_MAX+1];
+ unsigned long opt_key_offset;
+ struct ovs_vxlan_opts opts;
+ int err;
+
+ BUILD_BUG_ON(sizeof(opts) > sizeof(match->key->tun_opts));
+
+ err = nla_parse_nested(tb, OVS_VXLAN_EXT_MAX, a, vxlan_opt_policy);
+ if (err < 0)
+ return err;
+
+ memset(&opts, 0, sizeof(opts));
+
+ if (tb[OVS_VXLAN_EXT_GBP])
+ opts.gbp = nla_get_u32(tb[OVS_VXLAN_EXT_GBP]);
+
+ if (!is_mask)
+ SW_FLOW_KEY_PUT(match, tun_opts_len, sizeof(opts), false);
+ else
+ SW_FLOW_KEY_PUT(match, tun_opts_len, 0xff, true);
+
+ opt_key_offset = TUN_METADATA_OFFSET(sizeof(opts));
+ SW_FLOW_KEY_MEMCPY_OFFSET(match, opt_key_offset, &opts, sizeof(opts),
+ is_mask);
+ return 0;
+}
+
static int ipv4_tun_from_nlattr(const struct nlattr *attr,
struct sw_flow_match *match, bool is_mask,
bool log)
@@ -468,6 +508,7 @@ static int ipv4_tun_from_nlattr(const struct nlattr *attr,
int rem;
bool ttl = false;
__be16 tun_flags = 0;
+ int opts_type = 0;
nla_for_each_nested(a, attr, rem) {
int type = nla_type(a);
@@ -527,11 +568,30 @@ static int ipv4_tun_from_nlattr(const struct nlattr *attr,
tun_flags |= TUNNEL_OAM;
break;
case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
+ if (opts_type) {
+ OVS_NLERR(log, "Multiple metadata blocks provided");
+ return -EINVAL;
+ }
+
err = genev_tun_opt_from_nlattr(a, match, is_mask, log);
if (err)
return err;
- tun_flags |= TUNNEL_OPTIONS_PRESENT;
+ tun_flags |= TUNNEL_GENEVE_OPT;
+ opts_type = type;
+ break;
+ case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
+ if (opts_type) {
+ OVS_NLERR(log, "Multiple metadata blocks provided");
+ return -EINVAL;
+ }
+
+ err = vxlan_tun_opt_from_nlattr(a, match, is_mask, log);
+ if (err)
+ return err;
+
+ tun_flags |= TUNNEL_VXLAN_OPT;
+ opts_type = type;
break;
default:
OVS_NLERR(log, "Unknown IPv4 tunnel attribute %d",
@@ -560,6 +620,23 @@ static int ipv4_tun_from_nlattr(const struct nlattr *attr,
}
}
+ return opts_type;
+}
+
+static int vxlan_opt_to_nlattr(struct sk_buff *skb,
+ const void *tun_opts, int swkey_tun_opts_len)
+{
+ const struct ovs_vxlan_opts *opts = tun_opts;
+ struct nlattr *nla;
+
+ nla = nla_nest_start(skb, OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS);
+ if (!nla)
+ return -EMSGSIZE;
+
+ if (nla_put_u32(skb, OVS_VXLAN_EXT_GBP, opts->gbp) < 0)
+ return -EMSGSIZE;
+
+ nla_nest_end(skb, nla);
return 0;
}
@@ -596,10 +673,15 @@ static int __ipv4_tun_to_nlattr(struct sk_buff *skb,
if ((output->tun_flags & TUNNEL_OAM) &&
nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_OAM))
return -EMSGSIZE;
- if (tun_opts &&
- nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
- swkey_tun_opts_len, tun_opts))
- return -EMSGSIZE;
+ if (tun_opts) {
+ if (output->tun_flags & TUNNEL_GENEVE_OPT &&
