[PATCH iwl-next] ice: fix FDIR CTRL VSI resource leak in ice_reset_all_vfs()

[PATCH iwl-next] ice: fix FDIR CTRL VSI resource leak in ice_reset_all_vfs()

[Aleksandr Loktionov]

From: Dawid Osuchowski <dawid.osuchowski@linux.intel.com>

Resetting all VFs causes resource leak on VFs with FDIR filters
enabled as CTRL VSIs are only invalidated and not freed. Fix by using
ice_vf_ctrl_vsi_release() instead of ice_vf_ctrl_invalidate_vsi() which
aligns behavior with the ice_reset_vf() function.

Reproduction:
  echo 1 > /sys/class/net/$pf/device/sriov_numvfs
  ethtool -N $vf flow-type ether proto 0x9000 action 0
  echo 1 > /sys/class/net/$pf/device/reset

Fixes: da62c5ff9dcd ("ice: Add support for per VF ctrl VSI enabling")
Signed-off-by: Dawid Osuchowski <dawid.osuchowski@linux.intel.com>
Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
---

 drivers/net/ethernet/intel/ice/ice_vf_lib.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_vf_lib.c b/drivers/net/ethernet/intel/ice/ice_vf_lib.c
index c8bc952f..7d33f09 100644
--- a/drivers/net/ethernet/intel/ice/ice_vf_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_vf_lib.c
@@ -801,7 +801,7 @@ void ice_reset_all_vfs(struct ice_pf *pf)
 		 * setup only when VF creates its first FDIR rule.
 		 */
 		if (vf->ctrl_vsi_idx != ICE_NO_VSI)
-			ice_vf_ctrl_invalidate_vsi(vf);
+			ice_vf_ctrl_vsi_release(vf);
 
 		ice_vf_pre_vsi_rebuild(vf);
 		ice_vf_rebuild_vsi(vf);
-- 
2.52.0

[PATCH iwl-next] ice: do not reset MDD counters on VF reset

[Aleksandr Loktionov]

From: Mark Rustad <mark.d.rustad@intel.com>

Do not clear MDD event counters on VF reset, to be consistent with
how VF statistics are not reset either. This allows accumulation of
MDD events across resets so that persistent misbehavior can be
tracked and reported more accurately.

Fixes: 12bb018c538c ("ice: Refactor VF reset")
Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
---

 drivers/net/ethernet/intel/ice/ice_vf_lib.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_vf_lib.c b/drivers/net/ethernet/intel/ice/ice_vf_lib.c
index 7d33f09..9bcc739 100644
--- a/drivers/net/ethernet/intel/ice/ice_vf_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_vf_lib.c
@@ -227,8 +227,6 @@ static void ice_vf_clear_counters(struct ice_vf *vf)
 
 	vf->num_mac = 0;
 	vf->num_mac_lldp = 0;
-	memset(&vf->mdd_tx_events, 0, sizeof(vf->mdd_tx_events));
-	memset(&vf->mdd_rx_events, 0, sizeof(vf->mdd_rx_events));
 }
 
 /**
-- 
2.52.0

Re: [PATCH iwl-next] ice: do not reset MDD counters on VF reset

[Tony Nguyen]

On 3/27/2026 12:22 AM, Aleksandr Loktionov wrote:
> From: Mark Rustad <mark.d.rustad@intel.com>
> 
> Do not clear MDD event counters on VF reset, to be consistent with
> how VF statistics are not reset either. This allows accumulation of
> MDD events across resets so that persistent misbehavior can be
> tracked and reported more accurately.
> 
> Fixes: 12bb018c538c ("ice: Refactor VF reset")
> Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
> Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
> ---
> 
>   drivers/net/ethernet/intel/ice/ice_vf_lib.c | 2 --
>   1 file changed, 2 deletions(-)
> 
> diff --git a/drivers/net/ethernet/intel/ice/ice_vf_lib.c b/drivers/net/ethernet/intel/ice/ice_vf_lib.c
> index 7d33f09..9bcc739 100644
> --- a/drivers/net/ethernet/intel/ice/ice_vf_lib.c
> +++ b/drivers/net/ethernet/intel/ice/ice_vf_lib.c
> @@ -227,8 +227,6 @@ static void ice_vf_clear_counters(struct ice_vf *vf)
>   
>   	vf->num_mac = 0;
>   	vf->num_mac_lldp = 0;
> -	memset(&vf->mdd_tx_events, 0, sizeof(vf->mdd_tx_events));
> -	memset(&vf->mdd_rx_events, 0, sizeof(vf->mdd_rx_events));
>   }
>   
>   /**

