[PATCH net-next 0/8][pull request] ice: support dynamic interrupt allocation

[PATCH net-next 0/8][pull request] ice: support dynamic interrupt allocation

[Tony Nguyen]

Piotr Raczynski says:

This patchset reimplements MSIX interrupt allocation logic to allow dynamic
interrupt allocation after MSIX has been initially enabled. This allows
current and future features to allocate and free interrupts as needed and
will help to drastically decrease number of initially preallocated
interrupts (even down to the API hard limit of 1). Although this patchset
does not change behavior in terms of actual number of allocated interrupts
during probe, it will be subject to change.

First few patches prepares to introduce dynamic allocation by moving
interrupt allocation code to separate file and update allocation API used
in the driver to the currently preferred one.

Due to the current contract between ice and irdma driver which is directly
accessing msix entries allocated by ice driver, even after moving away from
older pci_enable_msix_range function, still keep msix_entries array for
irdma use.

Next patches refactors and removes redundant code from SRIOV related logic
as it also make it easier to move away from static allocation scheme.

Last patches actually enables dynamic allocation of MSIX interrupts. First,
introduce functions to allocate and free interrupts individually. This sets
ground for the rest of the changes even if that patch still allocates the
interrupts from the preallocated pool. Since this patch starts to keep
interrupt details in ice_q_vector structure we can get rid of functions
that calculates base vector number and register offset for the interrupt
as it is equal to the interrupt index. Only keep separate register offset
functions for the VF VSIs.

Next, replace homegrown interrupt tracker with much simpler xarray based
approach. As new API always allocate interrupts one by one, also track
interrupts in the same manner.

Lastly, extend the interrupt tracker to deal both with preallocated and
dynamically allocated vectors and use pci_msix_alloc_irq_at and
pci_msix_free_irq functions. Since not all architecture supports dynamic
allocation, check it before trying to allocate a new interrupt.

As previously mentioned, this patchset does not change number of initially
allocated interrupts during init phase but now it can and will likely be
changed.

Patch 1-3 -> move code around and use newer API
Patch 4-5 -> refactor and remove redundant SRIOV code
Patch 6   -> allocate every interrupt individually
Patch 7   -> replace homegrown interrupt tracker with xarray
Patch 8   -> allow dynamic interrupt allocation

The following are changes since commit 94e86ef1b801d213dfb8543633dec86abb1a457d:
  net: phy: dp83869: support mii mode when rgmii strap cfg is used
and are available in the git repository at:
  git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue 100GbE

Piotr Raczynski (8):
  ice: move interrupt related code to separate file
  ice: use pci_irq_vector helper function
  ice: use preferred MSIX allocation api
  ice: refactor VF control VSI interrupt handling
  ice: remove redundant SRIOV code
  ice: add individual interrupt allocation
  ice: track interrupt vectors with xarray
  ice: add dynamic interrupt allocation

 drivers/net/ethernet/intel/ice/Makefile      |   1 +
 drivers/net/ethernet/intel/ice/ice.h         |  24 +-
 drivers/net/ethernet/intel/ice/ice_arfs.c    |   5 +-
 drivers/net/ethernet/intel/ice/ice_base.c    |  41 +-
 drivers/net/ethernet/intel/ice/ice_ethtool.c |   2 +-
 drivers/net/ethernet/intel/ice/ice_idc.c     |  54 ++-
 drivers/net/ethernet/intel/ice/ice_irq.c     | 378 +++++++++++++++++++
 drivers/net/ethernet/intel/ice/ice_irq.h     |  25 ++
 drivers/net/ethernet/intel/ice/ice_lib.c     | 288 +-------------
 drivers/net/ethernet/intel/ice/ice_lib.h     |   5 -
 drivers/net/ethernet/intel/ice/ice_main.c    | 268 ++-----------
 drivers/net/ethernet/intel/ice/ice_ptp.c     |   2 +-
 drivers/net/ethernet/intel/ice/ice_sriov.c   |  43 +--
 drivers/net/ethernet/intel/ice/ice_vf_lib.c  |  32 ++
 drivers/net/ethernet/intel/ice/ice_vf_lib.h  |   7 +
 drivers/net/ethernet/intel/ice/ice_xsk.c     |   5 +-
 16 files changed, 560 insertions(+), 620 deletions(-)
 create mode 100644 drivers/net/ethernet/intel/ice/ice_irq.c
 create mode 100644 drivers/net/ethernet/intel/ice/ice_irq.h

-- 
2.38.1

[PATCH net-next 1/8] ice: move interrupt related code to separate file

[Tony Nguyen]

From: Piotr Raczynski <piotr.raczynski@intel.com>

Keep interrupt handling code in a dedicated file. This helps keep driver
structured better and prepares for more functionality added to this file.

Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Piotr Raczynski <piotr.raczynski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/ice/Makefile   |   1 +
 drivers/net/ethernet/intel/ice/ice.h      |   1 +
 drivers/net/ethernet/intel/ice/ice_irq.c  | 226 ++++++++++++++++++++++
 drivers/net/ethernet/intel/ice/ice_irq.h  |  10 +
 drivers/net/ethernet/intel/ice/ice_main.c | 218 ---------------------
 5 files changed, 238 insertions(+), 218 deletions(-)
 create mode 100644 drivers/net/ethernet/intel/ice/ice_irq.c
 create mode 100644 drivers/net/ethernet/intel/ice/ice_irq.h

diff --git a/drivers/net/ethernet/intel/ice/Makefile b/drivers/net/ethernet/intel/ice/Makefile
index 5d89392f969b..817977e3039d 100644
--- a/drivers/net/ethernet/intel/ice/Makefile
+++ b/drivers/net/ethernet/intel/ice/Makefile
@@ -18,6 +18,7 @@ ice-y := ice_main.o	\
 	 ice_txrx_lib.o	\
 	 ice_txrx.o	\
 	 ice_fltr.o	\
+	 ice_irq.o	\
 	 ice_pf_vsi_vlan_ops.o \
 	 ice_vsi_vlan_ops.o \
 	 ice_vsi_vlan_lib.o \
diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
index aa32111afd6e..4dc4c6bec081 100644
--- a/drivers/net/ethernet/intel/ice/ice.h
+++ b/drivers/net/ethernet/intel/ice/ice.h
@@ -74,6 +74,7 @@
 #include "ice_lag.h"
 #include "ice_vsi_vlan_ops.h"
 #include "ice_gnss.h"
+#include "ice_irq.h"
 
 #define ICE_BAR0		0
 #define ICE_REQ_DESC_MULTIPLE	32
diff --git a/drivers/net/ethernet/intel/ice/ice_irq.c b/drivers/net/ethernet/intel/ice/ice_irq.c
new file mode 100644
index 000000000000..1fc7daec9732
--- /dev/null
+++ b/drivers/net/ethernet/intel/ice/ice_irq.c
@@ -0,0 +1,226 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (C) 2023, Intel Corporation. */
+
+#include "ice.h"
+#include "ice_lib.h"
+#include "ice_irq.h"
+
+/**
+ * ice_reduce_msix_usage - Reduce usage of MSI-X vectors
+ * @pf: board private structure
+ * @v_remain: number of remaining MSI-X vectors to be distributed
+ *
+ * Reduce the usage of MSI-X vectors when entire request cannot be fulfilled.
+ * pf->num_lan_msix and pf->num_rdma_msix values are set based on number of
+ * remaining vectors.
+ */
+static void ice_reduce_msix_usage(struct ice_pf *pf, int v_remain)
+{
+	int v_rdma;
+
+	if (!ice_is_rdma_ena(pf)) {
+		pf->num_lan_msix = v_remain;
+		return;
+	}
+
+	/* RDMA needs at least 1 interrupt in addition to AEQ MSIX */
+	v_rdma = ICE_RDMA_NUM_AEQ_MSIX + 1;
+
+	if (v_remain < ICE_MIN_LAN_TXRX_MSIX + ICE_MIN_RDMA_MSIX) {
+		dev_warn(ice_pf_to_dev(pf), "Not enough MSI-X vectors to support RDMA.\n");
+		clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
+
+		pf->num_rdma_msix = 0;
+		pf->num_lan_msix = ICE_MIN_LAN_TXRX_MSIX;
+	} else if ((v_remain < ICE_MIN_LAN_TXRX_MSIX + v_rdma) ||
+		   (v_remain - v_rdma < v_rdma)) {
+		/* Support minimum RDMA and give remaining vectors to LAN MSIX
+		 */
+		pf->num_rdma_msix = ICE_MIN_RDMA_MSIX;
+		pf->num_lan_msix = v_remain - ICE_MIN_RDMA_MSIX;
+	} else {
+		/* Split remaining MSIX with RDMA after accounting for AEQ MSIX
+		 */
+		pf->num_rdma_msix = (v_remain - ICE_RDMA_NUM_AEQ_MSIX) / 2 +
+				    ICE_RDMA_NUM_AEQ_MSIX;
+		pf->num_lan_msix = v_remain - pf->num_rdma_msix;
+	}
+}
+
+/**
+ * ice_ena_msix_range - Request a range of MSIX vectors from the OS
+ * @pf: board private structure
+ *
+ * Compute the number of MSIX vectors wanted and request from the OS. Adjust
+ * device usage if there are not enough vectors. Return the number of vectors
+ * reserved or negative on failure.
+ */
+static int ice_ena_msix_range(struct ice_pf *pf)
+{
+	int num_cpus, hw_num_msix, v_other, v_wanted, v_actual;
+	struct device *dev = ice_pf_to_dev(pf);
+	int err, i;
+
+	hw_num_msix = pf->hw.func_caps.common_cap.num_msix_vectors;
+	num_cpus = num_online_cpus();
+
+	/* LAN miscellaneous handler */
+	v_other = ICE_MIN_LAN_OICR_MSIX;
+
+	/* Flow Director */
+	if (test_bit(ICE_FLAG_FD_ENA, pf->flags))
+		v_other += ICE_FDIR_MSIX;
+
+	/* switchdev */
+	v_other += ICE_ESWITCH_MSIX;
+
+	v_wanted = v_other;
+
+	/* LAN traffic */
+	pf->num_lan_msix = num_cpus;
+	v_wanted += pf->num_lan_msix;
+
+	/* RDMA auxiliary driver */
+	if (ice_is_rdma_ena(pf)) {
+		pf->num_rdma_msix = num_cpus + ICE_RDMA_NUM_AEQ_MSIX;
+		v_wanted += pf->num_rdma_msix;
+	}
+
+	if (v_wanted > hw_num_msix) {
+		int v_remain;
+
+		dev_warn(dev, "not enough device MSI-X vectors. wanted = %d, available = %d\n",
+			 v_wanted, hw_num_msix);
+
+		if (hw_num_msix < ICE_MIN_MSIX) {
+			err = -ERANGE;
+			goto exit_err;
+		}
+
+		v_remain = hw_num_msix - v_other;
+		if (v_remain < ICE_MIN_LAN_TXRX_MSIX) {
+			v_other = ICE_MIN_MSIX - ICE_MIN_LAN_TXRX_MSIX;
+			v_remain = ICE_MIN_LAN_TXRX_MSIX;
+		}
+
+		ice_reduce_msix_usage(pf, v_remain);
+		v_wanted = pf->num_lan_msix + pf->num_rdma_msix + v_other;
+
+		dev_notice(dev, "Reducing request to %d MSI-X vectors for LAN traffic.\n",
+			   pf->num_lan_msix);
+		if (ice_is_rdma_ena(pf))
+			dev_notice(dev, "Reducing request to %d MSI-X vectors for RDMA.\n",
+				   pf->num_rdma_msix);
+	}
+
+	pf->msix_entries = devm_kcalloc(dev, v_wanted,
+					sizeof(*pf->msix_entries), GFP_KERNEL);
+	if (!pf->msix_entries) {
+		err = -ENOMEM;
+		goto exit_err;
+	}
+
+	for (i = 0; i < v_wanted; i++)
+		pf->msix_entries[i].entry = i;
+
+	/* actually reserve the vectors */
+	v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
+					 ICE_MIN_MSIX, v_wanted);
+	if (v_actual < 0) {
+		dev_err(dev, "unable to reserve MSI-X vectors\n");
+		err = v_actual;
+		goto msix_err;
+	}
+
+	if (v_actual < v_wanted) {
+		dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
+			 v_wanted, v_actual);
+
+		if (v_actual < ICE_MIN_MSIX) {
+			/* error if we can't get minimum vectors */
+			pci_disable_msix(pf->pdev);
+			err = -ERANGE;
+			goto msix_err;
+		} else {
+			int v_remain = v_actual - v_other;
+
+			if (v_remain < ICE_MIN_LAN_TXRX_MSIX)
+				v_remain = ICE_MIN_LAN_TXRX_MSIX;
+
+			ice_reduce_msix_usage(pf, v_remain);
+
+			dev_notice(dev, "Enabled %d MSI-X vectors for LAN traffic.\n",
+				   pf->num_lan_msix);
+
+			if (ice_is_rdma_ena(pf))
+				dev_notice(dev, "Enabled %d MSI-X vectors for RDMA.\n",
+					   pf->num_rdma_msix);
+		}
+	}
+
+	return v_actual;
+
+msix_err:
+	devm_kfree(dev, pf->msix_entries);
+
+exit_err:
+	pf->num_rdma_msix = 0;
+	pf->num_lan_msix = 0;
+	return err;
+}
+
+/**
+ * ice_dis_msix - Disable MSI-X interrupt setup in OS
+ * @pf: board private structure
+ */
+static void ice_dis_msix(struct ice_pf *pf)
+{
+	pci_disable_msix(pf->pdev);
+	devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
+	pf->msix_entries = NULL;
+}
+
+/**
+ * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
+ * @pf: board private structure
+ */
+void ice_clear_interrupt_scheme(struct ice_pf *pf)
+{
+	ice_dis_msix(pf);
+
+	if (pf->irq_tracker) {
+		devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
+		pf->irq_tracker = NULL;
+	}
+}
+
+/**
+ * ice_init_interrupt_scheme - Determine proper interrupt scheme
+ * @pf: board private structure to initialize
+ */
+int ice_init_interrupt_scheme(struct ice_pf *pf)
+{
+	int vectors;
+
+	vectors = ice_ena_msix_range(pf);
+
+	if (vectors < 0)
+		return vectors;
+
+	/* set up vector assignment tracking */
+	pf->irq_tracker = devm_kzalloc(ice_pf_to_dev(pf),
+				       struct_size(pf->irq_tracker, list,
+						   vectors),
+				       GFP_KERNEL);
+	if (!pf->irq_tracker) {
+		ice_dis_msix(pf);
+		return -ENOMEM;
+	}
+
+	/* populate SW interrupts pool with number of OS granted IRQs. */
+	pf->num_avail_sw_msix = (u16)vectors;
+	pf->irq_tracker->num_entries = (u16)vectors;
+	pf->irq_tracker->end = pf->irq_tracker->num_entries;
+
+	return 0;
+}
diff --git a/drivers/net/ethernet/intel/ice/ice_irq.h b/drivers/net/ethernet/intel/ice/ice_irq.h
new file mode 100644
index 000000000000..82475162ab70
--- /dev/null
+++ b/drivers/net/ethernet/intel/ice/ice_irq.h
@@ -0,0 +1,10 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (C) 2023, Intel Corporation. */
+
+#ifndef _ICE_IRQ_H_
+#define _ICE_IRQ_H_
+
+int ice_init_interrupt_scheme(struct ice_pf *pf);
+void ice_clear_interrupt_scheme(struct ice_pf *pf);
+
+#endif
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index a1f7c8edc22f..c377bacc5e2e 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -3900,224 +3900,6 @@ static int ice_init_pf(struct ice_pf *pf)
 	return 0;
 }
 
-/**
- * ice_reduce_msix_usage - Reduce usage of MSI-X vectors
- * @pf: board private structure
- * @v_remain: number of remaining MSI-X vectors to be distributed
- *
- * Reduce the usage of MSI-X vectors when entire request cannot be fulfilled.
- * pf->num_lan_msix and pf->num_rdma_msix values are set based on number of
- * remaining vectors.
- */
-static void ice_reduce_msix_usage(struct ice_pf *pf, int v_remain)
-{
-	int v_rdma;
-
-	if (!ice_is_rdma_ena(pf)) {
-		pf->num_lan_msix = v_remain;
-		return;
-	}
-
-	/* RDMA needs at least 1 interrupt in addition to AEQ MSIX */
-	v_rdma = ICE_RDMA_NUM_AEQ_MSIX + 1;
-
-	if (v_remain < ICE_MIN_LAN_TXRX_MSIX + ICE_MIN_RDMA_MSIX) {
-		dev_warn(ice_pf_to_dev(pf), "Not enough MSI-X vectors to support RDMA.\n");
-		clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
-
-		pf->num_rdma_msix = 0;
-		pf->num_lan_msix = ICE_MIN_LAN_TXRX_MSIX;
-	} else if ((v_remain < ICE_MIN_LAN_TXRX_MSIX + v_rdma) ||
-		   (v_remain - v_rdma < v_rdma)) {
-		/* Support minimum RDMA and give remaining vectors to LAN MSIX */
-		pf->num_rdma_msix = ICE_MIN_RDMA_MSIX;
-		pf->num_lan_msix = v_remain - ICE_MIN_RDMA_MSIX;
-	} else {
-		/* Split remaining MSIX with RDMA after accounting for AEQ MSIX
-		 */
-		pf->num_rdma_msix = (v_remain - ICE_RDMA_NUM_AEQ_MSIX) / 2 +
-				    ICE_RDMA_NUM_AEQ_MSIX;
-		pf->num_lan_msix = v_remain - pf->num_rdma_msix;
-	}
-}
-
-/**
- * ice_ena_msix_range - Request a range of MSIX vectors from the OS
- * @pf: board private structure
- *
- * Compute the number of MSIX vectors wanted and request from the OS. Adjust
- * device usage if there are not enough vectors. Return the number of vectors
- * reserved or negative on failure.
- */
-static int ice_ena_msix_range(struct ice_pf *pf)
-{
-	int num_cpus, hw_num_msix, v_other, v_wanted, v_actual;
-	struct device *dev = ice_pf_to_dev(pf);
-	int err, i;
-
-	hw_num_msix = pf->hw.func_caps.common_cap.num_msix_vectors;
-	num_cpus = num_online_cpus();
-
-	/* LAN miscellaneous handler */
-	v_other = ICE_MIN_LAN_OICR_MSIX;
-
-	/* Flow Director */
-	if (test_bit(ICE_FLAG_FD_ENA, pf->flags))
-		v_other += ICE_FDIR_MSIX;
-
-	/* switchdev */
-	v_other += ICE_ESWITCH_MSIX;
-
-	v_wanted = v_other;
-
-	/* LAN traffic */
-	pf->num_lan_msix = num_cpus;
-	v_wanted += pf->num_lan_msix;
-
-	/* RDMA auxiliary driver */
-	if (ice_is_rdma_ena(pf)) {
-		pf->num_rdma_msix = num_cpus + ICE_RDMA_NUM_AEQ_MSIX;
-		v_wanted += pf->num_rdma_msix;
-	}
-
-	if (v_wanted > hw_num_msix) {
-		int v_remain;
-
-		dev_warn(dev, "not enough device MSI-X vectors. wanted = %d, available = %d\n",
-			 v_wanted, hw_num_msix);
-
-		if (hw_num_msix < ICE_MIN_MSIX) {
-			err = -ERANGE;
-			goto exit_err;
-		}
-
-		v_remain = hw_num_msix - v_other;
-		if (v_remain < ICE_MIN_LAN_TXRX_MSIX) {
-			v_other = ICE_MIN_MSIX - ICE_MIN_LAN_TXRX_MSIX;
-			v_remain = ICE_MIN_LAN_TXRX_MSIX;
-		}
-
-		ice_reduce_msix_usage(pf, v_remain);
-		v_wanted = pf->num_lan_msix + pf->num_rdma_msix + v_other;
-
-		dev_notice(dev, "Reducing request to %d MSI-X vectors for LAN traffic.\n",
-			   pf->num_lan_msix);
-		if (ice_is_rdma_ena(pf))
-			dev_notice(dev, "Reducing request to %d MSI-X vectors for RDMA.\n",
-				   pf->num_rdma_msix);
-	}
-
-	pf->msix_entries = devm_kcalloc(dev, v_wanted,
-					sizeof(*pf->msix_entries), GFP_KERNEL);
-	if (!pf->msix_entries) {
-		err = -ENOMEM;
-		goto exit_err;
-	}
-
-	for (i = 0; i < v_wanted; i++)
-		pf->msix_entries[i].entry = i;
-
-	/* actually reserve the vectors */
-	v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
-					 ICE_MIN_MSIX, v_wanted);
-	if (v_actual < 0) {
-		dev_err(dev, "unable to reserve MSI-X vectors\n");
-		err = v_actual;
-		goto msix_err;
-	}
-
-	if (v_actual < v_wanted) {
-		dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
-			 v_wanted, v_actual);
-
-		if (v_actual < ICE_MIN_MSIX) {
-			/* error if we can't get minimum vectors */
-			pci_disable_msix(pf->pdev);
-			err = -ERANGE;
-			goto msix_err;
-		} else {
-			int v_remain = v_actual - v_other;
-
-			if (v_remain < ICE_MIN_LAN_TXRX_MSIX)
-				v_remain = ICE_MIN_LAN_TXRX_MSIX;
-
-			ice_reduce_msix_usage(pf, v_remain);
-
-			dev_notice(dev, "Enabled %d MSI-X vectors for LAN traffic.\n",
-				   pf->num_lan_msix);
-
-			if (ice_is_rdma_ena(pf))
-				dev_notice(dev, "Enabled %d MSI-X vectors for RDMA.\n",
-					   pf->num_rdma_msix);
-		}
-	}
-
-	return v_actual;
-
-msix_err:
-	devm_kfree(dev, pf->msix_entries);
-
-exit_err:
-	pf->num_rdma_msix = 0;
-	pf->num_lan_msix = 0;
-	return err;
-}
-
-/**
- * ice_dis_msix - Disable MSI-X interrupt setup in OS
- * @pf: board private structure
- */
-static void ice_dis_msix(struct ice_pf *pf)
-{
-	pci_disable_msix(pf->pdev);
-	devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
-	pf->msix_entries = NULL;
-}
-
-/**
- * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
- * @pf: board private structure
- */
-static void ice_clear_interrupt_scheme(struct ice_pf *pf)
-{
-	ice_dis_msix(pf);
-
-	if (pf->irq_tracker) {
-		devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
-		pf->irq_tracker = NULL;
-	}
-}
-
-/**
- * ice_init_interrupt_scheme - Determine proper interrupt scheme
- * @pf: board private structure to initialize
- */
-static int ice_init_interrupt_scheme(struct ice_pf *pf)
-{
-	int vectors;
-
-	vectors = ice_ena_msix_range(pf);
-
-	if (vectors < 0)
-		return vectors;
-
-	/* set up vector assignment tracking */
-	pf->irq_tracker = devm_kzalloc(ice_pf_to_dev(pf),
-				       struct_size(pf->irq_tracker, list, vectors),
-				       GFP_KERNEL);
-	if (!pf->irq_tracker) {
-		ice_dis_msix(pf);
-		return -ENOMEM;
-	}
-
-	/* populate SW interrupts pool with number of OS granted IRQs. */
-	pf->num_avail_sw_msix = (u16)vectors;
-	pf->irq_tracker->num_entries = (u16)vectors;
-	pf->irq_tracker->end = pf->irq_tracker->num_entries;
-
-	return 0;
-}
-
 /**
  * ice_is_wol_supported - check if WoL is supported
  * @hw: pointer to hardware info
-- 
2.38.1

[PATCH net-next 2/8] ice: use pci_irq_vector helper function

[Tony Nguyen]

From: Piotr Raczynski <piotr.raczynski@intel.com>

Currently, driver gets interrupt number directly from ice_pf::msix_entries
array. Use helper function dedicated to do just that.

While at it use a variable to store interrupt number in
ice_free_irq_msix_misc instead of calling the helper function twice.

Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Piotr Raczynski <piotr.raczynski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/ice/ice_arfs.c |  2 +-
 drivers/net/ethernet/intel/ice/ice_lib.c  |  4 ++--
 drivers/net/ethernet/intel/ice/ice_main.c | 12 ++++++------
 drivers/net/ethernet/intel/ice/ice_ptp.c  |  2 +-
 drivers/net/ethernet/intel/ice/ice_xsk.c  |  2 +-
 5 files changed, 11 insertions(+), 11 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_arfs.c b/drivers/net/ethernet/intel/ice/ice_arfs.c
index fba178e07600..e81797344f5e 100644
--- a/drivers/net/ethernet/intel/ice/ice_arfs.c
+++ b/drivers/net/ethernet/intel/ice/ice_arfs.c
@@ -616,7 +616,7 @@ int ice_set_cpu_rx_rmap(struct ice_vsi *vsi)
 	base_idx = vsi->base_vector;
 	ice_for_each_q_vector(vsi, i)
 		if (irq_cpu_rmap_add(netdev->rx_cpu_rmap,
-				     pf->msix_entries[base_idx + i].vector)) {
+				     pci_irq_vector(pf->pdev, base_idx + i))) {
 			ice_free_cpu_rx_rmap(vsi);
 			return -EINVAL;
 		}
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index 450317dfcca7..79e1557f77e8 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -3056,7 +3056,7 @@ void ice_vsi_free_irq(struct ice_vsi *vsi)
 		u16 vector = i + base;
 		int irq_num;
 
-		irq_num = pf->msix_entries[vector].vector;
+		irq_num = pci_irq_vector(pf->pdev, vector);
 
 		/* free only the irqs that were actually requested */
 		if (!vsi->q_vectors[i] ||
@@ -3235,7 +3235,7 @@ void ice_vsi_dis_irq(struct ice_vsi *vsi)
 		return;
 
 	ice_for_each_q_vector(vsi, i)
-		synchronize_irq(pf->msix_entries[i + base].vector);
+		synchronize_irq(pci_irq_vector(pf->pdev, i + base));
 }
 
 /**
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index c377bacc5e2e..c103be660a9c 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -2501,7 +2501,7 @@ static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
 	for (vector = 0; vector < q_vectors; vector++) {
 		struct ice_q_vector *q_vector = vsi->q_vectors[vector];
 
-		irq_num = pf->msix_entries[base + vector].vector;
+		irq_num = pci_irq_vector(pf->pdev, base + vector);
 
 		if (q_vector->tx.tx_ring && q_vector->rx.rx_ring) {
 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
@@ -2557,7 +2557,7 @@ static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
 free_q_irqs:
 	while (vector) {
 		vector--;
-		irq_num = pf->msix_entries[base + vector].vector;
+		irq_num = pci_irq_vector(pf->pdev, base + vector);
 		if (!IS_ENABLED(CONFIG_RFS_ACCEL))
 			irq_set_affinity_notifier(irq_num, NULL);
 		irq_set_affinity_hint(irq_num, NULL);
@@ -3234,6 +3234,7 @@ static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
  */
 static void ice_free_irq_msix_misc(struct ice_pf *pf)
 {
+	int misc_irq_num = pci_irq_vector(pf->pdev, pf->oicr_idx);
 	struct ice_hw *hw = &pf->hw;
 
 	ice_dis_ctrlq_interrupts(hw);
@@ -3243,9 +3244,8 @@ static void ice_free_irq_msix_misc(struct ice_pf *pf)
 	ice_flush(hw);
 
 	if (pf->msix_entries) {
-		synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
-		devm_free_irq(ice_pf_to_dev(pf),
-			      pf->msix_entries[pf->oicr_idx].vector, pf);
+		synchronize_irq(misc_irq_num);
+		devm_free_irq(ice_pf_to_dev(pf), misc_irq_num, pf);
 	}
 
 	pf->num_avail_sw_msix += 1;
@@ -3317,7 +3317,7 @@ static int ice_req_irq_msix_misc(struct ice_pf *pf)
 	pf->oicr_idx = (u16)oicr_idx;
 
 	err = devm_request_threaded_irq(dev,
-					pf->msix_entries[pf->oicr_idx].vector,
+					pci_irq_vector(pf->pdev, pf->oicr_idx),
 					ice_misc_intr, ice_misc_intr_thread_fn,
 					0, pf->int_name, pf);
 	if (err) {
diff --git a/drivers/net/ethernet/intel/ice/ice_ptp.c b/drivers/net/ethernet/intel/ice/ice_ptp.c
index ac6f06f9a2ed..972d4f6fd615 100644
--- a/drivers/net/ethernet/intel/ice/ice_ptp.c
+++ b/drivers/net/ethernet/intel/ice/ice_ptp.c
@@ -911,7 +911,7 @@ ice_ptp_release_tx_tracker(struct ice_pf *pf, struct ice_ptp_tx *tx)
 	spin_unlock(&tx->lock);
 
 	/* wait for potentially outstanding interrupt to complete */
-	synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
+	synchronize_irq(pci_irq_vector(pf->pdev, pf->oicr_idx));
 
 	ice_ptp_flush_tx_tracker(pf, tx);
 
diff --git a/drivers/net/ethernet/intel/ice/ice_xsk.c b/drivers/net/ethernet/intel/ice/ice_xsk.c
index d1e489da7363..4102416d7a41 100644
--- a/drivers/net/ethernet/intel/ice/ice_xsk.c
+++ b/drivers/net/ethernet/intel/ice/ice_xsk.c
@@ -107,7 +107,7 @@ ice_qvec_dis_irq(struct ice_vsi *vsi, struct ice_rx_ring *rx_ring,
 
 		wr32(hw, GLINT_DYN_CTL(q_vector->reg_idx), 0);
 		ice_flush(hw);
-		synchronize_irq(pf->msix_entries[v_idx + base].vector);
+		synchronize_irq(pci_irq_vector(pf->pdev, v_idx + base));
 	}
 }
 
-- 
2.38.1

[PATCH net-next 3/8] ice: use preferred MSIX allocation api

[Tony Nguyen]

From: Piotr Raczynski <piotr.raczynski@intel.com>

Move away from using pci_enable_msix_range/pci_disable_msix and use
pci_alloc_irq_vectors/pci_free_irq_vectors instead.

As a result stop tracking msix_entries since with newer API entries are
handled by MSIX core. However, due to current design of communication
with RDMA driver which accesses ice_pf::msix_entries directly, keep
using the array just for RDMA driver use.

Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Piotr Raczynski <piotr.raczynski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/ice/ice_idc.c  | 29 ++++++++++++++--
 drivers/net/ethernet/intel/ice/ice_irq.c  | 40 +++++------------------
 drivers/net/ethernet/intel/ice/ice_main.c |  6 ++--
 3 files changed, 37 insertions(+), 38 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_idc.c b/drivers/net/ethernet/intel/ice/ice_idc.c
index e6bc2285071e..1000759505d7 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc.c
+++ b/drivers/net/ethernet/intel/ice/ice_idc.c
@@ -235,14 +235,33 @@ EXPORT_SYMBOL_GPL(ice_get_qos_params);
 static int ice_reserve_rdma_qvector(struct ice_pf *pf)
 {
 	if (ice_is_rdma_ena(pf)) {
-		int index;
+		int index, i;
 
 		index = ice_get_res(pf, pf->irq_tracker, pf->num_rdma_msix,
 				    ICE_RES_RDMA_VEC_ID);
 		if (index < 0)
 			return index;
+
+		pf->msix_entries = kcalloc(pf->num_rdma_msix,
+					   sizeof(*pf->msix_entries),
+						  GFP_KERNEL);
+		if (!pf->msix_entries) {
+			ice_free_res(pf->irq_tracker, pf->rdma_base_vector,
+				     ICE_RES_RDMA_VEC_ID);
+			return -ENOMEM;
+		}
+
 		pf->num_avail_sw_msix -= pf->num_rdma_msix;
-		pf->rdma_base_vector = (u16)index;
+
+		/* RDMA is the only user of pf->msix_entries array */
+		pf->rdma_base_vector = 0;
+
+		for (i = 0; i < pf->num_rdma_msix; i++, index++) {
+			struct msix_entry *entry = &pf->msix_entries[i];
+
+			entry->entry = index;
+			entry->vector = pci_irq_vector(pf->pdev, index);
+		}
 	}
 	return 0;
 }
@@ -253,6 +272,12 @@ static int ice_reserve_rdma_qvector(struct ice_pf *pf)
  */
 static void ice_free_rdma_qvector(struct ice_pf *pf)
 {
+	if (!pf->msix_entries)
+		return;
+
+	kfree(pf->msix_entries);
+	pf->msix_entries = NULL;
+
 	pf->num_avail_sw_msix -= pf->num_rdma_msix;
 	ice_free_res(pf->irq_tracker, pf->rdma_base_vector,
 		     ICE_RES_RDMA_VEC_ID);
diff --git a/drivers/net/ethernet/intel/ice/ice_irq.c b/drivers/net/ethernet/intel/ice/ice_irq.c
index 1fc7daec9732..f61be5d76373 100644
--- a/drivers/net/ethernet/intel/ice/ice_irq.c
+++ b/drivers/net/ethernet/intel/ice/ice_irq.c
@@ -59,7 +59,7 @@ static int ice_ena_msix_range(struct ice_pf *pf)
 {
 	int num_cpus, hw_num_msix, v_other, v_wanted, v_actual;
 	struct device *dev = ice_pf_to_dev(pf);
-	int err, i;
+	int err;
 
 	hw_num_msix = pf->hw.func_caps.common_cap.num_msix_vectors;
 	num_cpus = num_online_cpus();
@@ -113,23 +113,13 @@ static int ice_ena_msix_range(struct ice_pf *pf)
 				   pf->num_rdma_msix);
 	}
 
-	pf->msix_entries = devm_kcalloc(dev, v_wanted,
-					sizeof(*pf->msix_entries), GFP_KERNEL);
-	if (!pf->msix_entries) {
-		err = -ENOMEM;
-		goto exit_err;
-	}
-
-	for (i = 0; i < v_wanted; i++)
-		pf->msix_entries[i].entry = i;
-
 	/* actually reserve the vectors */
-	v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
-					 ICE_MIN_MSIX, v_wanted);
+	v_actual = pci_alloc_irq_vectors(pf->pdev, ICE_MIN_MSIX, v_wanted,
+					 PCI_IRQ_MSIX);
 	if (v_actual < 0) {
 		dev_err(dev, "unable to reserve MSI-X vectors\n");
 		err = v_actual;
-		goto msix_err;
+		goto exit_err;
 	}
 
 	if (v_actual < v_wanted) {
@@ -138,9 +128,9 @@ static int ice_ena_msix_range(struct ice_pf *pf)
 
 		if (v_actual < ICE_MIN_MSIX) {
 			/* error if we can't get minimum vectors */
-			pci_disable_msix(pf->pdev);
+			pci_free_irq_vectors(pf->pdev);
 			err = -ERANGE;
-			goto msix_err;
+			goto exit_err;
 		} else {
 			int v_remain = v_actual - v_other;
 
@@ -160,33 +150,19 @@ static int ice_ena_msix_range(struct ice_pf *pf)
 
 	return v_actual;
 
-msix_err:
-	devm_kfree(dev, pf->msix_entries);
-
 exit_err:
 	pf->num_rdma_msix = 0;
 	pf->num_lan_msix = 0;
 	return err;
 }
 
-/**
- * ice_dis_msix - Disable MSI-X interrupt setup in OS
- * @pf: board private structure
- */
-static void ice_dis_msix(struct ice_pf *pf)
-{
-	pci_disable_msix(pf->pdev);
-	devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
-	pf->msix_entries = NULL;
-}
-
 /**
  * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
  * @pf: board private structure
  */
 void ice_clear_interrupt_scheme(struct ice_pf *pf)
 {
-	ice_dis_msix(pf);
+	pci_free_irq_vectors(pf->pdev);
 
 	if (pf->irq_tracker) {
 		devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
@@ -213,7 +189,7 @@ int ice_init_interrupt_scheme(struct ice_pf *pf)
 						   vectors),
 				       GFP_KERNEL);
 	if (!pf->irq_tracker) {
-		ice_dis_msix(pf);
+		pci_free_irq_vectors(pf->pdev);
 		return -ENOMEM;
 	}
 
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index c103be660a9c..ce8cd49ae10c 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -3243,10 +3243,8 @@ static void ice_free_irq_msix_misc(struct ice_pf *pf)
 	wr32(hw, PFINT_OICR_ENA, 0);
 	ice_flush(hw);
 
-	if (pf->msix_entries) {
-		synchronize_irq(misc_irq_num);
-		devm_free_irq(ice_pf_to_dev(pf), misc_irq_num, pf);
-	}
+	synchronize_irq(misc_irq_num);
+	devm_free_irq(ice_pf_to_dev(pf), misc_irq_num, pf);
 
 	pf->num_avail_sw_msix += 1;
 	ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
-- 
2.38.1

[PATCH net-next 4/8] ice: refactor VF control VSI interrupt handling

[Tony Nguyen]

From: Piotr Raczynski <piotr.raczynski@intel.com>

All VF control VSIs share the same interrupt vector. Currently, a helper
function dedicated for that directly sets ice_vsi::base_vector.

Use helper that returns pointer to first found VF control VSI instead.

Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Signed-off-by: Piotr Raczynski <piotr.raczynski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/ice/ice_lib.c    | 76 ++++-----------------
 drivers/net/ethernet/intel/ice/ice_vf_lib.c | 32 +++++++++
 drivers/net/ethernet/intel/ice/ice_vf_lib.h |  7 ++
 3 files changed, 52 insertions(+), 63 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index 79e1557f77e8..f8c9c2c20e6d 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -1473,36 +1473,6 @@ ice_get_res(struct ice_pf *pf, struct ice_res_tracker *res, u16 needed, u16 id)
 	return ice_search_res(res, needed, id);
 }
 
-/**
- * ice_get_vf_ctrl_res - Get VF control VSI resource
- * @pf: pointer to the PF structure
- * @vsi: the VSI to allocate a resource for
- *
- * Look up whether another VF has already allocated the control VSI resource.
- * If so, re-use this resource so that we share it among all VFs.
- *
- * Otherwise, allocate the resource and return it.
- */
-static int ice_get_vf_ctrl_res(struct ice_pf *pf, struct ice_vsi *vsi)
-{
-	struct ice_vf *vf;
-	unsigned int bkt;
-	int base;
-
-	rcu_read_lock();
-	ice_for_each_vf_rcu(pf, bkt, vf) {
-		if (vf != vsi->vf && vf->ctrl_vsi_idx != ICE_NO_VSI) {
-			base = pf->vsi[vf->ctrl_vsi_idx]->base_vector;
-			rcu_read_unlock();
-			return base;
-		}
-	}
-	rcu_read_unlock();
-
-	return ice_get_res(pf, pf->irq_tracker, vsi->num_q_vectors,
-			   ICE_RES_VF_CTRL_VEC_ID);
-}
-
 /**
  * ice_vsi_setup_vector_base - Set up the base vector for the given VSI
  * @vsi: ptr to the VSI
@@ -1536,7 +1506,14 @@ static int ice_vsi_setup_vector_base(struct ice_vsi *vsi)
 	num_q_vectors = vsi->num_q_vectors;
 	/* reserve slots from OS requested IRQs */
 	if (vsi->type == ICE_VSI_CTRL && vsi->vf) {
-		base = ice_get_vf_ctrl_res(pf, vsi);
+		struct ice_vsi *ctrl_vsi = ice_get_vf_ctrl_vsi(pf, vsi);
+
+		if (ctrl_vsi)
+			base = ctrl_vsi->base_vector;
+		else
+			base = ice_get_res(pf, pf->irq_tracker,
+					   vsi->num_q_vectors,
+					   ICE_RES_VF_CTRL_VEC_ID);
 	} else {
 		base = ice_get_res(pf, pf->irq_tracker, num_q_vectors,
 				   vsi->idx);
@@ -2611,37 +2588,6 @@ static void ice_set_agg_vsi(struct ice_vsi *vsi)
 		vsi->agg_node->num_vsis);
 }
 
-/**
- * ice_free_vf_ctrl_res - Free the VF control VSI resource
- * @pf: pointer to PF structure
- * @vsi: the VSI to free resources for
- *
- * Check if the VF control VSI resource is still in use. If no VF is using it
- * any more, release the VSI resource. Otherwise, leave it to be cleaned up
- * once no other VF uses it.
- */
-static void ice_free_vf_ctrl_res(struct ice_pf *pf,  struct ice_vsi *vsi)
-{
-	struct ice_vf *vf;
-	unsigned int bkt;
-
-	rcu_read_lock();
-	ice_for_each_vf_rcu(pf, bkt, vf) {
-		if (vf != vsi->vf && vf->ctrl_vsi_idx != ICE_NO_VSI) {
-			rcu_read_unlock();
-			return;
-		}
-	}
-	rcu_read_unlock();
-
-	/* No other VFs left that have control VSI. It is now safe to reclaim
-	 * SW interrupts back to the common pool.
-	 */
-	ice_free_res(pf->irq_tracker, vsi->base_vector,
-		     ICE_RES_VF_CTRL_VEC_ID);
-	pf->num_avail_sw_msix += vsi->num_q_vectors;
-}
-
 static int ice_vsi_cfg_tc_lan(struct ice_pf *pf, struct ice_vsi *vsi)
 {
 	u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
@@ -2916,7 +2862,11 @@ void ice_vsi_decfg(struct ice_vsi *vsi)
 	 * cleared in the same manner.
 	 */
 	if (vsi->type == ICE_VSI_CTRL && vsi->vf) {
-		ice_free_vf_ctrl_res(pf, vsi);
+		struct ice_vsi *ctrl_vsi = ice_get_vf_ctrl_vsi(pf, vsi);
+
+		if (!ctrl_vsi)
+			ice_free_res(pf->irq_tracker, vsi->base_vector,
+				     ICE_RES_VF_CTRL_VEC_ID);
 	} else if (vsi->type != ICE_VSI_VF) {
 		/* reclaim SW interrupts back to the common pool */
 		ice_free_res(pf->irq_tracker, vsi->base_vector, vsi->idx);
diff --git a/drivers/net/ethernet/intel/ice/ice_vf_lib.c b/drivers/net/ethernet/intel/ice/ice_vf_lib.c
index 89fd6982df09..68142facc85d 100644
--- a/drivers/net/ethernet/intel/ice/ice_vf_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_vf_lib.c
@@ -1310,3 +1310,35 @@ void ice_vf_set_initialized(struct ice_vf *vf)
 	set_bit(ICE_VF_STATE_INIT, vf->vf_states);
 	memset(&vf->vlan_v2_caps, 0, sizeof(vf->vlan_v2_caps));
 }
+
+/**
+ * ice_get_vf_ctrl_vsi - Get first VF control VSI pointer
+ * @pf: the PF private structure
+ * @vsi: pointer to the VSI
+ *
+ * Return first found VF control VSI other than the vsi
+ * passed by parameter. This function is used to determine
+ * whether new resources have to be allocated for control VSI
+ * or they can be shared with existing one.
+ *
+ * Return found VF control VSI pointer other itself. Return
+ * NULL Otherwise.
+ *
+ */
+struct ice_vsi *ice_get_vf_ctrl_vsi(struct ice_pf *pf, struct ice_vsi *vsi)
+{
+	struct ice_vsi *ctrl_vsi = NULL;
+	struct ice_vf *vf;
+	unsigned int bkt;
+
+	rcu_read_lock();
+	ice_for_each_vf_rcu(pf, bkt, vf) {
+		if (vf != vsi->vf && vf->ctrl_vsi_idx != ICE_NO_VSI) {
+			ctrl_vsi = pf->vsi[vf->ctrl_vsi_idx];
+			break;
+		}
+	}
+
+	rcu_read_unlock();
+	return ctrl_vsi;
+}
diff --git a/drivers/net/ethernet/intel/ice/ice_vf_lib.h b/drivers/net/ethernet/intel/ice/ice_vf_lib.h
index e3cda6fb71ab..48fea6fa0362 100644
--- a/drivers/net/ethernet/intel/ice/ice_vf_lib.h
+++ b/drivers/net/ethernet/intel/ice/ice_vf_lib.h
@@ -226,6 +226,7 @@ int
 ice_vf_clear_vsi_promisc(struct ice_vf *vf, struct ice_vsi *vsi, u8 promisc_m);
 int ice_reset_vf(struct ice_vf *vf, u32 flags);
 void ice_reset_all_vfs(struct ice_pf *pf);
+struct ice_vsi *ice_get_vf_ctrl_vsi(struct ice_pf *pf, struct ice_vsi *vsi);
 #else /* CONFIG_PCI_IOV */
 static inline struct ice_vf *ice_get_vf_by_id(struct ice_pf *pf, u16 vf_id)
 {
@@ -290,6 +291,12 @@ static inline int ice_reset_vf(struct ice_vf *vf, u32 flags)
 static inline void ice_reset_all_vfs(struct ice_pf *pf)
 {
 }
+
+static inline struct ice_vsi *
+ice_get_vf_ctrl_vsi(struct ice_pf *pf, struct ice_vsi *vsi)
+{
+	return NULL;
+}
 #endif /* !CONFIG_PCI_IOV */
 
 #endif /* _ICE_VF_LIB_H_ */
-- 
2.38.1

[PATCH net-next 5/8] ice: remove redundant SRIOV code

[Tony Nguyen]

From: Piotr Raczynski <piotr.raczynski@intel.com>

Remove redundant code from ice_get_max_valid_res_idx that has no effect.
ice_pf::irq_tracker is initialized during driver probe, there is no reason
to check it again. Also it is not possible for pf::sriov_base_vector to be
lower than the tracker length, remove WARN_ON that will never happen.

Get rid of ice_get_max_valid_res_idx helper function completely since it
can never return negative value.

Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Signed-off-by: Piotr Raczynski <piotr.raczynski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/ice/ice_sriov.c | 36 ----------------------
 1 file changed, 36 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_sriov.c b/drivers/net/ethernet/intel/ice/ice_sriov.c
index f1dca59bd844..65f971b74717 100644
--- a/drivers/net/ethernet/intel/ice/ice_sriov.c
+++ b/drivers/net/ethernet/intel/ice/ice_sriov.c
@@ -135,18 +135,9 @@ static void ice_dis_vf_mappings(struct ice_vf *vf)
  */
 static int ice_sriov_free_msix_res(struct ice_pf *pf)
 {
-	struct ice_res_tracker *res;
-
 	if (!pf)
 		return -EINVAL;
 
-	res = pf->irq_tracker;
-	if (!res)
-		return -EINVAL;
-
-	/* give back irq_tracker resources used */
-	WARN_ON(pf->sriov_base_vector < res->num_entries);
-
 	pf->sriov_base_vector = 0;
 
 	return 0;
@@ -409,29 +400,6 @@ int ice_calc_vf_reg_idx(struct ice_vf *vf, struct ice_q_vector *q_vector)
 		q_vector->v_idx + 1;
 }
 
-/**
- * ice_get_max_valid_res_idx - Get the max valid resource index
- * @res: pointer to the resource to find the max valid index for
- *
- * Start from the end of the ice_res_tracker and return right when we find the
- * first res->list entry with the ICE_RES_VALID_BIT set. This function is only
- * valid for SR-IOV because it is the only consumer that manipulates the
- * res->end and this is always called when res->end is set to res->num_entries.
- */
-static int ice_get_max_valid_res_idx(struct ice_res_tracker *res)
-{
-	int i;
-
-	if (!res)
-		return -EINVAL;
-
-	for (i = res->num_entries - 1; i >= 0; i--)
-		if (res->list[i] & ICE_RES_VALID_BIT)
-			return i;
-
-	return 0;
-}
-
 /**
  * ice_sriov_set_msix_res - Set any used MSIX resources
  * @pf: pointer to PF structure
@@ -490,7 +458,6 @@ static int ice_sriov_set_msix_res(struct ice_pf *pf, u16 num_msix_needed)
  */
 static int ice_set_per_vf_res(struct ice_pf *pf, u16 num_vfs)
 {
-	int max_valid_res_idx = ice_get_max_valid_res_idx(pf->irq_tracker);
 	u16 num_msix_per_vf, num_txq, num_rxq, avail_qs;
 	int msix_avail_per_vf, msix_avail_for_sriov;
 	struct device *dev = ice_pf_to_dev(pf);
@@ -501,9 +468,6 @@ static int ice_set_per_vf_res(struct ice_pf *pf, u16 num_vfs)
 	if (!num_vfs)
 		return -EINVAL;
 
-	if (max_valid_res_idx < 0)
-		return -ENOSPC;
-
 	/* determine MSI-X resources per VF */
 	msix_avail_for_sriov = pf->hw.func_caps.common_cap.num_msix_vectors -
 		pf->irq_tracker->num_entries;
-- 
2.38.1

[PATCH net-next 6/8] ice: add individual interrupt allocation

[Tony Nguyen]

From: Piotr Raczynski <piotr.raczynski@intel.com>

Currently interrupt allocations, depending on a feature are distributed
in batches. Also, after allocation there is a series of operations that
distributes per irq settings through that batch of interrupts.

Although driver does not yet support dynamic interrupt allocation, keep
allocated interrupts in a pool and add allocation abstraction logic to
make code more flexible. Keep per interrupt information in the
ice_q_vector structure, which yields ice_vsi::base_vector redundant.
Also, as a result there are a few functions that can be removed.

Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Piotr Raczynski <piotr.raczynski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/ice/ice.h         |  11 +-
 drivers/net/ethernet/intel/ice/ice_arfs.c    |   5 +-
 drivers/net/ethernet/intel/ice/ice_base.c    |  41 +++-
 drivers/net/ethernet/intel/ice/ice_ethtool.c |   2 +-
 drivers/net/ethernet/intel/ice/ice_idc.c     |  45 ++--
 drivers/net/ethernet/intel/ice/ice_irq.c     |  46 +++-
 drivers/net/ethernet/intel/ice/ice_irq.h     |   3 +
 drivers/net/ethernet/intel/ice/ice_lib.c     | 225 ++-----------------
 drivers/net/ethernet/intel/ice/ice_lib.h     |   4 +-
 drivers/net/ethernet/intel/ice/ice_main.c    |  44 ++--
 drivers/net/ethernet/intel/ice/ice_ptp.c     |   2 +-
 drivers/net/ethernet/intel/ice/ice_sriov.c   |   2 +-
 drivers/net/ethernet/intel/ice/ice_xsk.c     |   5 +-
 13 files changed, 156 insertions(+), 279 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
index 4dc4c6bec081..d8dde291491e 100644
--- a/drivers/net/ethernet/intel/ice/ice.h
+++ b/drivers/net/ethernet/intel/ice/ice.h
@@ -105,10 +105,6 @@
 #define ICE_Q_WAIT_MAX_RETRY	(5 * ICE_Q_WAIT_RETRY_LIMIT)
 #define ICE_MAX_LG_RSS_QS	256
 #define ICE_RES_VALID_BIT	0x8000
-#define ICE_RES_MISC_VEC_ID	(ICE_RES_VALID_BIT - 1)
-#define ICE_RES_RDMA_VEC_ID	(ICE_RES_MISC_VEC_ID - 1)
-/* All VF control VSIs share the same IRQ, so assign a unique ID for them */
-#define ICE_RES_VF_CTRL_VEC_ID	(ICE_RES_RDMA_VEC_ID - 1)
 #define ICE_INVAL_Q_INDEX	0xffff
 
 #define ICE_MAX_RXQS_PER_TC		256	/* Used when setting VSI context per TC Rx queues */
@@ -349,7 +345,6 @@ struct ice_vsi {
 	u32 rx_buf_failed;
 	u32 rx_page_failed;
 	u16 num_q_vectors;
-	u16 base_vector;		/* IRQ base for OS reserved vectors */
 	enum ice_vsi_type type;
 	u16 vsi_num;			/* HW (absolute) index of this VSI */
 	u16 idx;			/* software index in pf->vsi[] */
@@ -480,6 +475,7 @@ struct ice_q_vector {
 	char name[ICE_INT_NAME_STR_LEN];
 
 	u16 total_events;	/* net_dim(): number of interrupts processed */
+	struct msi_map irq;
 } ____cacheline_internodealigned_in_smp;
 
 enum ice_pf_flags {
@@ -584,8 +580,7 @@ struct ice_pf {
 
 	u32 hw_csum_rx_error;
 	u32 oicr_err_reg;
-	u16 oicr_idx;		/* Other interrupt cause MSIX vector index */
-	u16 num_avail_sw_msix;	/* remaining MSIX SW vectors left unclaimed */
+	struct msi_map oicr_irq;	/* Other interrupt cause MSIX vector */
 	u16 max_pf_txqs;	/* Total Tx queues PF wide */
 	u16 max_pf_rxqs;	/* Total Rx queues PF wide */
 	u16 num_lan_msix;	/* Total MSIX vectors for base driver */
@@ -671,7 +666,7 @@ ice_irq_dynamic_ena(struct ice_hw *hw, struct ice_vsi *vsi,
 		    struct ice_q_vector *q_vector)
 {
 	u32 vector = (vsi && q_vector) ? q_vector->reg_idx :
-				((struct ice_pf *)hw->back)->oicr_idx;
+				((struct ice_pf *)hw->back)->oicr_irq.index;
 	int itr = ICE_ITR_NONE;
 	u32 val;
 
diff --git a/drivers/net/ethernet/intel/ice/ice_arfs.c b/drivers/net/ethernet/intel/ice/ice_arfs.c
index e81797344f5e..cca0e753f38f 100644
--- a/drivers/net/ethernet/intel/ice/ice_arfs.c
+++ b/drivers/net/ethernet/intel/ice/ice_arfs.c
@@ -596,7 +596,7 @@ int ice_set_cpu_rx_rmap(struct ice_vsi *vsi)
 {
 	struct net_device *netdev;
 	struct ice_pf *pf;
-	int base_idx, i;
+	int i;
 
 	if (!vsi || vsi->type != ICE_VSI_PF)
 		return 0;
@@ -613,10 +613,9 @@ int ice_set_cpu_rx_rmap(struct ice_vsi *vsi)
 	if (unlikely(!netdev->rx_cpu_rmap))
 		return -EINVAL;
 
-	base_idx = vsi->base_vector;
 	ice_for_each_q_vector(vsi, i)
 		if (irq_cpu_rmap_add(netdev->rx_cpu_rmap,
-				     pci_irq_vector(pf->pdev, base_idx + i))) {
+				     vsi->q_vectors[i]->irq.virq)) {
 			ice_free_cpu_rx_rmap(vsi);
 			return -EINVAL;
 		}
diff --git a/drivers/net/ethernet/intel/ice/ice_base.c b/drivers/net/ethernet/intel/ice/ice_base.c
index 1911d644dfa8..7dd7a0f32471 100644
--- a/drivers/net/ethernet/intel/ice/ice_base.c
+++ b/drivers/net/ethernet/intel/ice/ice_base.c
@@ -105,8 +105,7 @@ static int ice_vsi_alloc_q_vector(struct ice_vsi *vsi, u16 v_idx)
 	struct ice_q_vector *q_vector;
 
 	/* allocate q_vector */
-	q_vector = devm_kzalloc(ice_pf_to_dev(pf), sizeof(*q_vector),
-				GFP_KERNEL);
+	q_vector = kzalloc(sizeof(*q_vector), GFP_KERNEL);
 	if (!q_vector)
 		return -ENOMEM;
 
@@ -118,9 +117,31 @@ static int ice_vsi_alloc_q_vector(struct ice_vsi *vsi, u16 v_idx)
 	q_vector->rx.itr_mode = ITR_DYNAMIC;
 	q_vector->tx.type = ICE_TX_CONTAINER;
 	q_vector->rx.type = ICE_RX_CONTAINER;
+	q_vector->irq.index = -ENOENT;
 
-	if (vsi->type == ICE_VSI_VF)
+	if (vsi->type == ICE_VSI_VF) {
+		q_vector->reg_idx = ice_calc_vf_reg_idx(vsi->vf, q_vector);
 		goto out;
+	} else if (vsi->type == ICE_VSI_CTRL && vsi->vf) {
+		struct ice_vsi *ctrl_vsi = ice_get_vf_ctrl_vsi(pf, vsi);
+
+		if (ctrl_vsi) {
+			if (unlikely(!ctrl_vsi->q_vectors))
+				return -ENOENT;
+			q_vector->irq = ctrl_vsi->q_vectors[0]->irq;
+			goto skip_alloc;
+		}
+	}
+
+	q_vector->irq = ice_alloc_irq(pf);
+	if (q_vector->irq.index < 0) {
+		kfree(q_vector);
+		return -ENOMEM;
+	}
+
+skip_alloc:
+	q_vector->reg_idx = q_vector->irq.index;
+
 	/* only set affinity_mask if the CPU is online */
 	if (cpu_online(v_idx))
 		cpumask_set_cpu(v_idx, &q_vector->affinity_mask);
@@ -168,7 +189,19 @@ static void ice_free_q_vector(struct ice_vsi *vsi, int v_idx)
 	if (vsi->netdev)
 		netif_napi_del(&q_vector->napi);
 
-	devm_kfree(dev, q_vector);
+	/* release MSIX interrupt if q_vector had interrupt allocated */
+	if (q_vector->irq.index < 0)
+		goto free_q_vector;
+
+	/* only free last VF ctrl vsi interrupt */
+	if (vsi->type == ICE_VSI_CTRL && vsi->vf &&
+	    ice_get_vf_ctrl_vsi(pf, vsi))
+		goto free_q_vector;
+
+	ice_free_irq(pf, q_vector->irq);
+
+free_q_vector:
+	kfree(q_vector);
 	vsi->q_vectors[v_idx] = NULL;
 }
 
diff --git a/drivers/net/ethernet/intel/ice/ice_ethtool.c b/drivers/net/ethernet/intel/ice/ice_ethtool.c
index f86e814354a3..8407c7175cf6 100644
--- a/drivers/net/ethernet/intel/ice/ice_ethtool.c
+++ b/drivers/net/ethernet/intel/ice/ice_ethtool.c
@@ -956,7 +956,7 @@ static u64 ice_intr_test(struct net_device *netdev)
 
 	netdev_info(netdev, "interrupt test\n");
 
-	wr32(&pf->hw, GLINT_DYN_CTL(pf->oicr_idx),
+	wr32(&pf->hw, GLINT_DYN_CTL(pf->oicr_irq.index),
 	     GLINT_DYN_CTL_SW_ITR_INDX_M |
 	     GLINT_DYN_CTL_INTENA_MSK_M |
 	     GLINT_DYN_CTL_SWINT_TRIG_M);
diff --git a/drivers/net/ethernet/intel/ice/ice_idc.c b/drivers/net/ethernet/intel/ice/ice_idc.c
index 1000759505d7..bc016bb4440c 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc.c
+++ b/drivers/net/ethernet/intel/ice/ice_idc.c
@@ -229,38 +229,33 @@ void ice_get_qos_params(struct ice_pf *pf, struct iidc_qos_params *qos)
 EXPORT_SYMBOL_GPL(ice_get_qos_params);
 
 /**
- * ice_reserve_rdma_qvector - Reserve vector resources for RDMA driver
+ * ice_alloc_rdma_qvectors - Allocate vector resources for RDMA driver
  * @pf: board private structure to initialize
  */
-static int ice_reserve_rdma_qvector(struct ice_pf *pf)
+static int ice_alloc_rdma_qvectors(struct ice_pf *pf)
 {
 	if (ice_is_rdma_ena(pf)) {
-		int index, i;
-
-		index = ice_get_res(pf, pf->irq_tracker, pf->num_rdma_msix,
-				    ICE_RES_RDMA_VEC_ID);
-		if (index < 0)
-			return index;
+		int i;
 
 		pf->msix_entries = kcalloc(pf->num_rdma_msix,
 					   sizeof(*pf->msix_entries),
 						  GFP_KERNEL);
-		if (!pf->msix_entries) {
-			ice_free_res(pf->irq_tracker, pf->rdma_base_vector,
-				     ICE_RES_RDMA_VEC_ID);
+		if (!pf->msix_entries)
 			return -ENOMEM;
-		}
-
-		pf->num_avail_sw_msix -= pf->num_rdma_msix;
 
 		/* RDMA is the only user of pf->msix_entries array */
 		pf->rdma_base_vector = 0;
 
-		for (i = 0; i < pf->num_rdma_msix; i++, index++) {
+		for (i = 0; i < pf->num_rdma_msix; i++) {
 			struct msix_entry *entry = &pf->msix_entries[i];
+			struct msi_map map;
 
-			entry->entry = index;
-			entry->vector = pci_irq_vector(pf->pdev, index);
+			map = ice_alloc_irq(pf);
+			if (map.index < 0)
+				break;
+
+			entry->entry = map.index;
+			entry->vector = map.virq;
 		}
 	}
 	return 0;
@@ -272,15 +267,21 @@ static int ice_reserve_rdma_qvector(struct ice_pf *pf)
  */
 static void ice_free_rdma_qvector(struct ice_pf *pf)
 {
+	int i;
+
 	if (!pf->msix_entries)
 		return;
 
+	for (i = 0; i < pf->num_rdma_msix; i++) {
+		struct msi_map map;
+
+		map.index = pf->msix_entries[i].entry;
+		map.virq = pf->msix_entries[i].vector;
+		ice_free_irq(pf, map);
+	}
+
 	kfree(pf->msix_entries);
 	pf->msix_entries = NULL;
-
-	pf->num_avail_sw_msix -= pf->num_rdma_msix;
-	ice_free_res(pf->irq_tracker, pf->rdma_base_vector,
-		     ICE_RES_RDMA_VEC_ID);
 }
 
 /**
@@ -382,7 +383,7 @@ int ice_init_rdma(struct ice_pf *pf)
 	}
 
 	/* Reserve vector resources */
-	ret = ice_reserve_rdma_qvector(pf);
+	ret = ice_alloc_rdma_qvectors(pf);
 	if (ret < 0) {
 		dev_err(dev, "failed to reserve vectors for RDMA\n");
 		goto err_reserve_rdma_qvector;
diff --git a/drivers/net/ethernet/intel/ice/ice_irq.c b/drivers/net/ethernet/intel/ice/ice_irq.c
index f61be5d76373..ca1a1de26766 100644
--- a/drivers/net/ethernet/intel/ice/ice_irq.c
+++ b/drivers/net/ethernet/intel/ice/ice_irq.c
@@ -194,9 +194,53 @@ int ice_init_interrupt_scheme(struct ice_pf *pf)
 	}
 
 	/* populate SW interrupts pool with number of OS granted IRQs. */
-	pf->num_avail_sw_msix = (u16)vectors;
 	pf->irq_tracker->num_entries = (u16)vectors;
 	pf->irq_tracker->end = pf->irq_tracker->num_entries;
 
 	return 0;
 }
+
+/**
+ * ice_alloc_irq - Allocate new interrupt vector
+ * @pf: board private structure
+ *
+ * Allocate new interrupt vector for a given owner id.
+ * return struct msi_map with interrupt details and track
+ * allocated interrupt appropriately.
+ *
+ * This function mimics individual interrupt allocation,
+ * even interrupts are actually already allocated with
+ * pci_alloc_irq_vectors. Individual allocation helps
+ * to track interrupts and simplifies interrupt related
+ * handling.
+ *
+ * On failure, return map with negative .index. The caller
+ * is expected to check returned map index.
+ *
+ */
+struct msi_map ice_alloc_irq(struct ice_pf *pf)
+{
+	struct msi_map map = { .index = -ENOENT };
+	int entry;
+
+	entry = ice_get_res(pf, pf->irq_tracker);
+	if (entry < 0)
+		return map;
+
+	map.index = entry;
+	map.virq = pci_irq_vector(pf->pdev, map.index);
+
+	return map;
+}
+
+/**
+ * ice_free_irq - Free interrupt vector
+ * @pf: board private structure
+ * @map: map with interrupt details
+ *
+ * Remove allocated interrupt from the interrupt tracker
+ */
+void ice_free_irq(struct ice_pf *pf, struct msi_map map)
+{
+	ice_free_res(pf->irq_tracker, map.index);
+}
diff --git a/drivers/net/ethernet/intel/ice/ice_irq.h b/drivers/net/ethernet/intel/ice/ice_irq.h
index 82475162ab70..26e80dfe22b5 100644
--- a/drivers/net/ethernet/intel/ice/ice_irq.h
+++ b/drivers/net/ethernet/intel/ice/ice_irq.h
@@ -7,4 +7,7 @@
 int ice_init_interrupt_scheme(struct ice_pf *pf);
 void ice_clear_interrupt_scheme(struct ice_pf *pf);
 
+struct msi_map ice_alloc_irq(struct ice_pf *pf);
+void ice_free_irq(struct ice_pf *pf, struct msi_map map);
+
 #endif
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index f8c9c2c20e6d..fe908cf6da6a 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -1374,161 +1374,45 @@ static int ice_vsi_init(struct ice_vsi *vsi, u32 vsi_flags)
  * ice_free_res - free a block of resources
  * @res: pointer to the resource
  * @index: starting index previously returned by ice_get_res
- * @id: identifier to track owner
  *
  * Returns number of resources freed
  */
-int ice_free_res(struct ice_res_tracker *res, u16 index, u16 id)
+int ice_free_res(struct ice_res_tracker *res, u16 index)
 {
-	int count = 0;
-	int i;
-
 	if (!res || index >= res->end)
 		return -EINVAL;
 
-	id |= ICE_RES_VALID_BIT;
-	for (i = index; i < res->end && res->list[i] == id; i++) {
-		res->list[i] = 0;
-		count++;
-	}
-
-	return count;
-}
-
-/**
- * ice_search_res - Search the tracker for a block of resources
- * @res: pointer to the resource
- * @needed: size of the block needed
- * @id: identifier to track owner
- *
- * Returns the base item index of the block, or -ENOMEM for error
- */
-static int ice_search_res(struct ice_res_tracker *res, u16 needed, u16 id)
-{
-	u16 start = 0, end = 0;
-
-	if (needed > res->end)
-		return -ENOMEM;
-
-	id |= ICE_RES_VALID_BIT;
-
-	do {
-		/* skip already allocated entries */
-		if (res->list[end++] & ICE_RES_VALID_BIT) {
-			start = end;
-			if ((start + needed) > res->end)
-				break;
-		}
-
-		if (end == (start + needed)) {
-			int i = start;
-
-			/* there was enough, so assign it to the requestor */
-			while (i != end)
-				res->list[i++] = id;
-
-			return start;
-		}
-	} while (end < res->end);
+	res->list[index] = 0;
 
-	return -ENOMEM;
-}
-
-/**
- * ice_get_free_res_count - Get free count from a resource tracker
- * @res: Resource tracker instance
- */
-static u16 ice_get_free_res_count(struct ice_res_tracker *res)
-{
-	u16 i, count = 0;
-
-	for (i = 0; i < res->end; i++)
-		if (!(res->list[i] & ICE_RES_VALID_BIT))
-			count++;
-
-	return count;
+	return 0;
 }
 
 /**
- * ice_get_res - get a block of resources
+ * ice_get_res - get a resource from the tracker
  * @pf: board private structure
  * @res: pointer to the resource
- * @needed: size of the block needed
- * @id: identifier to track owner
  *
- * Returns the base item index of the block, or negative for error
+ * Returns the item index, or negative for error
  */
 int
-ice_get_res(struct ice_pf *pf, struct ice_res_tracker *res, u16 needed, u16 id)
+ice_get_res(struct ice_pf *pf, struct ice_res_tracker *res)
 {
-	if (!res || !pf)
-		return -EINVAL;
+	u16 i;
 
-	if (!needed || needed > res->num_entries || id >= ICE_RES_VALID_BIT) {
-		dev_err(ice_pf_to_dev(pf), "param err: needed=%d, num_entries = %d id=0x%04x\n",
-			needed, res->num_entries, id);
+	if (!res || !pf)
 		return -EINVAL;
-	}
 
-	return ice_search_res(res, needed, id);
-}
-
-/**
- * ice_vsi_setup_vector_base - Set up the base vector for the given VSI
- * @vsi: ptr to the VSI
- *
- * This should only be called after ice_vsi_alloc_def() which allocates the
- * corresponding SW VSI structure and initializes num_queue_pairs for the
- * newly allocated VSI.
- *
- * Returns 0 on success or negative on failure
- */
-static int ice_vsi_setup_vector_base(struct ice_vsi *vsi)
-{
-	struct ice_pf *pf = vsi->back;
-	struct device *dev;
-	u16 num_q_vectors;
-	int base;
-
-	dev = ice_pf_to_dev(pf);
-	/* SRIOV doesn't grab irq_tracker entries for each VSI */
-	if (vsi->type == ICE_VSI_VF)
-		return 0;
-	if (vsi->type == ICE_VSI_CHNL)
-		return 0;
-
-	if (vsi->base_vector) {
-		dev_dbg(dev, "VSI %d has non-zero base vector %d\n",
-			vsi->vsi_num, vsi->base_vector);
-		return -EEXIST;
-	}
-
-	num_q_vectors = vsi->num_q_vectors;
-	/* reserve slots from OS requested IRQs */
-	if (vsi->type == ICE_VSI_CTRL && vsi->vf) {
-		struct ice_vsi *ctrl_vsi = ice_get_vf_ctrl_vsi(pf, vsi);
+	/* skip already allocated entries */
+	for (i = 0; i < res->end; i++)
+		if (!(res->list[i] & ICE_RES_VALID_BIT))
+			break;
 
-		if (ctrl_vsi)
-			base = ctrl_vsi->base_vector;
-		else
-			base = ice_get_res(pf, pf->irq_tracker,
-					   vsi->num_q_vectors,
-					   ICE_RES_VF_CTRL_VEC_ID);
+	if (i < res->end) {
+		res->list[i] = ICE_RES_VALID_BIT;
+		return i;
 	} else {
-		base = ice_get_res(pf, pf->irq_tracker, num_q_vectors,
-				   vsi->idx);
-	}
-
-	if (base < 0) {
-		dev_err(dev, "%d MSI-X interrupts available. %s %d failed to get %d MSI-X vectors\n",
-			ice_get_free_res_count(pf->irq_tracker),
-			ice_vsi_type_str(vsi->type), vsi->idx, num_q_vectors);
-		return -ENOENT;
+		return -ENOMEM;
 	}
-	vsi->base_vector = (u16)base;
-	pf->num_avail_sw_msix -= num_q_vectors;
-
-	return 0;
 }
 
 /**
@@ -2386,50 +2270,6 @@ static void ice_vsi_set_tc_cfg(struct ice_vsi *vsi)
 	ice_vsi_set_dcb_tc_cfg(vsi);
 }
 
-/**
- * ice_vsi_set_q_vectors_reg_idx - set the HW register index for all q_vectors
- * @vsi: VSI to set the q_vectors register index on
- */
-static int
-ice_vsi_set_q_vectors_reg_idx(struct ice_vsi *vsi)
-{
-	u16 i;
-
-	if (!vsi || !vsi->q_vectors)
-		return -EINVAL;
-
-	ice_for_each_q_vector(vsi, i) {
-		struct ice_q_vector *q_vector = vsi->q_vectors[i];
-
-		if (!q_vector) {
-			dev_err(ice_pf_to_dev(vsi->back), "Failed to set reg_idx on q_vector %d VSI %d\n",
-				i, vsi->vsi_num);
-			goto clear_reg_idx;
-		}
-
-		if (vsi->type == ICE_VSI_VF) {
-			struct ice_vf *vf = vsi->vf;
-
-			q_vector->reg_idx = ice_calc_vf_reg_idx(vf, q_vector);
-		} else {
-			q_vector->reg_idx =
-				q_vector->v_idx + vsi->base_vector;
-		}
-	}
-
-	return 0;
-
-clear_reg_idx:
-	ice_for_each_q_vector(vsi, i) {
-		struct ice_q_vector *q_vector = vsi->q_vectors[i];
-
-		if (q_vector)
-			q_vector->reg_idx = 0;
-	}
-
-	return -EINVAL;
-}
-
 /**
  * ice_cfg_sw_lldp - Config switch rules for LLDP packet handling
  * @vsi: the VSI being configured
@@ -2674,14 +2514,6 @@ ice_vsi_cfg_def(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params)
 		if (ret)
 			goto unroll_vsi_init;
 
-		ret = ice_vsi_setup_vector_base(vsi);
-		if (ret)
-			goto unroll_alloc_q_vector;
-
-		ret = ice_vsi_set_q_vectors_reg_idx(vsi);
-		if (ret)
-			goto unroll_vector_base;
-
 		ret = ice_vsi_alloc_rings(vsi);
 		if (ret)
 			goto unroll_vector_base;
@@ -2732,10 +2564,6 @@ ice_vsi_cfg_def(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params)
 		if (ret)
 			goto unroll_alloc_q_vector;
 
-		ret = ice_vsi_set_q_vectors_reg_idx(vsi);
-		if (ret)
-			goto unroll_vector_base;
-
 		ret = ice_vsi_alloc_ring_stats(vsi);
 		if (ret)
 			goto unroll_vector_base;
@@ -2768,8 +2596,6 @@ ice_vsi_cfg_def(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params)
 
 unroll_vector_base:
 	/* reclaim SW interrupts back to the common pool */
-	ice_free_res(pf->irq_tracker, vsi->base_vector, vsi->idx);
-	pf->num_avail_sw_msix += vsi->num_q_vectors;
 unroll_alloc_q_vector:
 	ice_vsi_free_q_vectors(vsi);
 unroll_vsi_init:
@@ -2861,18 +2687,6 @@ void ice_vsi_decfg(struct ice_vsi *vsi)
 	 * many interrupts each VF needs. SR-IOV MSIX resources are also
 	 * cleared in the same manner.
 	 */
-	if (vsi->type == ICE_VSI_CTRL && vsi->vf) {
-		struct ice_vsi *ctrl_vsi = ice_get_vf_ctrl_vsi(pf, vsi);
-
-		if (!ctrl_vsi)
-			ice_free_res(pf->irq_tracker, vsi->base_vector,
-				     ICE_RES_VF_CTRL_VEC_ID);
-	} else if (vsi->type != ICE_VSI_VF) {
-		/* reclaim SW interrupts back to the common pool */
-		ice_free_res(pf->irq_tracker, vsi->base_vector, vsi->idx);
-		pf->num_avail_sw_msix += vsi->num_q_vectors;
-		vsi->base_vector = 0;
-	}
 
 	if (vsi->type == ICE_VSI_VF &&
 	    vsi->agg_node && vsi->agg_node->valid)
@@ -2989,7 +2803,6 @@ static void ice_vsi_release_msix(struct ice_vsi *vsi)
 void ice_vsi_free_irq(struct ice_vsi *vsi)
 {
 	struct ice_pf *pf = vsi->back;
-	int base = vsi->base_vector;
 	int i;
 
 	if (!vsi->q_vectors || !vsi->irqs_ready)
@@ -3003,10 +2816,9 @@ void ice_vsi_free_irq(struct ice_vsi *vsi)
 	ice_free_cpu_rx_rmap(vsi);
 
 	ice_for_each_q_vector(vsi, i) {
-		u16 vector = i + base;
 		int irq_num;
 
-		irq_num = pci_irq_vector(pf->pdev, vector);
+		irq_num = vsi->q_vectors[i]->irq.virq;
 
 		/* free only the irqs that were actually requested */
 		if (!vsi->q_vectors[i] ||
@@ -3138,7 +2950,6 @@ void ice_dis_vsi(struct ice_vsi *vsi, bool locked)
  */
 void ice_vsi_dis_irq(struct ice_vsi *vsi)
 {
-	int base = vsi->base_vector;
 	struct ice_pf *pf = vsi->back;
 	struct ice_hw *hw = &pf->hw;
 	u32 val;
@@ -3185,7 +2996,7 @@ void ice_vsi_dis_irq(struct ice_vsi *vsi)
 		return;
 
 	ice_for_each_q_vector(vsi, i)
-		synchronize_irq(pci_irq_vector(pf->pdev, i + base));
+		synchronize_irq(vsi->q_vectors[i]->irq.virq);
 }
 
 /**
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.h b/drivers/net/ethernet/intel/ice/ice_lib.h
index 75221478f2dc..2f52f9e32858 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.h
+++ b/drivers/net/ethernet/intel/ice/ice_lib.h
@@ -104,10 +104,10 @@ int ice_ena_vsi(struct ice_vsi *vsi, bool locked);
 void ice_vsi_decfg(struct ice_vsi *vsi);
 void ice_dis_vsi(struct ice_vsi *vsi, bool locked);
 
-int ice_free_res(struct ice_res_tracker *res, u16 index, u16 id);
+int ice_free_res(struct ice_res_tracker *res, u16 index);
 
 int
-ice_get_res(struct ice_pf *pf, struct ice_res_tracker *res, u16 needed, u16 id);
+ice_get_res(struct ice_pf *pf, struct ice_res_tracker *res);
 
 int ice_vsi_rebuild(struct ice_vsi *vsi, u32 vsi_flags);
 int ice_vsi_cfg(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params);
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index ce8cd49ae10c..efc621c0bd6c 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -2490,7 +2490,6 @@ static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
 {
 	int q_vectors = vsi->num_q_vectors;
 	struct ice_pf *pf = vsi->back;
-	int base = vsi->base_vector;
 	struct device *dev;
 	int rx_int_idx = 0;
 	int tx_int_idx = 0;
@@ -2501,7 +2500,7 @@ static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
 	for (vector = 0; vector < q_vectors; vector++) {
 		struct ice_q_vector *q_vector = vsi->q_vectors[vector];
 
-		irq_num = pci_irq_vector(pf->pdev, base + vector);
+		irq_num = q_vector->irq.virq;
 
 		if (q_vector->tx.tx_ring && q_vector->rx.rx_ring) {
 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
@@ -2555,9 +2554,8 @@ static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
 	return 0;
 
 free_q_irqs:
-	while (vector) {
-		vector--;
-		irq_num = pci_irq_vector(pf->pdev, base + vector);
+	while (vector--) {
+		irq_num = vsi->q_vectors[vector]->irq.virq;
 		if (!IS_ENABLED(CONFIG_RFS_ACCEL))
 			irq_set_affinity_notifier(irq_num, NULL);
 		irq_set_affinity_hint(irq_num, NULL);
@@ -3047,7 +3045,7 @@ static void ice_ena_misc_vector(struct ice_pf *pf)
 	wr32(hw, PFINT_OICR_ENA, val);
 
 	/* SW_ITR_IDX = 0, but don't change INTENA */
-	wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
+	wr32(hw, GLINT_DYN_CTL(pf->oicr_irq.index),
 	     GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
 }
 
@@ -3234,7 +3232,7 @@ static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
  */
 static void ice_free_irq_msix_misc(struct ice_pf *pf)
 {
-	int misc_irq_num = pci_irq_vector(pf->pdev, pf->oicr_idx);
+	int misc_irq_num = pf->oicr_irq.virq;
 	struct ice_hw *hw = &pf->hw;
 
 	ice_dis_ctrlq_interrupts(hw);
@@ -3246,8 +3244,7 @@ static void ice_free_irq_msix_misc(struct ice_pf *pf)
 	synchronize_irq(misc_irq_num);
 	devm_free_irq(ice_pf_to_dev(pf), misc_irq_num, pf);
 
-	pf->num_avail_sw_msix += 1;
-	ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
+	ice_free_irq(pf, pf->oicr_irq);
 }
 
 /**
@@ -3293,7 +3290,8 @@ static int ice_req_irq_msix_misc(struct ice_pf *pf)
 {
 	struct device *dev = ice_pf_to_dev(pf);
 	struct ice_hw *hw = &pf->hw;
-	int oicr_idx, err = 0;
+	struct msi_map oicr_irq;
+	int err = 0;
 
 	if (!pf->int_name[0])
 		snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
@@ -3307,30 +3305,26 @@ static int ice_req_irq_msix_misc(struct ice_pf *pf)
 		goto skip_req_irq;
 
 	/* reserve one vector in irq_tracker for misc interrupts */
-	oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
-	if (oicr_idx < 0)
-		return oicr_idx;
-
-	pf->num_avail_sw_msix -= 1;
-	pf->oicr_idx = (u16)oicr_idx;
-
-	err = devm_request_threaded_irq(dev,
-					pci_irq_vector(pf->pdev, pf->oicr_idx),
-					ice_misc_intr, ice_misc_intr_thread_fn,
-					0, pf->int_name, pf);
+	oicr_irq = ice_alloc_irq(pf);
+	if (oicr_irq.index < 0)
+		return oicr_irq.index;
+
+	pf->oicr_irq = oicr_irq;
+	err = devm_request_threaded_irq(dev, pf->oicr_irq.virq, ice_misc_intr,
+					ice_misc_intr_thread_fn, 0,
+					pf->int_name, pf);
 	if (err) {
 		dev_err(dev, "devm_request_threaded_irq for %s failed: %d\n",
 			pf->int_name, err);
-		ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
-		pf->num_avail_sw_msix += 1;
+		ice_free_irq(pf, pf->oicr_irq);
 		return err;
 	}
 
 skip_req_irq:
 	ice_ena_misc_vector(pf);
 
-	ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
-	wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
+	ice_ena_ctrlq_interrupts(hw, pf->oicr_irq.index);
+	wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_irq.index),
 	     ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
 
 	ice_flush(hw);
diff --git a/drivers/net/ethernet/intel/ice/ice_ptp.c b/drivers/net/ethernet/intel/ice/ice_ptp.c
index 972d4f6fd615..d4b6c997141d 100644
--- a/drivers/net/ethernet/intel/ice/ice_ptp.c
+++ b/drivers/net/ethernet/intel/ice/ice_ptp.c
@@ -911,7 +911,7 @@ ice_ptp_release_tx_tracker(struct ice_pf *pf, struct ice_ptp_tx *tx)
 	spin_unlock(&tx->lock);
 
 	/* wait for potentially outstanding interrupt to complete */
-	synchronize_irq(pci_irq_vector(pf->pdev, pf->oicr_idx));
+	synchronize_irq(pf->oicr_irq.virq);
 
 	ice_ptp_flush_tx_tracker(pf, tx);
 
diff --git a/drivers/net/ethernet/intel/ice/ice_sriov.c b/drivers/net/ethernet/intel/ice/ice_sriov.c
index 65f971b74717..0fc2b26a2fa6 100644
--- a/drivers/net/ethernet/intel/ice/ice_sriov.c
+++ b/drivers/net/ethernet/intel/ice/ice_sriov.c
@@ -835,7 +835,7 @@ static int ice_ena_vfs(struct ice_pf *pf, u16 num_vfs)
 	int ret;
 
 	/* Disable global interrupt 0 so we don't try to handle the VFLR. */
-	wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
+	wr32(hw, GLINT_DYN_CTL(pf->oicr_irq.index),
 	     ICE_ITR_NONE << GLINT_DYN_CTL_ITR_INDX_S);
 	set_bit(ICE_OICR_INTR_DIS, pf->state);
 	ice_flush(hw);
diff --git a/drivers/net/ethernet/intel/ice/ice_xsk.c b/drivers/net/ethernet/intel/ice/ice_xsk.c
index 4102416d7a41..a7fe2b4ce655 100644
--- a/drivers/net/ethernet/intel/ice/ice_xsk.c
+++ b/drivers/net/ethernet/intel/ice/ice_xsk.c
@@ -90,7 +90,6 @@ ice_qvec_dis_irq(struct ice_vsi *vsi, struct ice_rx_ring *rx_ring,
 {
 	struct ice_pf *pf = vsi->back;
 	struct ice_hw *hw = &pf->hw;
-	int base = vsi->base_vector;
 	u16 reg;
 	u32 val;
 
@@ -103,11 +102,9 @@ ice_qvec_dis_irq(struct ice_vsi *vsi, struct ice_rx_ring *rx_ring,
 	wr32(hw, QINT_RQCTL(reg), val);
 
 	if (q_vector) {
-		u16 v_idx = q_vector->v_idx;
-
 		wr32(hw, GLINT_DYN_CTL(q_vector->reg_idx), 0);
 		ice_flush(hw);
-		synchronize_irq(pci_irq_vector(pf->pdev, v_idx + base));
+		synchronize_irq(q_vector->irq.virq);
 	}
 }
 
-- 
2.38.1

[PATCH net-next 7/8] ice: track interrupt vectors with xarray

[Tony Nguyen]

From: Piotr Raczynski <piotr.raczynski@intel.com>

Replace custom interrupt tracker with generic xarray data structure.
Remove all code responsible for searching for a new entry with xa_alloc,
which always tries to allocate at the lowes possible index. As a result
driver is always using a contiguous region of the MSIX vector table.

New tracker keeps ice_irq_entry entries in xarray as opaque for the rest
of the driver hiding the entry details from the caller.

Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Piotr Raczynski <piotr.raczynski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/ice/ice.h       |   9 +-
 drivers/net/ethernet/intel/ice/ice_irq.c   | 101 ++++++++++++++++-----
 drivers/net/ethernet/intel/ice/ice_irq.h   |   9 ++
 drivers/net/ethernet/intel/ice/ice_lib.c   |  45 ---------
 drivers/net/ethernet/intel/ice/ice_lib.h   |   5 -
 drivers/net/ethernet/intel/ice/ice_sriov.c |   4 +-
 6 files changed, 89 insertions(+), 84 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
index d8dde291491e..8541d986ec7f 100644
--- a/drivers/net/ethernet/intel/ice/ice.h
+++ b/drivers/net/ethernet/intel/ice/ice.h
@@ -104,7 +104,6 @@
 #define ICE_Q_WAIT_RETRY_LIMIT	10
 #define ICE_Q_WAIT_MAX_RETRY	(5 * ICE_Q_WAIT_RETRY_LIMIT)
 #define ICE_MAX_LG_RSS_QS	256
-#define ICE_RES_VALID_BIT	0x8000
 #define ICE_INVAL_Q_INDEX	0xffff
 
 #define ICE_MAX_RXQS_PER_TC		256	/* Used when setting VSI context per TC Rx queues */
@@ -242,12 +241,6 @@ struct ice_tc_cfg {
 	struct ice_tc_info tc_info[ICE_MAX_TRAFFIC_CLASS];
 };
 
-struct ice_res_tracker {
-	u16 num_entries;
-	u16 end;
-	u16 list[];
-};
-
 struct ice_qs_cfg {
 	struct mutex *qs_mutex;  /* will be assigned to &pf->avail_q_mutex */
 	unsigned long *pf_map;
@@ -536,7 +529,7 @@ struct ice_pf {
 
 	/* OS reserved IRQ details */
 	struct msix_entry *msix_entries;
-	struct ice_res_tracker *irq_tracker;
+	struct ice_irq_tracker irq_tracker;
 	/* First MSIX vector used by SR-IOV VFs. Calculated by subtracting the
 	 * number of MSIX vectors needed for all SR-IOV VFs from the number of
 	 * MSIX vectors allowed on this PF.
diff --git a/drivers/net/ethernet/intel/ice/ice_irq.c b/drivers/net/ethernet/intel/ice/ice_irq.c
index ca1a1de26766..1713347c577f 100644
--- a/drivers/net/ethernet/intel/ice/ice_irq.c
+++ b/drivers/net/ethernet/intel/ice/ice_irq.c
@@ -5,6 +5,75 @@
 #include "ice_lib.h"
 #include "ice_irq.h"
 
+/**
+ * ice_init_irq_tracker - initialize interrupt tracker
+ * @pf: board private structure
+ * @max_vectors: maximum number of vectors that tracker can hold
+ */
+static void
+ice_init_irq_tracker(struct ice_pf *pf, unsigned int max_vectors)
+{
+	pf->irq_tracker.num_entries = max_vectors;
+	xa_init_flags(&pf->irq_tracker.entries, XA_FLAGS_ALLOC);
+}
+
+/**
+ * ice_deinit_irq_tracker - free xarray tracker
+ * @pf: board private structure
+ */
+static void ice_deinit_irq_tracker(struct ice_pf *pf)
+{
+	xa_destroy(&pf->irq_tracker.entries);
+}
+
+/**
+ * ice_free_irq_res - free a block of resources
+ * @pf: board private structure
+ * @index: starting index previously returned by ice_get_res
+ */
+static void ice_free_irq_res(struct ice_pf *pf, u16 index)
+{
+	struct ice_irq_entry *entry;
+
+	entry = xa_erase(&pf->irq_tracker.entries, index);
+	kfree(entry);
+}
+
+/**
+ * ice_get_irq_res - get an interrupt resource
+ * @pf: board private structure
+ *
+ * Allocate new irq entry in the free slot of the tracker. Since xarray
+ * is used, always allocate new entry at the lowest possible index. Set
+ * proper allocation limit for maximum tracker entries.
+ *
+ * Returns allocated irq entry or NULL on failure.
+ */
+static struct ice_irq_entry *ice_get_irq_res(struct ice_pf *pf)
+{
+	struct xa_limit limit = { .max = pf->irq_tracker.num_entries,
+				  .min = 0 };
+	struct ice_irq_entry *entry;
+	unsigned int index;
+	int ret;
+
+	entry = kzalloc(sizeof(*entry), GFP_KERNEL);
+	if (!entry)
+		return NULL;
+
+	ret = xa_alloc(&pf->irq_tracker.entries, &index, entry, limit,
+		       GFP_KERNEL);
+
+	if (ret) {
+		kfree(entry);
+		entry = NULL;
+	} else {
+		entry->index = index;
+	}
+
+	return entry;
+}
+
 /**
  * ice_reduce_msix_usage - Reduce usage of MSI-X vectors
  * @pf: board private structure
@@ -163,11 +232,7 @@ static int ice_ena_msix_range(struct ice_pf *pf)
 void ice_clear_interrupt_scheme(struct ice_pf *pf)
 {
 	pci_free_irq_vectors(pf->pdev);
-
-	if (pf->irq_tracker) {
-		devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
-		pf->irq_tracker = NULL;
-	}
+	ice_deinit_irq_tracker(pf);
 }
 
 /**
@@ -183,19 +248,7 @@ int ice_init_interrupt_scheme(struct ice_pf *pf)
 	if (vectors < 0)
 		return vectors;
 
-	/* set up vector assignment tracking */
-	pf->irq_tracker = devm_kzalloc(ice_pf_to_dev(pf),
-				       struct_size(pf->irq_tracker, list,
-						   vectors),
-				       GFP_KERNEL);
-	if (!pf->irq_tracker) {
-		pci_free_irq_vectors(pf->pdev);
-		return -ENOMEM;
-	}
-
-	/* populate SW interrupts pool with number of OS granted IRQs. */
-	pf->irq_tracker->num_entries = (u16)vectors;
-	pf->irq_tracker->end = pf->irq_tracker->num_entries;
+	ice_init_irq_tracker(pf, vectors);
 
 	return 0;
 }
@@ -221,13 +274,13 @@ int ice_init_interrupt_scheme(struct ice_pf *pf)
 struct msi_map ice_alloc_irq(struct ice_pf *pf)
 {
 	struct msi_map map = { .index = -ENOENT };
-	int entry;
+	struct ice_irq_entry *entry;
 
-	entry = ice_get_res(pf, pf->irq_tracker);
-	if (entry < 0)
+	entry = ice_get_irq_res(pf);
+	if (!entry)
 		return map;
 
-	map.index = entry;
+	map.index = entry->index;
 	map.virq = pci_irq_vector(pf->pdev, map.index);
 
 	return map;
@@ -238,9 +291,9 @@ struct msi_map ice_alloc_irq(struct ice_pf *pf)
  * @pf: board private structure
  * @map: map with interrupt details
  *
- * Remove allocated interrupt from the interrupt tracker
+ * Remove allocated interrupt from the interrupt tracker.
  */
 void ice_free_irq(struct ice_pf *pf, struct msi_map map)
 {
-	ice_free_res(pf->irq_tracker, map.index);
+	ice_free_irq_res(pf, map.index);
 }
diff --git a/drivers/net/ethernet/intel/ice/ice_irq.h b/drivers/net/ethernet/intel/ice/ice_irq.h
index 26e80dfe22b5..da5cdb1f0d3a 100644
--- a/drivers/net/ethernet/intel/ice/ice_irq.h
+++ b/drivers/net/ethernet/intel/ice/ice_irq.h
@@ -4,6 +4,15 @@
 #ifndef _ICE_IRQ_H_
 #define _ICE_IRQ_H_
 
+struct ice_irq_entry {
+	unsigned int index;
+};
+
+struct ice_irq_tracker {
+	struct xarray entries;
+	u16 num_entries;	/* total vectors available */
+};
+
 int ice_init_interrupt_scheme(struct ice_pf *pf);
 void ice_clear_interrupt_scheme(struct ice_pf *pf);
 
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index fe908cf6da6a..387bb9cbafbe 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -1370,51 +1370,6 @@ static int ice_vsi_init(struct ice_vsi *vsi, u32 vsi_flags)
 	return ret;
 }
 
-/**
- * ice_free_res - free a block of resources
- * @res: pointer to the resource
- * @index: starting index previously returned by ice_get_res
- *
- * Returns number of resources freed
- */
-int ice_free_res(struct ice_res_tracker *res, u16 index)
-{
-	if (!res || index >= res->end)
-		return -EINVAL;
-
-	res->list[index] = 0;
-
-	return 0;
-}
-
-/**
- * ice_get_res - get a resource from the tracker
- * @pf: board private structure
- * @res: pointer to the resource
- *
- * Returns the item index, or negative for error
- */
-int
-ice_get_res(struct ice_pf *pf, struct ice_res_tracker *res)
-{
-	u16 i;
-
-	if (!res || !pf)
-		return -EINVAL;
-
-	/* skip already allocated entries */
-	for (i = 0; i < res->end; i++)
-		if (!(res->list[i] & ICE_RES_VALID_BIT))
-			break;
-
-	if (i < res->end) {
-		res->list[i] = ICE_RES_VALID_BIT;
-		return i;
-	} else {
-		return -ENOMEM;
-	}
-}
-
 /**
  * ice_vsi_clear_rings - Deallocates the Tx and Rx rings for VSI
  * @vsi: the VSI having rings deallocated
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.h b/drivers/net/ethernet/intel/ice/ice_lib.h
index 2f52f9e32858..e985766e6bb5 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.h
+++ b/drivers/net/ethernet/intel/ice/ice_lib.h
@@ -104,11 +104,6 @@ int ice_ena_vsi(struct ice_vsi *vsi, bool locked);
 void ice_vsi_decfg(struct ice_vsi *vsi);
 void ice_dis_vsi(struct ice_vsi *vsi, bool locked);
 
-int ice_free_res(struct ice_res_tracker *res, u16 index);
-
-int
-ice_get_res(struct ice_pf *pf, struct ice_res_tracker *res);
-
 int ice_vsi_rebuild(struct ice_vsi *vsi, u32 vsi_flags);
 int ice_vsi_cfg(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params);
 
diff --git a/drivers/net/ethernet/intel/ice/ice_sriov.c b/drivers/net/ethernet/intel/ice/ice_sriov.c
index 0fc2b26a2fa6..195105ce9039 100644
--- a/drivers/net/ethernet/intel/ice/ice_sriov.c
+++ b/drivers/net/ethernet/intel/ice/ice_sriov.c
@@ -418,7 +418,7 @@ int ice_calc_vf_reg_idx(struct ice_vf *vf, struct ice_q_vector *q_vector)
 static int ice_sriov_set_msix_res(struct ice_pf *pf, u16 num_msix_needed)
 {
 	u16 total_vectors = pf->hw.func_caps.common_cap.num_msix_vectors;
-	int vectors_used = pf->irq_tracker->num_entries;
+	int vectors_used = pf->irq_tracker.num_entries;
 	int sriov_base_vector;
 
 	sriov_base_vector = total_vectors - num_msix_needed;
@@ -470,7 +470,7 @@ static int ice_set_per_vf_res(struct ice_pf *pf, u16 num_vfs)
 
 	/* determine MSI-X resources per VF */
 	msix_avail_for_sriov = pf->hw.func_caps.common_cap.num_msix_vectors -
-		pf->irq_tracker->num_entries;
+		pf->irq_tracker.num_entries;
 	msix_avail_per_vf = msix_avail_for_sriov / num_vfs;
 	if (msix_avail_per_vf >= ICE_NUM_VF_MSIX_MED) {
 		num_msix_per_vf = ICE_NUM_VF_MSIX_MED;
-- 
2.38.1

[PATCH net-next 8/8] ice: add dynamic interrupt allocation

[Tony Nguyen]

From: Piotr Raczynski <piotr.raczynski@intel.com>

Currently driver can only allocate interrupt vectors during init phase by
calling pci_alloc_irq_vectors. Change that and make use of new
pci_msix_alloc_irq_at/pci_msix_free_irq API and enable to allocate and free
more interrupts after MSIX has been enabled. Since not all platforms
supports dynamic allocation, check it with pci_msix_can_alloc_dyn.

Extend the tracker to keep track how many interrupts are allocated
initially so when all such vectors are already used, additional interrupts
are automatically allocated dynamically. Remember each interrupt allocation
method to then free appropriately. Since some features may require
interrupts allocated dynamically add appropriate VSI flag and take it into
account when allocating new interrupt.

Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Piotr Raczynski <piotr.raczynski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/ice/ice.h       |   3 +
 drivers/net/ethernet/intel/ice/ice_base.c  |   2 +-
 drivers/net/ethernet/intel/ice/ice_idc.c   |   2 +-
 drivers/net/ethernet/intel/ice/ice_irq.c   | 109 ++++++++++++++++++---
 drivers/net/ethernet/intel/ice/ice_irq.h   |   5 +-
 drivers/net/ethernet/intel/ice/ice_main.c  |   2 +-
 drivers/net/ethernet/intel/ice/ice_sriov.c |   5 +-
 7 files changed, 107 insertions(+), 21 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
index 8541d986ec7f..d637032c8139 100644
--- a/drivers/net/ethernet/intel/ice/ice.h
+++ b/drivers/net/ethernet/intel/ice/ice.h
@@ -338,6 +338,9 @@ struct ice_vsi {
 	u32 rx_buf_failed;
 	u32 rx_page_failed;
 	u16 num_q_vectors;
+	/* tell if only dynamic irq allocation is allowed */
+	bool irq_dyn_alloc;
+
 	enum ice_vsi_type type;
 	u16 vsi_num;			/* HW (absolute) index of this VSI */
 	u16 idx;			/* software index in pf->vsi[] */
diff --git a/drivers/net/ethernet/intel/ice/ice_base.c b/drivers/net/ethernet/intel/ice/ice_base.c
index 7dd7a0f32471..998dc7d81e51 100644
--- a/drivers/net/ethernet/intel/ice/ice_base.c
+++ b/drivers/net/ethernet/intel/ice/ice_base.c
@@ -133,7 +133,7 @@ static int ice_vsi_alloc_q_vector(struct ice_vsi *vsi, u16 v_idx)
 		}
 	}
 
-	q_vector->irq = ice_alloc_irq(pf);
+	q_vector->irq = ice_alloc_irq(pf, vsi->irq_dyn_alloc);
 	if (q_vector->irq.index < 0) {
 		kfree(q_vector);
 		return -ENOMEM;
diff --git a/drivers/net/ethernet/intel/ice/ice_idc.c b/drivers/net/ethernet/intel/ice/ice_idc.c
index bc016bb4440c..145b27f2a4ce 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc.c
+++ b/drivers/net/ethernet/intel/ice/ice_idc.c
@@ -250,7 +250,7 @@ static int ice_alloc_rdma_qvectors(struct ice_pf *pf)
 			struct msix_entry *entry = &pf->msix_entries[i];
 			struct msi_map map;
 
-			map = ice_alloc_irq(pf);
+			map = ice_alloc_irq(pf, false);
 			if (map.index < 0)
 				break;
 
diff --git a/drivers/net/ethernet/intel/ice/ice_irq.c b/drivers/net/ethernet/intel/ice/ice_irq.c
index 1713347c577f..ad82ff7d1995 100644
--- a/drivers/net/ethernet/intel/ice/ice_irq.c
+++ b/drivers/net/ethernet/intel/ice/ice_irq.c
@@ -9,11 +9,14 @@
  * ice_init_irq_tracker - initialize interrupt tracker
  * @pf: board private structure
  * @max_vectors: maximum number of vectors that tracker can hold
+ * @num_static: number of preallocated interrupts
  */
 static void
-ice_init_irq_tracker(struct ice_pf *pf, unsigned int max_vectors)
+ice_init_irq_tracker(struct ice_pf *pf, unsigned int max_vectors,
+		     unsigned int num_static)
 {
 	pf->irq_tracker.num_entries = max_vectors;
+	pf->irq_tracker.num_static = num_static;
 	xa_init_flags(&pf->irq_tracker.entries, XA_FLAGS_ALLOC);
 }
 
@@ -42,6 +45,7 @@ static void ice_free_irq_res(struct ice_pf *pf, u16 index)
 /**
  * ice_get_irq_res - get an interrupt resource
  * @pf: board private structure
+ * @dyn_only: force entry to be dynamically allocated
  *
  * Allocate new irq entry in the free slot of the tracker. Since xarray
  * is used, always allocate new entry at the lowest possible index. Set
@@ -49,10 +53,11 @@ static void ice_free_irq_res(struct ice_pf *pf, u16 index)
  *
  * Returns allocated irq entry or NULL on failure.
  */
-static struct ice_irq_entry *ice_get_irq_res(struct ice_pf *pf)
+static struct ice_irq_entry *ice_get_irq_res(struct ice_pf *pf, bool dyn_only)
 {
 	struct xa_limit limit = { .max = pf->irq_tracker.num_entries,
 				  .min = 0 };
+	unsigned int num_static = pf->irq_tracker.num_static;
 	struct ice_irq_entry *entry;
 	unsigned int index;
 	int ret;
@@ -61,6 +66,10 @@ static struct ice_irq_entry *ice_get_irq_res(struct ice_pf *pf)
 	if (!entry)
 		return NULL;
 
+	/* skip preallocated entries if the caller says so */
+	if (dyn_only)
+		limit.min = num_static;
+
 	ret = xa_alloc(&pf->irq_tracker.entries, &index, entry, limit,
 		       GFP_KERNEL);
 
@@ -69,6 +78,7 @@ static struct ice_irq_entry *ice_get_irq_res(struct ice_pf *pf)
 		entry = NULL;
 	} else {
 		entry->index = index;
+		entry->dynamic = index >= num_static;
 	}
 
 	return entry;
@@ -241,14 +251,20 @@ void ice_clear_interrupt_scheme(struct ice_pf *pf)
  */
 int ice_init_interrupt_scheme(struct ice_pf *pf)
 {
-	int vectors;
+	int total_vectors = pf->hw.func_caps.common_cap.num_msix_vectors;
+	int vectors, max_vectors;
 
 	vectors = ice_ena_msix_range(pf);
 
 	if (vectors < 0)
-		return vectors;
+		return -ENOMEM;
+
+	if (pci_msix_can_alloc_dyn(pf->pdev))
+		max_vectors = total_vectors;
+	else
+		max_vectors = vectors;
 
-	ice_init_irq_tracker(pf, vectors);
+	ice_init_irq_tracker(pf, max_vectors, vectors);
 
 	return 0;
 }
@@ -256,33 +272,55 @@ int ice_init_interrupt_scheme(struct ice_pf *pf)
 /**
  * ice_alloc_irq - Allocate new interrupt vector
  * @pf: board private structure
+ * @dyn_only: force dynamic allocation of the interrupt
  *
  * Allocate new interrupt vector for a given owner id.
  * return struct msi_map with interrupt details and track
  * allocated interrupt appropriately.
  *
- * This function mimics individual interrupt allocation,
- * even interrupts are actually already allocated with
- * pci_alloc_irq_vectors. Individual allocation helps
- * to track interrupts and simplifies interrupt related
- * handling.
+ * This function reserves new irq entry from the irq_tracker.
+ * if according to the tracker information all interrupts that
+ * were allocated with ice_pci_alloc_irq_vectors are already used
+ * and dynamically allocated interrupts are supported then new
+ * interrupt will be allocated with pci_msix_alloc_irq_at.
+ *
+ * Some callers may only support dynamically allocated interrupts.
+ * This is indicated with dyn_only flag.
  *
  * On failure, return map with negative .index. The caller
  * is expected to check returned map index.
  *
  */
-struct msi_map ice_alloc_irq(struct ice_pf *pf)
+struct msi_map ice_alloc_irq(struct ice_pf *pf, bool dyn_only)
 {
+	int sriov_base_vector = pf->sriov_base_vector;
 	struct msi_map map = { .index = -ENOENT };
+	struct device *dev = ice_pf_to_dev(pf);
 	struct ice_irq_entry *entry;
 
-	entry = ice_get_irq_res(pf);
+	entry = ice_get_irq_res(pf, dyn_only);
 	if (!entry)
 		return map;
 
-	map.index = entry->index;
-	map.virq = pci_irq_vector(pf->pdev, map.index);
+	/* fail if we're about to violate SRIOV vectors space */
+	if (sriov_base_vector && entry->index >= sriov_base_vector)
+		goto exit_free_res;
+
+	if (pci_msix_can_alloc_dyn(pf->pdev) && entry->dynamic) {
+		map = pci_msix_alloc_irq_at(pf->pdev, entry->index, NULL);
+		if (map.index < 0)
+			goto exit_free_res;
+		dev_dbg(dev, "allocated new irq at index %d\n", map.index);
+	} else {
+		map.index = entry->index;
+		map.virq = pci_irq_vector(pf->pdev, map.index);
+	}
+
+	return map;
 
+exit_free_res:
+	dev_err(dev, "Could not allocate irq at idx %d\n", entry->index);
+	ice_free_irq_res(pf, entry->index);
 	return map;
 }
 
@@ -291,9 +329,50 @@ struct msi_map ice_alloc_irq(struct ice_pf *pf)
  * @pf: board private structure
  * @map: map with interrupt details
  *
- * Remove allocated interrupt from the interrupt tracker.
+ * Remove allocated interrupt from the interrupt tracker. If interrupt was
+ * allocated dynamically, free respective interrupt vector.
  */
 void ice_free_irq(struct ice_pf *pf, struct msi_map map)
 {
+	struct ice_irq_entry *entry;
+
+	entry = xa_load(&pf->irq_tracker.entries, map.index);
+
+	if (!entry) {
+		dev_err(ice_pf_to_dev(pf), "Failed to get MSIX interrupt entry at index %d",
+			map.index);
+		return;
+	}
+
+	dev_dbg(ice_pf_to_dev(pf), "Free irq at index %d\n", map.index);
+
+	if (entry->dynamic)
+		pci_msix_free_irq(pf->pdev, map);
+
 	ice_free_irq_res(pf, map.index);
 }
+
+/**
+ * ice_get_max_used_msix_vector - Get the max used interrupt vector
+ * @pf: board private structure
+ *
+ * Return index of maximum used interrupt vectors with respect to the
+ * beginning of the MSIX table. Take into account that some interrupts
+ * may have been dynamically allocated after MSIX was initially enabled.
+ */
+int ice_get_max_used_msix_vector(struct ice_pf *pf)
+{
+	unsigned long start, index, max_idx;
+	void *entry;
+
+	/* Treat all preallocated interrupts as used */
+	start = pf->irq_tracker.num_static;
+	max_idx = start - 1;
+
+	xa_for_each_start(&pf->irq_tracker.entries, index, entry, start) {
+		if (index > max_idx)
+			max_idx = index;
+	}
+
+	return max_idx;
+}
diff --git a/drivers/net/ethernet/intel/ice/ice_irq.h b/drivers/net/ethernet/intel/ice/ice_irq.h
index da5cdb1f0d3a..f35efc08575e 100644
--- a/drivers/net/ethernet/intel/ice/ice_irq.h
+++ b/drivers/net/ethernet/intel/ice/ice_irq.h
@@ -6,17 +6,20 @@
 
 struct ice_irq_entry {
 	unsigned int index;
+	bool dynamic;	/* allocation type flag */
 };
 
 struct ice_irq_tracker {
 	struct xarray entries;
 	u16 num_entries;	/* total vectors available */
+	u16 num_static;	/* preallocated entries */
 };
 
 int ice_init_interrupt_scheme(struct ice_pf *pf);
 void ice_clear_interrupt_scheme(struct ice_pf *pf);
 