+ nla_put(skb, OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS,
+ swkey_tun_opts_len, tun_opts))
+ return -EMSGSIZE;
+ else if (output->tun_flags & TUNNEL_VXLAN_OPT &&
+ vxlan_opt_to_nlattr(skb, tun_opts, swkey_tun_opts_len))
+ return -EMSGSIZE;
+ }
return 0;
}
@@ -680,7 +762,7 @@ static int metadata_from_nlattrs(struct sw_flow_match *match, u64 *attrs,
}
if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
- is_mask, log))
+ is_mask, log) < 0)
return -EINVAL;
*attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
}
@@ -1578,17 +1660,23 @@ static int validate_and_copy_set_tun(const struct nlattr *attr,
struct sw_flow_key key;
struct ovs_tunnel_info *tun_info;
struct nlattr *a;
- int err, start;
+ int err, start, opts_type;
ovs_match_init(&match, &key, NULL);
- err = ipv4_tun_from_nlattr(nla_data(attr), &match, false, log);
- if (err)
- return err;
+ opts_type = ipv4_tun_from_nlattr(nla_data(attr), &match, false, log);
+ if (opts_type < 0)
+ return opts_type;
if (key.tun_opts_len) {
- err = validate_geneve_opts(&key);
- if (err < 0)
- return err;
+ switch (opts_type) {
+ case OVS_TUNNEL_KEY_ATTR_GENEVE_OPTS:
+ err = validate_geneve_opts(&key);
+ if (err < 0)
+ return err;
+ break;
+ case OVS_TUNNEL_KEY_ATTR_VXLAN_OPTS:
+ break;
+ }
};
start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET, log);
diff --git a/net/openvswitch/vport-geneve.c b/net/openvswitch/vport-geneve.c
index 2daf144..17b0840 100644
--- a/net/openvswitch/vport-geneve.c
+++ b/net/openvswitch/vport-geneve.c
@@ -88,7 +88,7 @@ static void geneve_rcv(struct geneve_sock *gs, struct sk_buff *skb)
opts_len = geneveh->opt_len * 4;
- flags = TUNNEL_KEY | TUNNEL_OPTIONS_PRESENT |
+ flags = TUNNEL_KEY | TUNNEL_GENEVE_OPT |
(udp_hdr(skb)->check != 0 ? TUNNEL_CSUM : 0) |
(geneveh->oam ? TUNNEL_OAM : 0) |
(geneveh->critical ? TUNNEL_CRIT_OPT : 0);
@@ -178,7 +178,7 @@ static int geneve_tnl_send(struct vport *vport, struct sk_buff *skb)
__be16 sport;
struct rtable *rt;
struct flowi4 fl;
- u8 vni[3];
+ u8 vni[3], opts_len, *opts;
__be16 df;
int err;
@@ -209,11 +209,18 @@ static int geneve_tnl_send(struct vport *vport, struct sk_buff *skb)
tunnel_id_to_vni(tun_key->tun_id, vni);
skb->ignore_df = 1;
+ if (tun_key->tun_flags & TUNNEL_GENEVE_OPT) {
+ opts = (u8 *)tun_info->options;
+ opts_len = tun_info->options_len;
+ } else {
+ opts = NULL;
+ opts_len = 0;
+ }
+
err = geneve_xmit_skb(geneve_port->gs, rt, skb, fl.saddr,
tun_key->ipv4_dst, tun_key->ipv4_tos,
tun_key->ipv4_ttl, df, sport, dport,
- tun_key->tun_flags, vni,
- tun_info->options_len, (u8 *)tun_info->options,
+ tun_key->tun_flags, vni, opts_len, opts,
false);
if (err < 0)
ip_rt_put(rt);
diff --git a/net/openvswitch/vport-vxlan.c b/net/openvswitch/vport-vxlan.c