AI Review says:

There's a documentation issue in 
drivers/net/ethernet/intel/ice/ice_main.c that should be addressed. The 
comment in ice_mdd_maybe_reset_vf() states:

 > /* VF MDD event counters will be cleared by reset, so print the event
 >  * prior to reset.
 >  */

This comment contradicts the new behavior introduced by this patch. 
Since the memset calls are removed, MDD event counters now persist 
across VF resets rather than being cleared. The comment should be 
updated to reflect that counters are accumulated across resets to track 
persistent misbehavior, which is the intent documented in the commit 
message.



Also, the threading on your sends seems to be off. If you are going to 
send them together, please make them part of a series. If you are 
treating them individually, please send them that way.

The target trees also seem off; some 'net' and some 'iwl-next' patches 
with Fixes: that sound to be bug fixes. I'm working through and sorting 
what's been sent, but please try to tag things properly on future sends.

Thanks,
Tony

[PATCH iwl-next] ice: check cross-timestamp timeout bits

[Aleksandr Loktionov]

From: Karol Kolacinski <karol.kolacinski@intel.com>

Polling for cross-timestamp active bit depends on HW scheduling and
actual timeout may happen before the driver finishes polling.

Check cross-timestamp timeout bits to ensure that the driver finishes
the operation earlier when HW indicates timeout.

Fixes: 92456e795ac6 ("ice: Add unified ice_capture_crosststamp")
Signed-off-by: Karol Kolacinski <karol.kolacinski@intel.com>
Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
---

 drivers/net/ethernet/intel/ice/ice_hw_autogen.h |  3 +++
 drivers/net/ethernet/intel/ice/ice_ptp.c        | 12 ++++++++++--
 2 files changed, 13 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_hw_autogen.h b/drivers/net/ethernet/intel/ice/ice_hw_autogen.h
index 082ad33..8b4884e 100644
--- a/drivers/net/ethernet/intel/ice/ice_hw_autogen.h
+++ b/drivers/net/ethernet/intel/ice/ice_hw_autogen.h
@@ -499,6 +499,8 @@
 #define PRTRPB_RDPC				0x000AC260
 #define GLHH_ART_CTL				0x000A41D4
 #define GLHH_ART_CTL_ACTIVE_M			BIT(0)
+#define GLHH_ART_CTL_TIME_OUT1_M		BIT(1)
+#define GLHH_ART_CTL_TIME_OUT2_M		BIT(2)
 #define GLHH_ART_TIME_H				0x000A41D8
 #define GLHH_ART_TIME_L				0x000A41DC
 #define GLTSYN_AUX_IN_0(_i)			(0x000889D8 + ((_i) * 4))
@@ -564,6 +566,7 @@
 #define E830_PRTTSYN_TXTIME_L(_i)		(0x001E5000 + ((_i) * 32))
 #define E830_GLPTM_ART_CTL			0x00088B50
 #define E830_GLPTM_ART_CTL_ACTIVE_M		BIT(0)
+#define E830_GLPTM_ART_CTL_TIME_OUT_M		BIT(1)
 #define E830_GLPTM_ART_TIME_H			0x00088B54
 #define E830_GLPTM_ART_TIME_L			0x00088B58
 #define E830_GLTSYN_PTMTIME_H(_i)		(0x00088B48 + ((_i) * 4))
diff --git a/drivers/net/ethernet/intel/ice/ice_ptp.c b/drivers/net/ethernet/intel/ice/ice_ptp.c
index 6848b1c..8b0530b 100644
--- a/drivers/net/ethernet/intel/ice/ice_ptp.c
+++ b/drivers/net/ethernet/intel/ice/ice_ptp.c
@@ -2011,6 +2011,7 @@ static int ice_ptp_adjtime(struct ptp_clock_info *info, s64 delta)
  * @lock_busy: Bit in the semaphore lock indicating the lock is busy
  * @ctl_reg: The hardware register to request cross timestamp
  * @ctl_active: Bit in the control register to request cross timestamp
+ * @ctl_timeout: Bits in the control register to indicate HW timeout
  * @art_time_l: Lower 32-bits of ART system time
  * @art_time_h: Upper 32-bits of ART system time
  * @dev_time_l: Lower 32-bits of device time (per timer index)
@@ -2024,6 +2025,7 @@ struct ice_crosststamp_cfg {
 	/* Capture control register */
 	u32 ctl_reg;
 	u32 ctl_active;
+	u32 ctl_timeout;
 
