diff mbox series

[v2] i40e: avoid redundant VF link state updates

Message ID 20251022161143.370040-1-robert.malz@canonical.com
State Accepted
Delegated to: Anthony Nguyen
Headers show
Series [v2] i40e: avoid redundant VF link state updates | expand

Commit Message

Robert Malz Oct. 22, 2025, 4:11 p.m. UTC
From: Jay Vosburgh <jay.vosburgh@canonical.com>

Multiple sources can request VF link state changes with identical
parameters. For example, OpenStack Neutron may request to set the VF link
state to IFLA_VF_LINK_STATE_AUTO during every initialization or user can
issue: `ip link set <ifname> vf 0 state auto` multiple times. Currently,
the i40e driver processes each of these requests, even if the requested
state is the same as the current one. This leads to unnecessary VF resets
and can cause performance degradation or instability in the VF driver,
particularly in environment using Data Plane Development Kit (DPDK).

With this patch i40e will skip VF link state change requests when the
desired link state matches the current configuration. This prevents
unnecessary VF resets and reduces PF-VF communication overhead.

To reproduce the problem run following command multiple times
on the same interface: 'ip link set <ifname> vf 0 state auto'
Every time command is executed, PF driver will trigger VF reset.

Co-developed-by: Robert Malz <robert.malz@canonical.com>
Signed-off-by: Robert Malz <robert.malz@canonical.com>
Signed-off-by: Jay Vosburgh <jay.vosburgh@canonical.com>

---
V1 -> V2: updated commit message, added information how to reproduce

 drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

Comments

Loktionov, Aleksandr Oct. 23, 2025, 5:42 a.m. UTC | #1
> -----Original Message-----
> From: Robert Malz <robert.malz@canonical.com>
> Sent: Wednesday, October 22, 2025 6:12 PM
> To: intel-wired-lan@lists.osuosl.org; netdev@vger.kernel.org
> Cc: Nguyen, Anthony L <anthony.l.nguyen@intel.com>; Kitszel,
> Przemyslaw <przemyslaw.kitszel@intel.com>; Andrew Lunn
> <andrew+netdev@lunn.ch>; David S . Miller <davem@davemloft.net>; Eric
> Dumazet <edumazet@google.com>; Jakub Kicinski <kuba@kernel.org>; Paolo
> Abeni <pabeni@redhat.com>; Simon Horman <horms@kernel.org>; Loktionov,
> Aleksandr <aleksandr.loktionov@intel.com>; Czapnik, Lukasz
> <lukasz.czapnik@intel.com>; Robert Malz <robert.malz@canonical.com>;
> Jamie Bainbridge <jamie.bainbridge@gmail.com>; Jay Vosburgh
> <jay.vosburgh@canonical.com>; Dennis Chen <dechen@redhat.com>; Keller,
> Jacob E <jacob.e.keller@intel.com>; Michal Swiatkowski
> <michal.swiatkowski@linux.intel.com>
> Subject: [PATCH v2] i40e: avoid redundant VF link state updates
> 
> From: Jay Vosburgh <jay.vosburgh@canonical.com>
> 
> Multiple sources can request VF link state changes with identical
> parameters. For example, OpenStack Neutron may request to set the VF
> link state to IFLA_VF_LINK_STATE_AUTO during every initialization or
> user can
> issue: `ip link set <ifname> vf 0 state auto` multiple times.
> Currently, the i40e driver processes each of these requests, even if
> the requested state is the same as the current one. This leads to
> unnecessary VF resets and can cause performance degradation or
> instability in the VF driver, particularly in environment using Data
> Plane Development Kit (DPDK).
> 
> With this patch i40e will skip VF link state change requests when the
> desired link state matches the current configuration. This prevents
> unnecessary VF resets and reduces PF-VF communication overhead.
> 
> To reproduce the problem run following command multiple times on the
> same interface: 'ip link set <ifname> vf 0 state auto'
> Every time command is executed, PF driver will trigger VF reset.
> 
> Co-developed-by: Robert Malz <robert.malz@canonical.com>
> Signed-off-by: Robert Malz <robert.malz@canonical.com>
> Signed-off-by: Jay Vosburgh <jay.vosburgh@canonical.com>
> 
> ---
> V1 -> V2: updated commit message, added information how to reproduce
> 
>  drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
> 
> diff --git a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c
> b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c
> index 081a4526a2f0..0fe0d52c796b 100644
> --- a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c
> +++ b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c
> @@ -4788,6 +4788,7 @@ int i40e_ndo_set_vf_link_state(struct net_device
> *netdev, int vf_id, int link)
>  	unsigned long q_map;
>  	struct i40e_vf *vf;
>  	int abs_vf_id;
> +	int old_link;
>  	int ret = 0;
>  	int tmp;
> 
> @@ -4806,6 +4807,17 @@ int i40e_ndo_set_vf_link_state(struct
> net_device *netdev, int vf_id, int link)
>  	vf = &pf->vf[vf_id];
>  	abs_vf_id = vf->vf_id + hw->func_caps.vf_base_id;
> 
> +	/* skip VF link state change if requested state is already set
> */
> +	if (!vf->link_forced)
> +		old_link = IFLA_VF_LINK_STATE_AUTO;
> +	else if (vf->link_up)
> +		old_link = IFLA_VF_LINK_STATE_ENABLE;
> +	else
> +		old_link = IFLA_VF_LINK_STATE_DISABLE;
> +
> +	if (link == old_link)
> +		goto error_out;
> +
>  	pfe.event = VIRTCHNL_EVENT_LINK_CHANGE;
>  	pfe.severity = PF_EVENT_SEVERITY_INFO;
> 
> --
> 2.34.1