index deed9e3..0beddd0 100644
--- a/net/openvswitch/vport-vxlan.c
+++ b/net/openvswitch/vport-vxlan.c
@@ -40,6 +40,7 @@
#include "datapath.h"
#include "vport.h"
+#include "vport-vxlan.h"
/**
* struct vxlan_port - Keeps track of open UDP ports
@@ -49,6 +50,7 @@
struct vxlan_port {
struct vxlan_sock *vs;
char name[IFNAMSIZ];
+ u32 exts; /* VXLAN_F_* in <net/vxlan.h> */
};
static struct vport_ops ovs_vxlan_vport_ops;
@@ -63,16 +65,26 @@ static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
struct vxlan_metadata *md)
{
struct ovs_tunnel_info tun_info;
+ struct vxlan_port *vxlan_port;
struct vport *vport = vs->data;
struct iphdr *iph;
+ struct ovs_vxlan_opts opts = {
+ .gbp = md->gbp,
+ };
__be64 key;
+ __be16 flags;
+
+ flags = TUNNEL_KEY;
+ vxlan_port = vxlan_vport(vport);
+ if (vxlan_port->exts & VXLAN_F_GBP)
+ flags |= TUNNEL_VXLAN_OPT;
/* Save outer tunnel values */
iph = ip_hdr(skb);
key = cpu_to_be64(ntohl(md->vni) >> 8);
ovs_flow_tun_info_init(&tun_info, iph,
udp_hdr(skb)->source, udp_hdr(skb)->dest,
- key, TUNNEL_KEY, NULL, 0);
+ key, flags, &opts, sizeof(opts));
ovs_vport_receive(vport, skb, &tun_info);
}
@@ -84,6 +96,21 @@ static int vxlan_get_options(const struct vport *vport, struct sk_buff *skb)
if (nla_put_u16(skb, OVS_TUNNEL_ATTR_DST_PORT, ntohs(dst_port)))
return -EMSGSIZE;
+
+ if (vxlan_port->exts) {
+ struct nlattr *exts;
+
+ exts = nla_nest_start(skb, OVS_TUNNEL_ATTR_EXTENSION);
+ if (!exts)
+ return -EMSGSIZE;
+
+ if (vxlan_port->exts & VXLAN_F_GBP &&
+ nla_put_flag(skb, OVS_VXLAN_EXT_GBP))
+ return -EMSGSIZE;
+
+ nla_nest_end(skb, exts);
+ }
+
return 0;
}
@@ -96,6 +123,31 @@ static void vxlan_tnl_destroy(struct vport *vport)
ovs_vport_deferred_free(vport);
}
+static const struct nla_policy exts_policy[OVS_VXLAN_EXT_MAX+1] = {
+ [OVS_VXLAN_EXT_GBP] = { .type = NLA_FLAG, },
+};
+
+static int vxlan_configure_exts(struct vport *vport, struct nlattr *attr)
+{
+ struct nlattr *exts[OVS_VXLAN_EXT_MAX+1];
+ struct vxlan_port *vxlan_port;
+ int err;
+
+ if (nla_len(attr) < sizeof(struct nlattr))
+ return -EINVAL;
+
+ err = nla_parse_nested(exts, OVS_VXLAN_EXT_MAX, attr, exts_policy);
+ if (err < 0)
+ return err;
+
+ vxlan_port = vxlan_vport(vport);
+
+ if (exts[OVS_VXLAN_EXT_GBP])
+ vxlan_port->exts |= VXLAN_F_GBP;
+
+ return 0;
+}
+
static struct vport *vxlan_tnl_create(const struct vport_parms *parms)
{
struct net *net = ovs_dp_get_net(parms->dp);
@@ -128,7 +180,17 @@ static struct vport *vxlan_tnl_create(const struct vport_parms *parms)
vxlan_port = vxlan_vport(vport);
strncpy(vxlan_port->name, parms->name, IFNAMSIZ);
- vs = vxlan_sock_add(net, htons(dst_port), vxlan_rcv, vport, true, 0);
+ a = nla_find_nested(options, OVS_TUNNEL_ATTR_EXTENSION);
+ if (a) {