 	/* Time storage */
 	u32 art_time_l;
@@ -2037,6 +2039,7 @@ static const struct ice_crosststamp_cfg ice_crosststamp_cfg_e82x = {
 	.lock_busy = PFHH_SEM_BUSY_M,
 	.ctl_reg = GLHH_ART_CTL,
 	.ctl_active = GLHH_ART_CTL_ACTIVE_M,
+	.ctl_timeout = GLHH_ART_CTL_TIME_OUT1_M | GLHH_ART_CTL_TIME_OUT2_M,
 	.art_time_l = GLHH_ART_TIME_L,
 	.art_time_h = GLHH_ART_TIME_H,
 	.dev_time_l[0] = GLTSYN_HHTIME_L(0),
@@ -2051,6 +2054,7 @@ static const struct ice_crosststamp_cfg ice_crosststamp_cfg_e830 = {
 	.lock_busy = E830_PFPTM_SEM_BUSY_M,
 	.ctl_reg = E830_GLPTM_ART_CTL,
 	.ctl_active = E830_GLPTM_ART_CTL_ACTIVE_M,
+	.ctl_timeout = E830_GLPTM_ART_CTL_TIME_OUT_M,
 	.art_time_l = E830_GLPTM_ART_TIME_L,
 	.art_time_h = E830_GLPTM_ART_TIME_H,
 	.dev_time_l[0] = E830_GLTSYN_PTMTIME_L(0),
@@ -2123,9 +2127,13 @@ static int ice_capture_crosststamp(ktime_t *device,
 	ctl |= cfg->ctl_active;
 	wr32(hw, cfg->ctl_reg, ctl);
 
-	/* Poll until hardware completes the capture */
-	err = rd32_poll_timeout(hw, cfg->ctl_reg, ctl, !(ctl & cfg->ctl_active),
+	/* Poll until hardware completes the capture or timeout occurs */
+	err = rd32_poll_timeout(hw, cfg->ctl_reg, ctl,
+				!(ctl & cfg->ctl_active) ||
+				(ctl & cfg->ctl_timeout),
 				5, 20 * USEC_PER_MSEC);
+	if (ctl & cfg->ctl_timeout)
+		err = -ETIMEDOUT;
 	if (err)
 		goto err_timeout;
 
-- 
2.52.0

Re: [Intel-wired-lan] [PATCH iwl-next] ice: check cross-timestamp timeout bits

[Jacob Keller]

On 3/27/2026 12:22 AM, Aleksandr Loktionov wrote:
> From: Karol Kolacinski <karol.kolacinski@intel.com>
> 
> Polling for cross-timestamp active bit depends on HW scheduling and
> actual timeout may happen before the driver finishes polling.
> 
> Check cross-timestamp timeout bits to ensure that the driver finishes
> the operation earlier when HW indicates timeout.
> 
> Fixes: 92456e795ac6 ("ice: Add unified ice_capture_crosststamp")
> Signed-off-by: Karol Kolacinski <karol.kolacinski@intel.com>
> Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
> ---
> 

Waste less time once hardware gives up. Makes sense.

Acked-by: Jacob Keller <jacob.e.keller@intel.com>

Re: [PATCH iwl-next] ice: check cross-timestamp timeout bits

[Simon Horman]

On Fri, Mar 27, 2026 at 08:22:34AM +0100, Aleksandr Loktionov wrote:
> From: Karol Kolacinski <karol.kolacinski@intel.com>
> 
> Polling for cross-timestamp active bit depends on HW scheduling and
> actual timeout may happen before the driver finishes polling.
> 
> Check cross-timestamp timeout bits to ensure that the driver finishes
> the operation earlier when HW indicates timeout.
> 
> Fixes: 92456e795ac6 ("ice: Add unified ice_capture_crosststamp")
> Signed-off-by: Karol Kolacinski <karol.kolacinski@intel.com>
> Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>

Reviewed-by: Simon Horman <horms@kernel.org>

[PATCH iwl-next] ice: fix AQ error code comparison in ice_set_pauseparam()

[Aleksandr Loktionov]

From: Lukasz Czapnik <lukasz.czapnik@intel.com>

Fix unreachable code: the conditionals in ice_set_pauseparam() used
the bitwise-AND operator suggesting aq_failures is a bitmap, but it
is actually an enum, making the third condition logically unreachable.