Reviewed-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Romanowski, Rafal Oct. 29, 2025, 9:44 a.m. UTC | #2
> -----Original Message-----
> From: Intel-wired-lan <intel-wired-lan-bounces@osuosl.org> On Behalf Of
> Robert Malz
> Sent: Wednesday, October 22, 2025 6:12 PM
> To: intel-wired-lan@lists.osuosl.org; netdev@vger.kernel.org
> Cc: Jamie Bainbridge <jamie.bainbridge@gmail.com>; Michal Swiatkowski
> <michal.swiatkowski@linux.intel.com>; Dennis Chen <dechen@redhat.com>;
> Robert Malz <robert.malz@canonical.com>; Kitszel, Przemyslaw
> <przemyslaw.kitszel@intel.com>; Czapnik, Lukasz
> <lukasz.czapnik@intel.com>; Loktionov, Aleksandr
> <aleksandr.loktionov@intel.com>; Andrew Lunn <andrew+netdev@lunn.ch>;
> Eric Dumazet <edumazet@google.com>; Nguyen, Anthony L
> <anthony.l.nguyen@intel.com>; Simon Horman <horms@kernel.org>; Keller,
> Jacob E <jacob.e.keller@intel.com>; Jakub Kicinski <kuba@kernel.org>; Paolo
> Abeni <pabeni@redhat.com>; David S . Miller <davem@davemloft.net>
> Subject: [Intel-wired-lan] [PATCH v2] i40e: avoid redundant VF link state
> updates
> 
> From: Jay Vosburgh <jay.vosburgh@canonical.com>
> 
> Multiple sources can request VF link state changes with identical parameters.
> For example, OpenStack Neutron may request to set the VF link state to
> IFLA_VF_LINK_STATE_AUTO during every initialization or user can
> issue: `ip link set <ifname> vf 0 state auto` multiple times. Currently, the i40e
> driver processes each of these requests, even if the requested state is the
> same as the current one. This leads to unnecessary VF resets and can cause
> performance degradation or instability in the VF driver, particularly in
> environment using Data Plane Development Kit (DPDK).
> 
> With this patch i40e will skip VF link state change requests when the desired
> link state matches the current configuration. This prevents unnecessary VF
> resets and reduces PF-VF communication overhead.
> 
> To reproduce the problem run following command multiple times on the same
> interface: 'ip link set <ifname> vf 0 state auto'
> Every time command is executed, PF driver will trigger VF reset.
> 
> Co-developed-by: Robert Malz <robert.malz@canonical.com>
> Signed-off-by: Robert Malz <robert.malz@canonical.com>
> Signed-off-by: Jay Vosburgh <jay.vosburgh@canonical.com>
> 
> ---
> V1 -> V2: updated commit message, added information how to reproduce
> 
>  drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
> 

Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
diff mbox series

Patch

diff --git a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c
index 081a4526a2f0..0fe0d52c796b 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c
@@ -4788,6 +4788,7 @@  int i40e_ndo_set_vf_link_state(struct net_device *netdev, int vf_id, int link)
 	unsigned long q_map;
 	struct i40e_vf *vf;
 	int abs_vf_id;
+	int old_link;
 	int ret = 0;
 	int tmp;
 
@@ -4806,6 +4807,17 @@  int i40e_ndo_set_vf_link_state(struct net_device *netdev, int vf_id, int link)
 	vf = &pf->vf[vf_id];
 	abs_vf_id = vf->vf_id + hw->func_caps.vf_base_id;
 
+	/* skip VF link state change if requested state is already set */
+	if (!vf->link_forced)
+		old_link = IFLA_VF_LINK_STATE_AUTO;
+	else if (vf->link_up)
+		old_link = IFLA_VF_LINK_STATE_ENABLE;
+	else
+		old_link = IFLA_VF_LINK_STATE_DISABLE;
+
+	if (link == old_link)
+		goto error_out;
+
 	pfe.event = VIRTCHNL_EVENT_LINK_CHANGE;
 	pfe.severity = PF_EVENT_SEVERITY_INFO;