+ err = vxlan_configure_exts(vport, a);
+ if (err) {
+ ovs_vport_free(vport);
+ goto error;
+ }
+ }
+
+ vs = vxlan_sock_add(net, htons(dst_port), vxlan_rcv, vport, true,
+ vxlan_port->exts);
if (IS_ERR(vs)) {
ovs_vport_free(vport);
return (void *)vs;
@@ -141,6 +203,21 @@ error:
return ERR_PTR(err);
}
+static int vxlan_ext_gbp(struct sk_buff *skb)
+{
+ const struct ovs_tunnel_info *tun_info;
+ const struct ovs_vxlan_opts *opts;
+
+ tun_info = OVS_CB(skb)->egress_tun_info;
+ opts = tun_info->options;
+
+ if (tun_info->tunnel.tun_flags & TUNNEL_VXLAN_OPT &&
+ tun_info->options_len >= sizeof(*opts))
+ return opts->gbp;
+ else
+ return 0;
+}
+
static int vxlan_tnl_send(struct vport *vport, struct sk_buff *skb)
{
struct net *net = ovs_dp_get_net(vport->dp);
@@ -181,6 +258,7 @@ static int vxlan_tnl_send(struct vport *vport, struct sk_buff *skb)
src_port = udp_flow_src_port(net, skb, 0, 0, true);
md.vni = htonl(be64_to_cpu(tun_key->tun_id) << 8);
+ md.gbp = vxlan_ext_gbp(skb);
err = vxlan_xmit_skb(vxlan_port->vs, rt, skb,
fl.saddr, tun_key->ipv4_dst,
diff --git a/net/openvswitch/vport-vxlan.h b/net/openvswitch/vport-vxlan.h
new file mode 100644
index 0000000..4b08233e
--- /dev/null
+++ b/net/openvswitch/vport-vxlan.h
@@ -0,0 +1,11 @@
+#ifndef VPORT_VXLAN_H
+#define VPORT_VXLAN_H 1
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+
+struct ovs_vxlan_opts {
+ __u32 gbp;
+};
+
+#endif
--
1.9.3
^ permalink raw reply related
* Re: [PATCH net-next v13 3/3] net: hisilicon: new hip04 ethernet driver
From: Ding Tianhong @ 2015-01-15 2:54 UTC (permalink / raw)
To: Arnd Bergmann
Cc: robh+dt, davem, grant.likely, agraf, sergei.shtylyov,
linux-arm-kernel, eric.dumazet, xuwei5, zhangfei.gao, netdev,
devicetree, linux
In-Reply-To: <3131780.HetDHI4Cfl@wuerfel>
On 2015/1/14 16:53, Arnd Bergmann wrote:
> On Wednesday 14 January 2015 14:34:14 Ding Tianhong wrote:
>> +#define HIP04_MAX_TX_COALESCE_USECS 200
>> +#define HIP04_MIN_TX_COALESCE_USECS 100
>> +#define HIP04_MAX_TX_COALESCE_FRAMES 200
>> +#define HIP04_MIN_TX_COALESCE_FRAMES 100
>
> It's not important, but in case you are creating another version of the
> patch, maybe the allowed range can be extended somewhat. The example values
> I picked when I sent my suggestion were really made up. It's great if
> they work fine, but users might want to tune this far more depending on
> their workloads, How about these
>
> #define HIP04_MAX_TX_COALESCE_USECS 100000
> #define HIP04_MIN_TX_COALESCE_USECS 1
> #define HIP04_MAX_TX_COALESCE_FRAMES (TX_DESC_NUM - 1)
> #define HIP04_MIN_TX_COALESCE_FRAMES 1
>
Is it really ok that the so big range may break the driver and hip04 could not work fine?
I am not sure it is ok, I will fix it in next version.
Ding
> Arnd
>
> .