Replace the if-else ladder with a switch statement.  Also move the
aq_failures initialization to the variable declaration and remove the
redundant zeroing from ice_set_fc().

Fixes: fcea6f3da546 ("ice: Add stats and ethtool support")
Signed-off-by: Lukasz Czapnik <lukasz.czapnik@intel.com>
Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
---

 drivers/net/ethernet/intel/ice/ice_common.c  |  1 -
 drivers/net/ethernet/intel/ice/ice_ethtool.c | 12 ++++++++----
 2 files changed, 8 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_common.c b/drivers/net/ethernet/intel/ice/ice_common.c
index f1a6601..6dad7d4 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.c
+++ b/drivers/net/ethernet/intel/ice/ice_common.c
@@ -3883,7 +3883,6 @@ ice_set_fc(struct ice_port_info *pi, u8 *aq_failures, bool ena_auto_link_update)
 	if (!pi || !aq_failures)
 		return -EINVAL;
 
-	*aq_failures = 0;
 	hw = pi->hw;
 
 	pcaps = kzalloc_obj(*pcaps);
diff --git a/drivers/net/ethernet/intel/ice/ice_ethtool.c b/drivers/net/ethernet/intel/ice/ice_ethtool.c
index 2a4f06f..0cf064c 100644
--- a/drivers/net/ethernet/intel/ice/ice_ethtool.c
+++ b/drivers/net/ethernet/intel/ice/ice_ethtool.c
@@ -3501,7 +3501,7 @@ ice_set_pauseparam(struct net_device *netdev, struct ethtool_pauseparam *pause)
 	struct ice_vsi *vsi = np->vsi;
 	struct ice_hw *hw = &pf->hw;
 	struct ice_port_info *pi;
-	u8 aq_failures;
+	u8 aq_failures = 0;
 	bool link_up;
 	u32 is_an;
 	int err;
@@ -3572,18 +3572,22 @@ ice_set_pauseparam(struct net_device *netdev, struct ethtool_pauseparam *pause)
 	/* Set the FC mode and only restart AN if link is up */
 	err = ice_set_fc(pi, &aq_failures, link_up);
 
-	if (aq_failures & ICE_SET_FC_AQ_FAIL_GET) {
+	switch (aq_failures) {
+	case ICE_SET_FC_AQ_FAIL_GET:
 		netdev_info(netdev, "Set fc failed on the get_phy_capabilities call with err %d aq_err %s\n",
 			    err, libie_aq_str(hw->adminq.sq_last_status));
 		err = -EAGAIN;
-	} else if (aq_failures & ICE_SET_FC_AQ_FAIL_SET) {
+		break;
+	case ICE_SET_FC_AQ_FAIL_SET:
 		netdev_info(netdev, "Set fc failed on the set_phy_config call with err %d aq_err %s\n",
 			    err, libie_aq_str(hw->adminq.sq_last_status));
 		err = -EAGAIN;
-	} else if (aq_failures & ICE_SET_FC_AQ_FAIL_UPDATE) {
+		break;
+	case ICE_SET_FC_AQ_FAIL_UPDATE:
 		netdev_info(netdev, "Set fc failed on the get_link_info call with err %d aq_err %s\n",
 			    err, libie_aq_str(hw->adminq.sq_last_status));
 		err = -EAGAIN;
+		break;
 	}
 
 	return err;
-- 
2.52.0

Re: [PATCH iwl-next] ice: fix AQ error code comparison in ice_set_pauseparam()

[Simon Horman]

On Fri, Mar 27, 2026 at 08:22:35AM +0100, Aleksandr Loktionov wrote:
> From: Lukasz Czapnik <lukasz.czapnik@intel.com>
> 
> Fix unreachable code: the conditionals in ice_set_pauseparam() used
> the bitwise-AND operator suggesting aq_failures is a bitmap, but it
> is actually an enum, making the third condition logically unreachable.
> 
> Replace the if-else ladder with a switch statement.  Also move the
> aq_failures initialization to the variable declaration and remove the
> redundant zeroing from ice_set_fc().

I think that this second part feels more like a clean-up
than part of the fix. But I don't feel strongly about it.