-struct msi_map ice_alloc_irq(struct ice_pf *pf);
+struct msi_map ice_alloc_irq(struct ice_pf *pf, bool dyn_only);
 void ice_free_irq(struct ice_pf *pf, struct msi_map map);
+int ice_get_max_used_msix_vector(struct ice_pf *pf);
 
 #endif
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index efc621c0bd6c..62e91512aeab 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -3305,7 +3305,7 @@ static int ice_req_irq_msix_misc(struct ice_pf *pf)
 		goto skip_req_irq;
 
 	/* reserve one vector in irq_tracker for misc interrupts */
-	oicr_irq = ice_alloc_irq(pf);
+	oicr_irq = ice_alloc_irq(pf, false);
 	if (oicr_irq.index < 0)
 		return oicr_irq.index;
 
diff --git a/drivers/net/ethernet/intel/ice/ice_sriov.c b/drivers/net/ethernet/intel/ice/ice_sriov.c
index 195105ce9039..80c643fb9f2f 100644
--- a/drivers/net/ethernet/intel/ice/ice_sriov.c
+++ b/drivers/net/ethernet/intel/ice/ice_sriov.c
@@ -418,7 +418,7 @@ int ice_calc_vf_reg_idx(struct ice_vf *vf, struct ice_q_vector *q_vector)
 static int ice_sriov_set_msix_res(struct ice_pf *pf, u16 num_msix_needed)
 {
 	u16 total_vectors = pf->hw.func_caps.common_cap.num_msix_vectors;
-	int vectors_used = pf->irq_tracker.num_entries;
+	int vectors_used = ice_get_max_used_msix_vector(pf);
 	int sriov_base_vector;
 
 	sriov_base_vector = total_vectors - num_msix_needed;
@@ -458,6 +458,7 @@ static int ice_sriov_set_msix_res(struct ice_pf *pf, u16 num_msix_needed)
  */
 static int ice_set_per_vf_res(struct ice_pf *pf, u16 num_vfs)
 {
+	int vectors_used = ice_get_max_used_msix_vector(pf);
 	u16 num_msix_per_vf, num_txq, num_rxq, avail_qs;
 	int msix_avail_per_vf, msix_avail_for_sriov;
 	struct device *dev = ice_pf_to_dev(pf);
@@ -470,7 +471,7 @@ static int ice_set_per_vf_res(struct ice_pf *pf, u16 num_vfs)
 
 	/* determine MSI-X resources per VF */
 	msix_avail_for_sriov = pf->hw.func_caps.common_cap.num_msix_vectors -
-		pf->irq_tracker.num_entries;
+		vectors_used;
 	msix_avail_per_vf = msix_avail_for_sriov / num_vfs;
 	if (msix_avail_per_vf >= ICE_NUM_VF_MSIX_MED) {
 		num_msix_per_vf = ICE_NUM_VF_MSIX_MED;
-- 
2.38.1

Re: [PATCH net-next 6/8] ice: add individual interrupt allocation

[Simon Horman]

On Tue, May 09, 2023 at 10:00:46AM -0700, Tony Nguyen wrote:
> From: Piotr Raczynski <piotr.raczynski@intel.com>
> 
> Currently interrupt allocations, depending on a feature are distributed
> in batches. Also, after allocation there is a series of operations that
> distributes per irq settings through that batch of interrupts.
> 
> Although driver does not yet support dynamic interrupt allocation, keep
> allocated interrupts in a pool and add allocation abstraction logic to
> make code more flexible. Keep per interrupt information in the
> ice_q_vector structure, which yields ice_vsi::base_vector redundant.
> Also, as a result there are a few functions that can be removed.
> 
> Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
> Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
> Reviewed-by: Simon Horman <simon.horman@corigine.com>
> Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
> Signed-off-by: Piotr Raczynski <piotr.raczynski@intel.com>
> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

...

> diff --git a/drivers/net/ethernet/intel/ice/ice_base.c b/drivers/net/ethernet/intel/ice/ice_base.c
> index 1911d644dfa8..7dd7a0f32471 100644
> --- a/drivers/net/ethernet/intel/ice/ice_base.c
> +++ b/drivers/net/ethernet/intel/ice/ice_base.c
> @@ -105,8 +105,7 @@ static int ice_vsi_alloc_q_vector(struct ice_vsi *vsi, u16 v_idx)
>  	struct ice_q_vector *q_vector;
>  
>  	/* allocate q_vector */
> -	q_vector = devm_kzalloc(ice_pf_to_dev(pf), sizeof(*q_vector),
> -				GFP_KERNEL);
> +	q_vector = kzalloc(sizeof(*q_vector), GFP_KERNEL);
>  	if (!q_vector)
>  		return -ENOMEM;
>  
> @@ -118,9 +117,31 @@ static int ice_vsi_alloc_q_vector(struct ice_vsi *vsi, u16 v_idx)
>  	q_vector->rx.itr_mode = ITR_DYNAMIC;
>  	q_vector->tx.type = ICE_TX_CONTAINER;
>  	q_vector->rx.type = ICE_RX_CONTAINER;
> +	q_vector->irq.index = -ENOENT;
>  
> -	if (vsi->type == ICE_VSI_VF)
> +	if (vsi->type == ICE_VSI_VF) {
> +		q_vector->reg_idx = ice_calc_vf_reg_idx(vsi->vf, q_vector);
>  		goto out;
> +	} else if (vsi->type == ICE_VSI_CTRL && vsi->vf) {
> +		struct ice_vsi *ctrl_vsi = ice_get_vf_ctrl_vsi(pf, vsi);
> +
> +		if (ctrl_vsi) {
> +			if (unlikely(!ctrl_vsi->q_vectors))
> +				return -ENOENT;

q_vector appears to be leaked here.

> +			q_vector->irq = ctrl_vsi->q_vectors[0]->irq;
> +			goto skip_alloc;
> +		}
> +	}
> +
> +	q_vector->irq = ice_alloc_irq(pf);
> +	if (q_vector->irq.index < 0) {
> +		kfree(q_vector);
> +		return -ENOMEM;
> +	}
> +
> +skip_alloc:
> +	q_vector->reg_idx = q_vector->irq.index;
> +
>  	/* only set affinity_mask if the CPU is online */
>  	if (cpu_online(v_idx))
>  		cpumask_set_cpu(v_idx, &q_vector->affinity_mask);

...

Re: [PATCH net-next 6/8] ice: add individual interrupt allocation

[Jacob Keller]

On 5/10/2023 1:54 AM, Simon Horman wrote:
> On Tue, May 09, 2023 at 10:00:46AM -0700, Tony Nguyen wrote:
>> From: Piotr Raczynski <piotr.raczynski@intel.com>
>>
>> Currently interrupt allocations, depending on a feature are distributed
>> in batches. Also, after allocation there is a series of operations that
>> distributes per irq settings through that batch of interrupts.
>>
>> Although driver does not yet support dynamic interrupt allocation, keep
>> allocated interrupts in a pool and add allocation abstraction logic to
>> make code more flexible. Keep per interrupt information in the
>> ice_q_vector structure, which yields ice_vsi::base_vector redundant.
>> Also, as a result there are a few functions that can be removed.
>>
>> Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
>> Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
>> Reviewed-by: Simon Horman <simon.horman@corigine.com>
>> Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
>> Signed-off-by: Piotr Raczynski <piotr.raczynski@intel.com>
>> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
> 
> ...
> 
>> diff --git a/drivers/net/ethernet/intel/ice/ice_base.c b/drivers/net/ethernet/intel/ice/ice_base.c
>> index 1911d644dfa8..7dd7a0f32471 100644
>> --- a/drivers/net/ethernet/intel/ice/ice_base.c
>> +++ b/drivers/net/ethernet/intel/ice/ice_base.c
>> @@ -105,8 +105,7 @@ static int ice_vsi_alloc_q_vector(struct ice_vsi *vsi, u16 v_idx)
>>  	struct ice_q_vector *q_vector;
>>  
>>  	/* allocate q_vector */
>> -	q_vector = devm_kzalloc(ice_pf_to_dev(pf), sizeof(*q_vector),
>> -				GFP_KERNEL);
>> +	q_vector = kzalloc(sizeof(*q_vector), GFP_KERNEL);

Especially since we moved away from devm_kzalloc here so it won't
automatically get released at driver unload. (Which is fine, I think
we're slowly moving away from devm here because we didn't really commit
to using it properly and had devm_kfree a lot anyways...).

>>  	if (!q_vector)
>>  		return -ENOMEM;
>>  
>> @@ -118,9 +117,31 @@ static int ice_vsi_alloc_q_vector(struct ice_vsi *vsi, u16 v_idx)
>>  	q_vector->rx.itr_mode = ITR_DYNAMIC;
>>  	q_vector->tx.type = ICE_TX_CONTAINER;
>>  	q_vector->rx.type = ICE_RX_CONTAINER;
>> +	q_vector->irq.index = -ENOENT;
>>  
>> -	if (vsi->type == ICE_VSI_VF)
>> +	if (vsi->type == ICE_VSI_VF) {
>> +		q_vector->reg_idx = ice_calc_vf_reg_idx(vsi->vf, q_vector);
>>  		goto out;
>> +	} else if (vsi->type == ICE_VSI_CTRL && vsi->vf) {
>> +		struct ice_vsi *ctrl_vsi = ice_get_vf_ctrl_vsi(pf, vsi);
>> +
>> +		if (ctrl_vsi) {
>> +			if (unlikely(!ctrl_vsi->q_vectors))
>> +				return -ENOENT;
> 
> q_vector appears to be leaked here.

Yea that seems like a leak to me too. We allocate q_vector near the
start of the function then perform some lookup checks here per VSI type
to get the index. We wanted to obtain the irq value from the CTRL VSI. I
bet this case is very rare since it would be unlikely that we have a
ctrl_vsi pointer but do *not* have the q_vectors array setup yet.

Probably best if this was refactored a bit to have a cleanup exit label
so that it was more difficult to miss the cleanup.

> 
>> +			q_vector->irq = ctrl_vsi->q_vectors[0]->irq;
>> +			goto skip_alloc;
>> +		}
>> +	}
>> +
>> +	q_vector->irq = ice_alloc_irq(pf);
>> +	if (q_vector->irq.index < 0) {
>> +		kfree(q_vector);
>> +		return -ENOMEM;
>> +	}
>> +
>> +skip_alloc:
>> +	q_vector->reg_idx = q_vector->irq.index;
>> +
>>  	/* only set affinity_mask if the CPU is online */
>>  	if (cpu_online(v_idx))
>>  		cpumask_set_cpu(v_idx, &q_vector->affinity_mask);
> 
> ...

Re: [PATCH net-next 6/8] ice: add individual interrupt allocation

[Simon Horman]

On Wed, May 10, 2023 at 11:33:33AM -0700, Jacob Keller wrote:
> 
> 
> On 5/10/2023 1:54 AM, Simon Horman wrote:
> > On Tue, May 09, 2023 at 10:00:46AM -0700, Tony Nguyen wrote:
> >> From: Piotr Raczynski <piotr.raczynski@intel.com>
> >>
> >> Currently interrupt allocations, depending on a feature are distributed
> >> in batches. Also, after allocation there is a series of operations that
> >> distributes per irq settings through that batch of interrupts.
> >>
> >> Although driver does not yet support dynamic interrupt allocation, keep
> >> allocated interrupts in a pool and add allocation abstraction logic to
> >> make code more flexible. Keep per interrupt information in the
> >> ice_q_vector structure, which yields ice_vsi::base_vector redundant.
> >> Also, as a result there are a few functions that can be removed.
> >>
> >> Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
> >> Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
> >> Reviewed-by: Simon Horman <simon.horman@corigine.com>
> >> Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
> >> Signed-off-by: Piotr Raczynski <piotr.raczynski@intel.com>
> >> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
> > 
> > ...
> > 
> >> diff --git a/drivers/net/ethernet/intel/ice/ice_base.c b/drivers/net/ethernet/intel/ice/ice_base.c
> >> index 1911d644dfa8..7dd7a0f32471 100644
> >> --- a/drivers/net/ethernet/intel/ice/ice_base.c
> >> +++ b/drivers/net/ethernet/intel/ice/ice_base.c
> >> @@ -105,8 +105,7 @@ static int ice_vsi_alloc_q_vector(struct ice_vsi *vsi, u16 v_idx)
> >>  	struct ice_q_vector *q_vector;
> >>  
> >>  	/* allocate q_vector */
> >> -	q_vector = devm_kzalloc(ice_pf_to_dev(pf), sizeof(*q_vector),
> >> -				GFP_KERNEL);
> >> +	q_vector = kzalloc(sizeof(*q_vector), GFP_KERNEL);
> 
> Especially since we moved away from devm_kzalloc here so it won't
> automatically get released at driver unload. (Which is fine, I think
> we're slowly moving away from devm here because we didn't really commit
> to using it properly and had devm_kfree a lot anyways...).
> 
> >>  	if (!q_vector)
> >>  		return -ENOMEM;
> >>  
> >> @@ -118,9 +117,31 @@ static int ice_vsi_alloc_q_vector(struct ice_vsi *vsi, u16 v_idx)
> >>  	q_vector->rx.itr_mode = ITR_DYNAMIC;
> >>  	q_vector->tx.type = ICE_TX_CONTAINER;
> >>  	q_vector->rx.type = ICE_RX_CONTAINER;
> >> +	q_vector->irq.index = -ENOENT;
> >>  
> >> -	if (vsi->type == ICE_VSI_VF)
> >> +	if (vsi->type == ICE_VSI_VF) {
> >> +		q_vector->reg_idx = ice_calc_vf_reg_idx(vsi->vf, q_vector);
> >>  		goto out;
> >> +	} else if (vsi->type == ICE_VSI_CTRL && vsi->vf) {
> >> +		struct ice_vsi *ctrl_vsi = ice_get_vf_ctrl_vsi(pf, vsi);
> >> +
> >> +		if (ctrl_vsi) {
> >> +			if (unlikely(!ctrl_vsi->q_vectors))
> >> +				return -ENOENT;
> > 
> > q_vector appears to be leaked here.
> 
> Yea that seems like a leak to me too. We allocate q_vector near the
> start of the function then perform some lookup checks here per VSI type
> to get the index. We wanted to obtain the irq value from the CTRL VSI. I
> bet this case is very rare since it would be unlikely that we have a
> ctrl_vsi pointer but do *not* have the q_vectors array setup yet.
> 
> Probably best if this was refactored a bit to have a cleanup exit label
> so that it was more difficult to miss the cleanup.

Thanks Jacob, agreed.

> >> +			q_vector->irq = ctrl_vsi->q_vectors[0]->irq;
> >> +			goto skip_alloc;
> >> +		}
> >> +	}
> >> +
> >> +	q_vector->irq = ice_alloc_irq(pf);
> >> +	if (q_vector->irq.index < 0) {
> >> +		kfree(q_vector);
> >> +		return -ENOMEM;
> >> +	}
> >> +
> >> +skip_alloc:
> >> +	q_vector->reg_idx = q_vector->irq.index;
> >> +
> >>  	/* only set affinity_mask if the CPU is online */
> >>  	if (cpu_online(v_idx))
> >>  		cpumask_set_cpu(v_idx, &q_vector->affinity_mask);
> > 
> > ...