>
^ permalink raw reply
* Re: [PATCH 1/5] vxlan: Group Policy extension
From: Thomas Graf @ 2015-01-15 2:55 UTC (permalink / raw)
To: Tom Herbert
Cc: David Miller, Jesse Gross, Stephen Hemminger, Pravin B Shelar,
Alexei Starovoitov, Nicolas Dichtel, Linux Netdev List,
dev@openvswitch.org
In-Reply-To: <20150115012804.GI2105@casper.infradead.org>
On 01/15/15 at 01:28am, Thomas Graf wrote:
> What exactly is the problem of having a distinct bitmap used by
> extensions? It is the least error prone method because it's clear that
> all extensions must match and we don't have to maintain an additional
> bitmask which can be forgotten to be updated.
>
> If you need to compare additional receive checksum settings for RCO
> then that should be separate because as you say it's not an extension.
Tom,
OK. I have changed it to use flags instead of exts in v6. You should
be able to add whatever RCO flags need to be matched in
vxlan_find_sock() to the VXLAN_F_UNSHAREABLE bitmask.
Hope this makes everybody happy ;-)
^ permalink raw reply
* Re: [PATCH net-next v13 3/3] net: hisilicon: new hip04 ethernet driver
From: Ding Tianhong @ 2015-01-15 2:56 UTC (permalink / raw)
To: Joe Perches
Cc: arnd, robh+dt, davem, grant.likely, agraf, sergei.shtylyov,
linux-arm-kernel, eric.dumazet, xuwei5, zhangfei.gao, netdev,
devicetree, linux
In-Reply-To: <1421222782.2847.1.camel@perches.com>
On 2015/1/14 16:06, Joe Perches wrote:
> On Wed, 2015-01-14 at 14:34 +0800, Ding Tianhong wrote:
>> Support Hisilicon hip04 ethernet driver, including 100M / 1000M controller.
>> The controller has no tx done interrupt, reclaim xmitted buffer in the poll.
>
> Single comment:
>
>> diff --git a/drivers/net/ethernet/hisilicon/hip04_eth.c b/drivers/net/ethernet/hisilicon/hip04_eth.c
> []
>> +static int hip04_rx_poll(struct napi_struct *napi, int budget)
>> +{
> []
>> + while (cnt && !last) {
>> + buf = priv->rx_buf[priv->rx_head];
>> + skb = build_skb(buf, priv->rx_buf_size);
>> + if (unlikely(!skb))
>> + net_dbg_ratelimited("build_skb failed\n");
>> +
>> + dma_unmap_single(&ndev->dev, priv->rx_phys[priv->rx_head],
>> + RX_BUF_SIZE, DMA_FROM_DEVICE);
>> + priv->rx_phys[priv->rx_head] = 0;
>> +
>> + desc = (struct rx_desc *)skb->data;
>
> Perhaps if (!skb) is possible, there shouldn't be
> a dereference of that known null here.
>
Yes, we should return and do something to avoid oops, thanks.
Ding
>
>
> .
>
^ permalink raw reply
* Re: [PATCH 1/5] vxlan: Group Policy extension
From: Tom Herbert @ 2015-01-15 2:59 UTC (permalink / raw)
To: Thomas Graf
Cc: David Miller, Jesse Gross, Stephen Hemminger, Pravin B Shelar,
Alexei Starovoitov, Nicolas Dichtel, Linux Netdev List,
dev@openvswitch.org
In-Reply-To: <20150115025539.GA20315@casper.infradead.org>
On Wed, Jan 14, 2015 at 6:55 PM, Thomas Graf <tgraf@suug.ch> wrote:
> On 01/15/15 at 01:28am, Thomas Graf wrote:
>> What exactly is the problem of having a distinct bitmap used by
>> extensions? It is the least error prone method because it's clear that
>> all extensions must match and we don't have to maintain an additional
>> bitmask which can be forgotten to be updated.
>>
>> If you need to compare additional receive checksum settings for RCO
>> then that should be separate because as you say it's not an extension.