> Fixes: fcea6f3da546 ("ice: Add stats and ethtool support")
> Signed-off-by: Lukasz Czapnik <lukasz.czapnik@intel.com>
> Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>

Reviewed-by: Simon Horman <horms@kernel.org>

> ---
> 
>  drivers/net/ethernet/intel/ice/ice_common.c  |  1 -
>  drivers/net/ethernet/intel/ice/ice_ethtool.c | 12 ++++++++----
>  2 files changed, 8 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/net/ethernet/intel/ice/ice_common.c b/drivers/net/ethernet/intel/ice/ice_common.c
> index f1a6601..6dad7d4 100644
> --- a/drivers/net/ethernet/intel/ice/ice_common.c
> +++ b/drivers/net/ethernet/intel/ice/ice_common.c
> @@ -3883,7 +3883,6 @@ ice_set_fc(struct ice_port_info *pi, u8 *aq_failures, bool ena_auto_link_update)
>  	if (!pi || !aq_failures)
>  		return -EINVAL;

FWIIW, I think the defensive code above could also be removed at some point.

>  
> -	*aq_failures = 0;
>  	hw = pi->hw;
>  
>  	pcaps = kzalloc_obj(*pcaps);

...

[PATCH iwl-next] ice: call netif_keep_dst() once when entering switchdev mode

[Aleksandr Loktionov]

From: Marcin Szycik <marcin.szycik@intel.com>

netif_keep_dst() only needs to be called once for the uplink VSI, not
once for each port representor.  Move it from ice_eswitch_setup_repr()
to ice_eswitch_enable_switchdev().

Fixes: defd52455aee ("ice: do Tx through PF netdev in slow-path")
Signed-off-by: Marcin Szycik <marcin.szycik@intel.com>
Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
---

 drivers/net/ethernet/intel/ice/ice_eswitch.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_eswitch.c b/drivers/net/ethernet/intel/ice/ice_eswitch.c
index 2e4f096..c30e27b 100644
--- a/drivers/net/ethernet/intel/ice/ice_eswitch.c
+++ b/drivers/net/ethernet/intel/ice/ice_eswitch.c
@@ -117,8 +117,6 @@ static int ice_eswitch_setup_repr(struct ice_pf *pf, struct ice_repr *repr)
 	if (!repr->dst)
 		return -ENOMEM;
 
-	netif_keep_dst(uplink_vsi->netdev);
-
 	dst = repr->dst;
 	dst->u.port_info.port_id = vsi->vsi_num;
 	dst->u.port_info.lower_dev = uplink_vsi->netdev;
@@ -312,6 +310,8 @@ static int ice_eswitch_enable_switchdev(struct ice_pf *pf)
 	if (ice_eswitch_br_offloads_init(pf))
 		goto err_br_offloads;
 
+	netif_keep_dst(uplink_vsi->netdev);
+
 	pf->eswitch.is_running = true;
 
 	return 0;
-- 
2.52.0

Re: [PATCH iwl-next] ice: call netif_keep_dst() once when entering switchdev mode

[Simon Horman]

On Fri, Mar 27, 2026 at 08:22:36AM +0100, Aleksandr Loktionov wrote:
> From: Marcin Szycik <marcin.szycik@intel.com>
> 
> netif_keep_dst() only needs to be called once for the uplink VSI, not
> once for each port representor.  Move it from ice_eswitch_setup_repr()
> to ice_eswitch_enable_switchdev().
> 
> Fixes: defd52455aee ("ice: do Tx through PF netdev in slow-path")

This problem seems to predate the cited commit.

> Signed-off-by: Marcin Szycik <marcin.szycik@intel.com>
> Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>

Re: [PATCH iwl-next] ice: fix FDIR CTRL VSI resource leak in ice_reset_all_vfs()

[Simon Horman]

On Fri, Mar 27, 2026 at 08:22:32AM +0100, Aleksandr Loktionov wrote:
> From: Dawid Osuchowski <dawid.osuchowski@linux.intel.com>
> 
> Resetting all VFs causes resource leak on VFs with FDIR filters
> enabled as CTRL VSIs are only invalidated and not freed. Fix by using
> ice_vf_ctrl_vsi_release() instead of ice_vf_ctrl_invalidate_vsi() which
> aligns behavior with the ice_reset_vf() function.
> 
> Reproduction:
>   echo 1 > /sys/class/net/$pf/device/sriov_numvfs
>   ethtool -N $vf flow-type ether proto 0x9000 action 0
>   echo 1 > /sys/class/net/$pf/device/reset
> 
> Fixes: da62c5ff9dcd ("ice: Add support for per VF ctrl VSI enabling")
> Signed-off-by: Dawid Osuchowski <dawid.osuchowski@linux.intel.com>
> Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>

Reviewed-by: Simon Horman <horms@kernel.org>