>
> Tom,
>
> OK. I have changed it to use flags instead of exts in v6. You should
> be able to add whatever RCO flags need to be matched in
> vxlan_find_sock() to the VXLAN_F_UNSHAREABLE bitmask.
>
> Hope this makes everybody happy ;-)
Awesome. Thanks!
^ permalink raw reply
* Re: [PATCH 1/5] vxlan: Group Policy extension
From: Tom Herbert @ 2015-01-15 3:06 UTC (permalink / raw)
To: Thomas Graf
Cc: David Miller, Jesse Gross, Stephen Hemminger, Pravin B Shelar,
Alexei Starovoitov, Nicolas Dichtel, Linux Netdev List,
dev@openvswitch.org
In-Reply-To: <80677f296e829f43ca57a4c0dc4f45782e2e6f8e.1421290198.git.tgraf@suug.ch>
> +struct vxlan_metadata {
> + __be32 vni;
> + u32 gbp;
Should this be __be32 also and use ntohl/htonl when setting to/from skb->mark?
> +};
> +
^ permalink raw reply
* [PATCH] net: core: Fix race by protecting process_queues at CPU hotplug
From: subashab @ 2015-01-15 3:13 UTC (permalink / raw)
To: netdev
I am seeing frequent crashes in high throughput conditions in a
multiprocessor system with kernel version 3.10 where cores are getting hot
plugged. I have pinned the network stack to a particular core using
receive packet steering (RPS). At the time of crash, it looks like a
contention of the process_queue between dev_cpu_callback and
process_backlog.
The following is the log at the moment of the crash
75241.598056: <6> Unable to handle kernel NULL pointer dereference at
virtual address 00000004
75241.598064: <6> pgd = c0004000
75241.598072: <2> [00000004] *pgd=00000000
75241.598082: <6> Internal error: Oops: 817 [#1] PREEMPT SMP ARM
75241.598096: <2> Modules linked in: wlan(O) [last unloaded: wlan]
75241.598106: <6> CPU: 0 PID: 3 Comm: ksoftirqd/0 Tainted: G W
O 3.10.49-g0b78c2e-00003-g2eab3e3 #1
75241.598113: <6> task: ee870a80 ti: ee888000 task.ti: ee888000
75241.598133: <2> PC is at process_backlog+0x78/0x140
75241.598142: <2> LR is at __netif_receive_skb_core+0x6fc/0x724
Here is the call stack involved in the crash
-006|__skb_unlink
-006|__skb_dequeue
-006|dev_cpu_callback
-007|notifier_call_chain
-008|__raw_notifier_call_chain
-009|notifier_to_errno
-009|__cpu_notify
-010|cpu_notify_nofail
-011|check_for_tasks
-011|cpu_down
-012|disable_nonboot_cpus()
-013|suspend_enter
-013|suspend_devices_and_enter
-014|enter_state
-014|pm_suspend
-015|state_store
-016|kobj_attr_store
-017|flush_write_buffer
-017|sysfs_write_file
-018|vfs_write
-019|SYSC_write
-019|sys_write
-020|ret_fast_syscall
I have a fix in mind in which the process queues are protected using spin
locks. I do not observe the crash with this patch. Since this patch is in
the hot path for incoming packet processing, I would like some reviews of
this patch. I would also like to know if there are any potential
performance bottlenecks due to this change.
net: core: Protect process_queues at CPU hotplug
When a CPU is hotplugged while processing incoming packets,
dev_cpu_callback() will copy the poll list from the offline
CPU and raise the softIRQ. In the same context, it will also
process process_queue of the offline CPU by de-queueing and
calling netif_rx. Due to this there is a potential for race
condition between process_backlog() and dev_cpu_callback()
accessing the same process_queue resource. This patch
protects this concurrent access by locking.
Signed-off-by: Prasad Sodagudi <psodagud@codeaurora.org>
Signed-off-by: Subash Abhinov Kasiviswanathan <subashab@codeaurora.org>
diff --git a/net/core/dev.c b/net/core/dev.c
index df0b522..aa8f503 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -3640,6 +3640,7 @@ static void flush_backlog(void *arg)
struct net_device *dev = arg;
struct softnet_data *sd = &__get_cpu_var(softnet_data);
struct sk_buff *skb, *tmp;
+ unsigned long flags;
rps_lock(sd);
skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) {
@@ -3651,6 +3652,7 @@ static void flush_backlog(void *arg)
}
rps_unlock(sd);
+ spin_lock_irqsave(&sd->process_queue.lock, flags);
skb_queue_walk_safe(&sd->process_queue, skb, tmp) {
if (skb->dev == dev) {
__skb_unlink(skb, &sd->process_queue);
@@ -3658,6 +3660,7 @@ static void flush_backlog(void *arg)
input_queue_head_incr(sd);
}
}
+ spin_unlock_irqrestore(&sd->process_queue.lock, flags);
}
static int napi_gro_complete(struct sk_buff *skb)
@@ -4021,7 +4024,7 @@ static int process_backlog(struct napi_struct *napi,
int quota)
{
int work = 0;
struct softnet_data *sd = container_of(napi, struct softnet_data,
backlog);
-
+ unsigned long flags;
#ifdef CONFIG_RPS
/* Check if we have pending ipi, its better to send them now,
* not waiting net_rx_action() end.
@@ -4032,18 +4035,19 @@ static int process_backlog(struct napi_struct
*napi, int quota)
}
#endif
napi->weight = weight_p;
- local_irq_disable();
+ spin_lock_irqsave(&sd->process_queue.lock, flags);
while (work < quota) {
struct sk_buff *skb;
unsigned int qlen;
while ((skb = __skb_dequeue(&sd->process_queue))) {
- local_irq_enable();
+ spin_unlock_irqrestore(&sd->process_queue.lock,
flags);
__netif_receive_skb(skb);
- local_irq_disable();
+ spin_lock_irqsave(&sd->process_queue.lock, flags);
input_queue_head_incr(sd);
if (++work >= quota) {
- local_irq_enable();
+
spin_unlock_irqrestore(&sd->process_queue.lock,
+ flags);
return work;
}
}
@@ -4054,6 +4058,7 @@ static int process_backlog(struct napi_struct *napi,
int quota)
skb_queue_splice_tail_init(&sd->input_pkt_queue,
&sd->process_queue);
+
if (qlen < quota - work) {
/*
* Inline a custom version of __napi_complete().
@@ -4069,7 +4074,7 @@ static int process_backlog(struct napi_struct *napi,
int quota)
}
rps_unlock(sd);
}
- local_irq_enable();
+ spin_unlock_irqrestore(&sd->process_queue.lock, flags);
return work;
}
@@ -5991,6 +5996,7 @@ static int dev_cpu_callback(struct notifier_block *nfb,
struct sk_buff *skb;
unsigned int cpu, oldcpu = (unsigned long)ocpu;
struct softnet_data *sd, *oldsd;
+ unsigned long flags;
if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
return NOTIFY_OK;
@@ -6024,11 +6030,15 @@ static int dev_cpu_callback(struct notifier_block
*nfb,
raise_softirq_irqoff(NET_TX_SOFTIRQ);
local_irq_enable();
+ spin_lock_irqsave(&oldsd->process_queue.lock, flags);
/* Process offline CPU's input_pkt_queue */
while ((skb = __skb_dequeue(&oldsd->process_queue))) {
+ spin_unlock_irqrestore(&oldsd->process_queue.lock, flags);
netif_rx(skb);
+ spin_lock_irqsave(&oldsd->process_queue.lock, flags);
input_queue_head_incr(oldsd);
}
+ spin_unlock_irqrestore(&oldsd->process_queue.lock, flags);
while ((skb = __skb_dequeue(&oldsd->input_pkt_queue))) {
netif_rx(skb);
input_queue_head_incr(oldsd);
Thanks
KS
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
^ permalink raw reply related
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox