[PATCH net-next v2 0/9][pull request] 1GbE Intel Wired LAN Driver Updates 2021-05-20

[PATCH net-next v2 0/9][pull request] 1GbE Intel Wired LAN Driver Updates 2021-05-20

[Tony Nguyen]

This series contains updates to igc driver only.

Andre Guedes says:

This series adds AF_XDP zero-copy feature to igc driver.

The initial patches do some code refactoring, preparing the code base to
land the AF_XDP zero-copy feature, avoiding code duplications. The last
patches of the series are the ones implementing the feature.

The last patch which indeed implements AF_XDP zero-copy support was
originally way too lengthy so, for the sake of code review, I broke it
up into two patches: one adding support for the RX functionality and the
other one adding TX support.
---
v2:
Patch 8/9 - "igc: Enable RX via AF_XDP zero-copy"
 * In XDP_PASS flow, copy metadata too into the skb.
 * When HW timestamp is added by the NIC, after copying it into
   a local variable, update xdp_buff->data_meta so that
   metadata length when XDP program is called 0.
 * In igc_xdp_enable_pool(), call xsk_pool_dma_unmap() on
   failure.

Known issues:
 When an XDP application is running in Tx-Only mode with Zero-Copy
 enabled, it is not expected to add the frames to the fill-queue. I have
 noticed the following two issues in this scenario:
 - If XDP_USE_NEED_WAKEUP flag is not set by application, igc_poll()
   will go into infinite loop because the buffer allocation resulting
   in igc_clean_rx_irq_zc() indicating that all work is not done and NAPI
   should keep polling. This does not occur if XDP_USE_NEED_WAKEUP
   flag is set.
 - Since there are no buffers allocated by userspace for the fill
   queue, there is no memory allocated for the NIC to copy the data
   to. If the packet received is destined to the hardware queue where
   XDP application is running, no packets are received even on other
   queues.
 Both these issues can be mitigated by adding a few frames to the
 fill queue. The second issue can also be mitigated by making sure no
 packets are being received on the hardware queue where Rx is running.

The following are changes since commit a49e72b3bda73d36664a084e47da9727a31b8095:
  net: qrtr: ns: Fix error return code in qrtr_ns_init()
and are available in the git repository at:
  git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue 1GbE

Andre Guedes (9):
  igc: Move igc_xdp_is_enabled()
  igc: Refactor __igc_xdp_run_prog()
  igc: Refactor igc_clean_rx_ring()
  igc: Refactor XDP rxq info registration
  igc: Introduce TX/RX stats helpers
  igc: Introduce igc_unmap_tx_buffer() helper
  igc: Replace IGC_TX_FLAGS_XDP flag by an enum
  igc: Enable RX via AF_XDP zero-copy
  igc: Enable TX via AF_XDP zero-copy

 drivers/net/ethernet/intel/igc/igc.h      |  33 +-
 drivers/net/ethernet/intel/igc/igc_base.h |   2 +
 drivers/net/ethernet/intel/igc/igc_main.c | 656 ++++++++++++++++++----
 drivers/net/ethernet/intel/igc/igc_xdp.c  | 109 +++-
 drivers/net/ethernet/intel/igc/igc_xdp.h  |   8 +-
 5 files changed, 681 insertions(+), 127 deletions(-)

-- 
2.26.2

[PATCH net-next v2 1/9] igc: Move igc_xdp_is_enabled()

[Tony Nguyen]

From: Andre Guedes <andre.guedes@intel.com>

Move the helper igc_xdp_is_enabled() to igc_xdp.h so it can be reused in
igc_xdp.c by upcoming patches that will introduce AF_XDP zero-copy
support to the driver.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: Jithu Joseph <jithu.joseph@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/igc/igc_main.c | 5 -----
 drivers/net/ethernet/intel/igc/igc_xdp.h  | 5 +++++
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethernet/intel/igc/igc_main.c
index 92c0701e2a36..edfe9d492071 100644
--- a/drivers/net/ethernet/intel/igc/igc_main.c
+++ b/drivers/net/ethernet/intel/igc/igc_main.c
@@ -515,11 +515,6 @@ static int igc_setup_all_rx_resources(struct igc_adapter *adapter)
 	return err;
 }
 
-static bool igc_xdp_is_enabled(struct igc_adapter *adapter)
-{
-	return !!adapter->xdp_prog;
-}
-
 /**
  * igc_configure_rx_ring - Configure a receive ring after Reset
  * @adapter: board private structure
diff --git a/drivers/net/ethernet/intel/igc/igc_xdp.h b/drivers/net/ethernet/intel/igc/igc_xdp.h
index cfecb515b718..412aa369e6ba 100644
--- a/drivers/net/ethernet/intel/igc/igc_xdp.h
+++ b/drivers/net/ethernet/intel/igc/igc_xdp.h
@@ -10,4 +10,9 @@ int igc_xdp_set_prog(struct igc_adapter *adapter, struct bpf_prog *prog,
 int igc_xdp_register_rxq_info(struct igc_ring *ring);
 void igc_xdp_unregister_rxq_info(struct igc_ring *ring);
 
+static inline bool igc_xdp_is_enabled(struct igc_adapter *adapter)
+{
+	return !!adapter->xdp_prog;
+}
+
 #endif /* _IGC_XDP_H_ */
-- 
2.26.2

[PATCH net-next v2 2/9] igc: Refactor __igc_xdp_run_prog()

[Tony Nguyen]

From: Andre Guedes <andre.guedes@intel.com>

Refactor __igc_xdp_run_prog() helper from igc_xdp_run_prog(),
preparing the code for AF_XDP zero-copy support which is added
by upcoming patches.

The existing igc_xdp_run_prog() caters to regular XDP rx path
which has to verify if bpf_prog is not NULL. Zero-copy
path assumes that bpf_prog is not NULL and hence this check is
not required. Therefore it makes sense to refactor the common
code into a helper function, to avoid code duplication.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: Jithu Joseph <jithu.joseph@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/igc/igc_main.c | 56 +++++++++++------------
 1 file changed, 28 insertions(+), 28 deletions(-)

diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethernet/intel/igc/igc_main.c
index edfe9d492071..1961cc667c3b 100644
--- a/drivers/net/ethernet/intel/igc/igc_main.c
+++ b/drivers/net/ethernet/intel/igc/igc_main.c
@@ -2020,38 +2020,22 @@ static int igc_xdp_xmit_back(struct igc_adapter *adapter, struct xdp_buff *xdp)
 	return res;
 }
 
-static struct sk_buff *igc_xdp_run_prog(struct igc_adapter *adapter,
-					struct xdp_buff *xdp)
+/* This function assumes rcu_read_lock() is held by the caller. */
+static int __igc_xdp_run_prog(struct igc_adapter *adapter,
+			      struct bpf_prog *prog,
+			      struct xdp_buff *xdp)
 {
-	struct bpf_prog *prog;
-	int res;
-	u32 act;
-
-	rcu_read_lock();
-
-	prog = READ_ONCE(adapter->xdp_prog);
-	if (!prog) {
-		res = IGC_XDP_PASS;
-		goto unlock;
-	}
+	u32 act = bpf_prog_run_xdp(prog, xdp);
 
-	act = bpf_prog_run_xdp(prog, xdp);
 	switch (act) {
 	case XDP_PASS:
-		res = IGC_XDP_PASS;
-		break;
+		return IGC_XDP_PASS;
 	case XDP_TX:
-		if (igc_xdp_xmit_back(adapter, xdp) < 0)
-			res = IGC_XDP_CONSUMED;
-		else
-			res = IGC_XDP_TX;
-		break;
+		return igc_xdp_xmit_back(adapter, xdp) < 0 ?
+			IGC_XDP_CONSUMED : IGC_XDP_TX;
 	case XDP_REDIRECT:
-		if (xdp_do_redirect(adapter->netdev, xdp, prog) < 0)
-			res = IGC_XDP_CONSUMED;
-		else
-			res = IGC_XDP_REDIRECT;
-		break;
+		return xdp_do_redirect(adapter->netdev, xdp, prog) < 0 ?
+			IGC_XDP_CONSUMED : IGC_XDP_REDIRECT;
 	default:
 		bpf_warn_invalid_xdp_action(act);
 		fallthrough;
@@ -2059,9 +2043,25 @@ static struct sk_buff *igc_xdp_run_prog(struct igc_adapter *adapter,
 		trace_xdp_exception(adapter->netdev, prog, act);
 		fallthrough;
 	case XDP_DROP:
-		res = IGC_XDP_CONSUMED;
-		break;
+		return IGC_XDP_CONSUMED;
 	}
+}
+
+static struct sk_buff *igc_xdp_run_prog(struct igc_adapter *adapter,
+					struct xdp_buff *xdp)
+{
+	struct bpf_prog *prog;
+	int res;
+
+	rcu_read_lock();
+
+	prog = READ_ONCE(adapter->xdp_prog);
+	if (!prog) {
+		res = IGC_XDP_PASS;
+		goto unlock;
+	}
+
+	res = __igc_xdp_run_prog(adapter, prog, xdp);
 
 unlock:
 	rcu_read_unlock();
-- 
2.26.2

[PATCH net-next v2 3/9] igc: Refactor igc_clean_rx_ring()

[Tony Nguyen]

From: Andre Guedes <andre.guedes@intel.com>

Refactor igc_clean_rx_ring() helper, preparing the code for AF_XDP
zero-copy support which is added by upcoming patches.

The refactor consists of encapsulating page-shared specific code into
its own helper, leaving common code that will be shared by both
page-shared and xsk pool in igc_clean_rx_ring().

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: Jithu Joseph <jithu.joseph@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/igc/igc_main.c | 23 ++++++++++++++---------
 1 file changed, 14 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethernet/intel/igc/igc_main.c
index 1961cc667c3b..5024f8dc98f9 100644
--- a/drivers/net/ethernet/intel/igc/igc_main.c
+++ b/drivers/net/ethernet/intel/igc/igc_main.c
@@ -346,11 +346,7 @@ static int igc_setup_all_tx_resources(struct igc_adapter *adapter)
 	return err;
 }
 
-/**
- * igc_clean_rx_ring - Free Rx Buffers per Queue
- * @rx_ring: ring to free buffers from
- */
-static void igc_clean_rx_ring(struct igc_ring *rx_ring)
+static void igc_clean_rx_ring_page_shared(struct igc_ring *rx_ring)
 {
 	u16 i = rx_ring->next_to_clean;
 
@@ -383,12 +379,21 @@ static void igc_clean_rx_ring(struct igc_ring *rx_ring)
 		if (i == rx_ring->count)
 			i = 0;
 	}
+}
+
+/**
+ * igc_clean_rx_ring - Free Rx Buffers per Queue
+ * @ring: ring to free buffers from
+ */
+static void igc_clean_rx_ring(struct igc_ring *ring)
+{
+	igc_clean_rx_ring_page_shared(ring);
 
-	clear_ring_uses_large_buffer(rx_ring);
+	clear_ring_uses_large_buffer(ring);
 
-	rx_ring->next_to_alloc = 0;
-	rx_ring->next_to_clean = 0;
-	rx_ring->next_to_use = 0;
+	ring->next_to_alloc = 0;
+	ring->next_to_clean = 0;
+	ring->next_to_use = 0;
 }
 
 /**
-- 
2.26.2

[PATCH net-next v2 4/9] igc: Refactor XDP rxq info registration

[Tony Nguyen]

From: Andre Guedes <andre.guedes@intel.com>

Refactor XDP rxq info registration code, preparing the driver for AF_XDP
zero-copy support which is added by upcoming patches.

Currently, xdp_rxq and memory model are both registered during RX
resource setup time by igc_xdp_register_rxq_info() helper. With AF_XDP,
we want to register the memory model later on while configuring the ring
because we will know which memory model type to register
(MEM_TYPE_PAGE_SHARED or MEM_TYPE_XSK_BUFF_POOL).

The helpers igc_xdp_register_rxq_info() and igc_xdp_unregister_rxq_
info() are not useful anymore so they are removed.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: Jithu Joseph <jithu.joseph@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/igc/igc_main.c | 16 ++++++++++----
 drivers/net/ethernet/intel/igc/igc_xdp.c  | 27 -----------------------
 drivers/net/ethernet/intel/igc/igc_xdp.h  |  3 ---
 3 files changed, 12 insertions(+), 34 deletions(-)

diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethernet/intel/igc/igc_main.c
index 5024f8dc98f9..90d8e4355f43 100644
--- a/drivers/net/ethernet/intel/igc/igc_main.c
+++ b/drivers/net/ethernet/intel/igc/igc_main.c
@@ -419,7 +419,7 @@ void igc_free_rx_resources(struct igc_ring *rx_ring)
 {
 	igc_clean_rx_ring(rx_ring);
 
-	igc_xdp_unregister_rxq_info(rx_ring);
+	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
 
 	vfree(rx_ring->rx_buffer_info);
 	rx_ring->rx_buffer_info = NULL;
@@ -458,11 +458,16 @@ int igc_setup_rx_resources(struct igc_ring *rx_ring)
 {
 	struct net_device *ndev = rx_ring->netdev;
 	struct device *dev = rx_ring->dev;
+	u8 index = rx_ring->queue_index;
 	int size, desc_len, res;
 
-	res = igc_xdp_register_rxq_info(rx_ring);
-	if (res < 0)
+	res = xdp_rxq_info_reg(&rx_ring->xdp_rxq, ndev, index,
+			       rx_ring->q_vector->napi.napi_id);
+	if (res < 0) {
+		netdev_err(ndev, "Failed to register xdp_rxq index %u\n",
+			   index);
 		return res;
+	}
 
 	size = sizeof(struct igc_rx_buffer) * rx_ring->count;
 	rx_ring->rx_buffer_info = vzalloc(size);
@@ -488,7 +493,7 @@ int igc_setup_rx_resources(struct igc_ring *rx_ring)
 	return 0;
 
 err:
-	igc_xdp_unregister_rxq_info(rx_ring);
+	xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
 	vfree(rx_ring->rx_buffer_info);
 	rx_ring->rx_buffer_info = NULL;
 	netdev_err(ndev, "Unable to allocate memory for Rx descriptor ring\n");
@@ -536,6 +541,9 @@ static void igc_configure_rx_ring(struct igc_adapter *adapter,
 	u32 srrctl = 0, rxdctl = 0;
 	u64 rdba = ring->dma;
 
+	WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
+					   MEM_TYPE_PAGE_SHARED, NULL));
+
 	if (igc_xdp_is_enabled(adapter))
 		set_ring_uses_large_buffer(ring);
 
diff --git a/drivers/net/ethernet/intel/igc/igc_xdp.c b/drivers/net/ethernet/intel/igc/igc_xdp.c
index 11133c4619bb..27c886a254f1 100644
--- a/drivers/net/ethernet/intel/igc/igc_xdp.c
+++ b/drivers/net/ethernet/intel/igc/igc_xdp.c
@@ -31,30 +31,3 @@ int igc_xdp_set_prog(struct igc_adapter *adapter, struct bpf_prog *prog,
 
 	return 0;
 }
-
-int igc_xdp_register_rxq_info(struct igc_ring *ring)
-{
-	struct net_device *dev = ring->netdev;
-	int err;
-
-	err = xdp_rxq_info_reg(&ring->xdp_rxq, dev, ring->queue_index, 0);
-	if (err) {
-		netdev_err(dev, "Failed to register xdp rxq info\n");
-		return err;
-	}
-
-	err = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq, MEM_TYPE_PAGE_SHARED,
-					 NULL);
-	if (err) {
-		netdev_err(dev, "Failed to register xdp rxq mem model\n");
-		xdp_rxq_info_unreg(&ring->xdp_rxq);
-		return err;
-	}
-
-	return 0;
-}
-
-void igc_xdp_unregister_rxq_info(struct igc_ring *ring)
-{
-	xdp_rxq_info_unreg(&ring->xdp_rxq);
-}
diff --git a/drivers/net/ethernet/intel/igc/igc_xdp.h b/drivers/net/ethernet/intel/igc/igc_xdp.h
index 412aa369e6ba..cdaa2c39b03a 100644
--- a/drivers/net/ethernet/intel/igc/igc_xdp.h
+++ b/drivers/net/ethernet/intel/igc/igc_xdp.h
@@ -7,9 +7,6 @@
 int igc_xdp_set_prog(struct igc_adapter *adapter, struct bpf_prog *prog,
 		     struct netlink_ext_ack *extack);
 
-int igc_xdp_register_rxq_info(struct igc_ring *ring);
-void igc_xdp_unregister_rxq_info(struct igc_ring *ring);
-
 static inline bool igc_xdp_is_enabled(struct igc_adapter *adapter)
 {
 	return !!adapter->xdp_prog;
-- 
2.26.2

[PATCH net-next v2 5/9] igc: Introduce TX/RX stats helpers

[Tony Nguyen]

From: Andre Guedes <andre.guedes@intel.com>

In preparation for AF_XDP zero-copy support, encapsulate the code that
updates the driver RX stats in its own local helper so it can be reused
in the zero-copy path. Likewise, encapsulate TX stats code as well.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: Jithu Joseph <jithu.joseph@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/igc/igc_main.c | 43 ++++++++++++++++-------
 1 file changed, 31 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethernet/intel/igc/igc_main.c
index 90d8e4355f43..5a262bab9116 100644
--- a/drivers/net/ethernet/intel/igc/igc_main.c
+++ b/drivers/net/ethernet/intel/igc/igc_main.c
@@ -2111,6 +2111,20 @@ static void igc_finalize_xdp(struct igc_adapter *adapter, int status)
 		xdp_do_flush();
 }
 
+static void igc_update_rx_stats(struct igc_q_vector *q_vector,
+				unsigned int packets, unsigned int bytes)
+{
+	struct igc_ring *ring = q_vector->rx.ring;
+
+	u64_stats_update_begin(&ring->rx_syncp);
+	ring->rx_stats.packets += packets;
+	ring->rx_stats.bytes += bytes;
+	u64_stats_update_end(&ring->rx_syncp);
+
+	q_vector->rx.total_packets += packets;
+	q_vector->rx.total_bytes += bytes;
+}
+
 static int igc_clean_rx_irq(struct igc_q_vector *q_vector, const int budget)
 {
 	unsigned int total_bytes = 0, total_packets = 0;
@@ -2231,12 +2245,7 @@ static int igc_clean_rx_irq(struct igc_q_vector *q_vector, const int budget)
 	/* place incomplete frames back on ring for completion */
 	rx_ring->skb = skb;
 
-	u64_stats_update_begin(&rx_ring->rx_syncp);
-	rx_ring->rx_stats.packets += total_packets;
-	rx_ring->rx_stats.bytes += total_bytes;
-	u64_stats_update_end(&rx_ring->rx_syncp);
-	q_vector->rx.total_packets += total_packets;
-	q_vector->rx.total_bytes += total_bytes;
+	igc_update_rx_stats(q_vector, total_packets, total_bytes);
 
 	if (cleaned_count)
 		igc_alloc_rx_buffers(rx_ring, cleaned_count);
@@ -2244,6 +2253,20 @@ static int igc_clean_rx_irq(struct igc_q_vector *q_vector, const int budget)
 	return total_packets;
 }
 
+static void igc_update_tx_stats(struct igc_q_vector *q_vector,
+				unsigned int packets, unsigned int bytes)
+{
+	struct igc_ring *ring = q_vector->tx.ring;
+
+	u64_stats_update_begin(&ring->tx_syncp);
+	ring->tx_stats.bytes += bytes;
+	ring->tx_stats.packets += packets;
+	u64_stats_update_end(&ring->tx_syncp);
+
+	q_vector->tx.total_bytes += bytes;
+	q_vector->tx.total_packets += packets;
+}
+
 /**
  * igc_clean_tx_irq - Reclaim resources after transmit completes
  * @q_vector: pointer to q_vector containing needed info
@@ -2346,12 +2369,8 @@ static bool igc_clean_tx_irq(struct igc_q_vector *q_vector, int napi_budget)
 
 	i += tx_ring->count;
 	tx_ring->next_to_clean = i;
-	u64_stats_update_begin(&tx_ring->tx_syncp);
-	tx_ring->tx_stats.bytes += total_bytes;
-	tx_ring->tx_stats.packets += total_packets;
-	u64_stats_update_end(&tx_ring->tx_syncp);
-	q_vector->tx.total_bytes += total_bytes;
-	q_vector->tx.total_packets += total_packets;
+
+	igc_update_tx_stats(q_vector, total_packets, total_bytes);
 
 	if (test_bit(IGC_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags)) {
 		struct igc_hw *hw = &adapter->hw;
-- 
2.26.2

[PATCH net-next v2 6/9] igc: Introduce igc_unmap_tx_buffer() helper

[Tony Nguyen]

From: Andre Guedes <andre.guedes@intel.com>

In preparation for AF_XDP zero-copy support, encapsulate the code that
unmaps Tx buffers into its own local helper so we can reuse it, avoiding
code duplication.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: Jithu Joseph <jithu.joseph@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/igc/igc_main.c | 49 +++++++----------------
 1 file changed, 15 insertions(+), 34 deletions(-)

diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethernet/intel/igc/igc_main.c
index 5a262bab9116..ebb57d82e4dd 100644
--- a/drivers/net/ethernet/intel/igc/igc_main.c
+++ b/drivers/net/ethernet/intel/igc/igc_main.c
@@ -171,6 +171,14 @@ static void igc_get_hw_control(struct igc_adapter *adapter)
 	     ctrl_ext | IGC_CTRL_EXT_DRV_LOAD);
 }
 
+static void igc_unmap_tx_buffer(struct device *dev, struct igc_tx_buffer *buf)
+{
+	dma_unmap_single(dev, dma_unmap_addr(buf, dma),
+			 dma_unmap_len(buf, len), DMA_TO_DEVICE);
+
+	dma_unmap_len_set(buf, len, 0);
+}
+
 /**
  * igc_clean_tx_ring - Free Tx Buffers
  * @tx_ring: ring to be cleaned
@@ -188,11 +196,7 @@ static void igc_clean_tx_ring(struct igc_ring *tx_ring)
 		else
 			dev_kfree_skb_any(tx_buffer->skb);
 
-		/* unmap skb header data */
-		dma_unmap_single(tx_ring->dev,
-				 dma_unmap_addr(tx_buffer, dma),
-				 dma_unmap_len(tx_buffer, len),
-				 DMA_TO_DEVICE);
+		igc_unmap_tx_buffer(tx_ring->dev, tx_buffer);
 
 		/* check for eop_desc to determine the end of the packet */
 		eop_desc = tx_buffer->next_to_watch;
@@ -211,10 +215,7 @@ static void igc_clean_tx_ring(struct igc_ring *tx_ring)
 
 			/* unmap any remaining paged data */
 			if (dma_unmap_len(tx_buffer, len))
-				dma_unmap_page(tx_ring->dev,
-					       dma_unmap_addr(tx_buffer, dma),
-					       dma_unmap_len(tx_buffer, len),
-					       DMA_TO_DEVICE);
+				igc_unmap_tx_buffer(tx_ring->dev, tx_buffer);
 		}
 
 		/* move us one more past the eop_desc for start of next pkt */
@@ -1219,11 +1220,7 @@ static int igc_tx_map(struct igc_ring *tx_ring,
 	/* clear dma mappings for failed tx_buffer_info map */
 	while (tx_buffer != first) {
 		if (dma_unmap_len(tx_buffer, len))
-			dma_unmap_page(tx_ring->dev,
-				       dma_unmap_addr(tx_buffer, dma),
-				       dma_unmap_len(tx_buffer, len),
-				       DMA_TO_DEVICE);
-		dma_unmap_len_set(tx_buffer, len, 0);
+			igc_unmap_tx_buffer(tx_ring->dev, tx_buffer);
 
 		if (i-- == 0)
 			i += tx_ring->count;
@@ -1231,11 +1228,7 @@ static int igc_tx_map(struct igc_ring *tx_ring,
 	}
 
 	if (dma_unmap_len(tx_buffer, len))
-		dma_unmap_single(tx_ring->dev,
-				 dma_unmap_addr(tx_buffer, dma),
-				 dma_unmap_len(tx_buffer, len),
-				 DMA_TO_DEVICE);
-	dma_unmap_len_set(tx_buffer, len, 0);
+		igc_unmap_tx_buffer(tx_ring->dev, tx_buffer);
 
 	dev_kfree_skb_any(tx_buffer->skb);
 	tx_buffer->skb = NULL;
@@ -2317,14 +2310,7 @@ static bool igc_clean_tx_irq(struct igc_q_vector *q_vector, int napi_budget)
 		else
 			napi_consume_skb(tx_buffer->skb, napi_budget);
 
-		/* unmap skb header data */
-		dma_unmap_single(tx_ring->dev,
-				 dma_unmap_addr(tx_buffer, dma),
-				 dma_unmap_len(tx_buffer, len),
-				 DMA_TO_DEVICE);
-
-		/* clear tx_buffer data */
-		dma_unmap_len_set(tx_buffer, len, 0);
+		igc_unmap_tx_buffer(tx_ring->dev, tx_buffer);
 
 		/* clear last DMA location and unmap remaining buffers */
 		while (tx_desc != eop_desc) {
@@ -2338,13 +2324,8 @@ static bool igc_clean_tx_irq(struct igc_q_vector *q_vector, int napi_budget)
 			}
 
 			/* unmap any remaining paged data */
-			if (dma_unmap_len(tx_buffer, len)) {
-				dma_unmap_page(tx_ring->dev,
-					       dma_unmap_addr(tx_buffer, dma),
-					       dma_unmap_len(tx_buffer, len),
-					       DMA_TO_DEVICE);
-				dma_unmap_len_set(tx_buffer, len, 0);
-			}
+			if (dma_unmap_len(tx_buffer, len))
+				igc_unmap_tx_buffer(tx_ring->dev, tx_buffer);
 		}
 
 		/* move us one more past the eop_desc for start of next pkt */
-- 
2.26.2

[PATCH net-next v2 7/9] igc: Replace IGC_TX_FLAGS_XDP flag by an enum

[Tony Nguyen]

From: Andre Guedes <andre.guedes@intel.com>

Up to this point, Tx buffers are associated with either a skb or a xdpf,
and the IGC_TX_FLAGS_XDP flag was enough to distinguish between these
two case. However, with upcoming patches that will add AF_XDP zero-copy
support, a third case will be introduced so this flag-based approach
won't fit well.

In preparation to land AF_XDP zero-copy support, replace the
IGC_TX_FLAGS_XDP flag by an enum which will be extended once zero-copy
support is introduced to the driver.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: Jithu Joseph <jithu.joseph@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/igc/igc.h      |  8 ++++++--
 drivers/net/ethernet/intel/igc/igc_main.c | 25 ++++++++++++++++++-----
 2 files changed, 26 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/intel/igc/igc.h b/drivers/net/ethernet/intel/igc/igc.h
index 25871351730b..4815c520652b 100644
--- a/drivers/net/ethernet/intel/igc/igc.h
+++ b/drivers/net/ethernet/intel/igc/igc.h
@@ -390,8 +390,6 @@ enum igc_tx_flags {
 	/* olinfo flags */
 	IGC_TX_FLAGS_IPV4	= 0x10,
 	IGC_TX_FLAGS_CSUM	= 0x20,
-
-	IGC_TX_FLAGS_XDP	= 0x100,
 };
 
 enum igc_boards {
@@ -408,12 +406,18 @@ enum igc_boards {
 #define TXD_USE_COUNT(S)	DIV_ROUND_UP((S), IGC_MAX_DATA_PER_TXD)
 #define DESC_NEEDED	(MAX_SKB_FRAGS + 4)
 
+enum igc_tx_buffer_type {
+	IGC_TX_BUFFER_TYPE_SKB,
+	IGC_TX_BUFFER_TYPE_XDP,
+};
+
 /* wrapper around a pointer to a socket buffer,
  * so a DMA handle can be stored along with the buffer
  */
 struct igc_tx_buffer {
 	union igc_adv_tx_desc *next_to_watch;
 	unsigned long time_stamp;
+	enum igc_tx_buffer_type type;
 	union {
 		struct sk_buff *skb;
 		struct xdp_frame *xdpf;
diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethernet/intel/igc/igc_main.c
index ebb57d82e4dd..4e1327a5a61e 100644
--- a/drivers/net/ethernet/intel/igc/igc_main.c
+++ b/drivers/net/ethernet/intel/igc/igc_main.c
@@ -191,10 +191,17 @@ static void igc_clean_tx_ring(struct igc_ring *tx_ring)
 	while (i != tx_ring->next_to_use) {
 		union igc_adv_tx_desc *eop_desc, *tx_desc;
 
-		if (tx_buffer->tx_flags & IGC_TX_FLAGS_XDP)
+		switch (tx_buffer->type) {
+		case IGC_TX_BUFFER_TYPE_XDP:
 			xdp_return_frame(tx_buffer->xdpf);
-		else
+			break;
+		case IGC_TX_BUFFER_TYPE_SKB:
 			dev_kfree_skb_any(tx_buffer->skb);
+			break;
+		default:
+			netdev_warn_once(tx_ring->netdev, "Unknown Tx buffer type\n");
+			break;
+		}
 
 		igc_unmap_tx_buffer(tx_ring->dev, tx_buffer);
 
@@ -1360,6 +1367,7 @@ static netdev_tx_t igc_xmit_frame_ring(struct sk_buff *skb,
 
 	/* record the location of the first descriptor for this packet */
 	first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
+	first->type = IGC_TX_BUFFER_TYPE_SKB;
 	first->skb = skb;
 	first->bytecount = skb->len;
 	first->gso_segs = 1;
@@ -1943,8 +1951,8 @@ static int igc_xdp_init_tx_buffer(struct igc_tx_buffer *buffer,
 		return -ENOMEM;
 	}
 
+	buffer->type = IGC_TX_BUFFER_TYPE_XDP;
 	buffer->xdpf = xdpf;
-	buffer->tx_flags = IGC_TX_FLAGS_XDP;
 	buffer->protocol = 0;
 	buffer->bytecount = xdpf->len;
 	buffer->gso_segs = 1;
@@ -2305,10 +2313,17 @@ static bool igc_clean_tx_irq(struct igc_q_vector *q_vector, int napi_budget)
 		total_bytes += tx_buffer->bytecount;
 		total_packets += tx_buffer->gso_segs;
 
-		if (tx_buffer->tx_flags & IGC_TX_FLAGS_XDP)
+		switch (tx_buffer->type) {
+		case IGC_TX_BUFFER_TYPE_XDP:
 			xdp_return_frame(tx_buffer->xdpf);
-		else
+			break;
+		case IGC_TX_BUFFER_TYPE_SKB:
 			napi_consume_skb(tx_buffer->skb, napi_budget);
+			break;
+		default:
+			netdev_warn_once(tx_ring->netdev, "Unknown Tx buffer type\n");
+			break;
+		}
 
 		igc_unmap_tx_buffer(tx_ring->dev, tx_buffer);
 
-- 
2.26.2

[PATCH net-next v2 8/9] igc: Enable RX via AF_XDP zero-copy

[Tony Nguyen]

From: Andre Guedes <andre.guedes@intel.com>

Add support for receiving packets via AF_XDP zero-copy mechanism.

Add a new flag to 'enum igc_ring_flags_t' to indicate the ring has
AF_XDP zero-copy enabled so proper ring setup is carried out during ring
configuration in igc_configure_rx_ring().

RX buffers can now be allocated via the shared pages mechanism (default
behavior of the driver) or via xsk pool (when AF_XDP zero-copy is
enabled) so a union is added to the 'struct igc_rx_buffer' to cover both
cases.

When AF_XDP zero-copy is enabled, rx buffers are allocated from the xsk
pool using the new helper igc_alloc_rx_buffers_zc() which is the
counterpart of igc_alloc_rx_buffers().

Likewise other Intel drivers that support AF_XDP zero-copy, in igc we
have a dedicated path for cleaning up rx irqs when zero-copy is enabled.
This avoids adding too many checks within igc_clean_rx_irq(), resulting
in a more readable and efficient code since this function is called from
the hot-path of the driver.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: Jithu Joseph <jithu.joseph@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/igc/igc.h      |  22 +-
 drivers/net/ethernet/intel/igc/igc_base.h |   1 +
 drivers/net/ethernet/intel/igc/igc_main.c | 344 +++++++++++++++++++++-
 drivers/net/ethernet/intel/igc/igc_xdp.c  | 100 +++++++
 drivers/net/ethernet/intel/igc/igc_xdp.h  |   2 +
 5 files changed, 450 insertions(+), 19 deletions(-)

diff --git a/drivers/net/ethernet/intel/igc/igc.h b/drivers/net/ethernet/intel/igc/igc.h
index 4815c520652b..cd6f4c94c4dd 100644
--- a/drivers/net/ethernet/intel/igc/igc.h
+++ b/drivers/net/ethernet/intel/igc/igc.h
@@ -118,6 +118,7 @@ struct igc_ring {
 	};
 
 	struct xdp_rxq_info xdp_rxq;
+	struct xsk_buff_pool *xsk_pool;
 } ____cacheline_internodealigned_in_smp;
 
 /* Board specific private data structure */
@@ -255,6 +256,9 @@ bool igc_has_link(struct igc_adapter *adapter);
 void igc_reset(struct igc_adapter *adapter);
 int igc_set_spd_dplx(struct igc_adapter *adapter, u32 spd, u8 dplx);
 void igc_update_stats(struct igc_adapter *adapter);
+void igc_disable_rx_ring(struct igc_ring *ring);
+void igc_enable_rx_ring(struct igc_ring *ring);
+int igc_xsk_wakeup(struct net_device *dev, u32 queue_id, u32 flags);
 
 /* igc_dump declarations */
 void igc_rings_dump(struct igc_adapter *adapter);
@@ -432,14 +436,19 @@ struct igc_tx_buffer {
 };
 
 struct igc_rx_buffer {
-	dma_addr_t dma;
-	struct page *page;
+	union {
+		struct {
+			dma_addr_t dma;
+			struct page *page;
 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
-	__u32 page_offset;
+			__u32 page_offset;
 #else
-	__u16 page_offset;
+			__u16 page_offset;
 #endif
-	__u16 pagecnt_bias;
+			__u16 pagecnt_bias;
+		};
+		struct xdp_buff *xdp;
+	};
 };
 
 struct igc_q_vector {
@@ -525,7 +534,8 @@ enum igc_ring_flags_t {
 	IGC_RING_FLAG_RX_SCTP_CSUM,
 	IGC_RING_FLAG_RX_LB_VLAN_BSWAP,
 	IGC_RING_FLAG_TX_CTX_IDX,
-	IGC_RING_FLAG_TX_DETECT_HANG
+	IGC_RING_FLAG_TX_DETECT_HANG,
+	IGC_RING_FLAG_AF_XDP_ZC,
 };
 
 #define ring_uses_large_buffer(ring) \
diff --git a/drivers/net/ethernet/intel/igc/igc_base.h b/drivers/net/ethernet/intel/igc/igc_base.h
index ea627ce52525..2ca028c1919f 100644
--- a/drivers/net/ethernet/intel/igc/igc_base.h
+++ b/drivers/net/ethernet/intel/igc/igc_base.h
@@ -81,6 +81,7 @@ union igc_adv_rx_desc {
 
 /* Additional Receive Descriptor Control definitions */
 #define IGC_RXDCTL_QUEUE_ENABLE	0x02000000 /* Ena specific Rx Queue */
+#define IGC_RXDCTL_SWFLUSH		0x04000000 /* Receive Software Flush */
 
 /* SRRCTL bit definitions */
 #define IGC_SRRCTL_BSIZEPKT_SHIFT		10 /* Shift _right_ */
diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethernet/intel/igc/igc_main.c
index 4e1327a5a61e..3ffc20fae4c6 100644
--- a/drivers/net/ethernet/intel/igc/igc_main.c
+++ b/drivers/net/ethernet/intel/igc/igc_main.c
@@ -11,7 +11,7 @@
 #include <linux/pm_runtime.h>
 #include <net/pkt_sched.h>
 #include <linux/bpf_trace.h>
-
+#include <net/xdp_sock_drv.h>
 #include <net/ipv6.h>
 
 #include "igc.h"
@@ -389,13 +389,31 @@ static void igc_clean_rx_ring_page_shared(struct igc_ring *rx_ring)
 	}
 }
 
+static void igc_clean_rx_ring_xsk_pool(struct igc_ring *ring)
+{
+	struct igc_rx_buffer *bi;
+	u16 i;
+
+	for (i = 0; i < ring->count; i++) {
+		bi = &ring->rx_buffer_info[i];
+		if (!bi->xdp)
+			continue;
+
+		xsk_buff_free(bi->xdp);
+		bi->xdp = NULL;
+	}
+}
+
 /**
  * igc_clean_rx_ring - Free Rx Buffers per Queue
  * @ring: ring to free buffers from
  */
 static void igc_clean_rx_ring(struct igc_ring *ring)
 {
-	igc_clean_rx_ring_page_shared(ring);
+	if (ring->xsk_pool)
+		igc_clean_rx_ring_xsk_pool(ring);
+	else
+		igc_clean_rx_ring_page_shared(ring);
 
 	clear_ring_uses_large_buffer(ring);
 
@@ -533,6 +551,16 @@ static int igc_setup_all_rx_resources(struct igc_adapter *adapter)
 	return err;
 }
 
+static struct xsk_buff_pool *igc_get_xsk_pool(struct igc_adapter *adapter,
+					      struct igc_ring *ring)
+{
+	if (!igc_xdp_is_enabled(adapter) ||
+	    !test_bit(IGC_RING_FLAG_AF_XDP_ZC, &ring->flags))
+		return NULL;
+
+	return xsk_get_pool_from_qid(ring->netdev, ring->queue_index);
+}
+
 /**
  * igc_configure_rx_ring - Configure a receive ring after Reset
  * @adapter: board private structure
@@ -548,9 +576,20 @@ static void igc_configure_rx_ring(struct igc_adapter *adapter,
 	int reg_idx = ring->reg_idx;
 	u32 srrctl = 0, rxdctl = 0;
 	u64 rdba = ring->dma;
-
-	WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
-					   MEM_TYPE_PAGE_SHARED, NULL));
+	u32 buf_size;
+
+	xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
+	ring->xsk_pool = igc_get_xsk_pool(adapter, ring);
+	if (ring->xsk_pool) {
+		WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
+						   MEM_TYPE_XSK_BUFF_POOL,
+						   NULL));
+		xsk_pool_set_rxq_info(ring->xsk_pool, &ring->xdp_rxq);
+	} else {
+		WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
+						   MEM_TYPE_PAGE_SHARED,
+						   NULL));
+	}
 
 	if (igc_xdp_is_enabled(adapter))
 		set_ring_uses_large_buffer(ring);
@@ -574,12 +613,15 @@ static void igc_configure_rx_ring(struct igc_adapter *adapter,
 	ring->next_to_clean = 0;
 	ring->next_to_use = 0;
 
-	/* set descriptor configuration */
-	srrctl = IGC_RX_HDR_LEN << IGC_SRRCTL_BSIZEHDRSIZE_SHIFT;
-	if (ring_uses_large_buffer(ring))
-		srrctl |= IGC_RXBUFFER_3072 >> IGC_SRRCTL_BSIZEPKT_SHIFT;
+	if (ring->xsk_pool)
+		buf_size = xsk_pool_get_rx_frame_size(ring->xsk_pool);
+	else if (ring_uses_large_buffer(ring))
+		buf_size = IGC_RXBUFFER_3072;
 	else
-		srrctl |= IGC_RXBUFFER_2048 >> IGC_SRRCTL_BSIZEPKT_SHIFT;
+		buf_size = IGC_RXBUFFER_2048;
+
+	srrctl = IGC_RX_HDR_LEN << IGC_SRRCTL_BSIZEHDRSIZE_SHIFT;
+	srrctl |= buf_size >> IGC_SRRCTL_BSIZEPKT_SHIFT;
 	srrctl |= IGC_SRRCTL_DESCTYPE_ADV_ONEBUF;
 
 	wr32(IGC_SRRCTL(reg_idx), srrctl);
@@ -1939,6 +1981,63 @@ static void igc_alloc_rx_buffers(struct igc_ring *rx_ring, u16 cleaned_count)
 	}
 }
 
+static bool igc_alloc_rx_buffers_zc(struct igc_ring *ring, u16 count)
+{
+	union igc_adv_rx_desc *desc;
+	u16 i = ring->next_to_use;
+	struct igc_rx_buffer *bi;
+	dma_addr_t dma;
+	bool ok = true;
+
+	if (!count)
+		return ok;
+
+	desc = IGC_RX_DESC(ring, i);
+	bi = &ring->rx_buffer_info[i];
+	i -= ring->count;
+
+	do {
+		bi->xdp = xsk_buff_alloc(ring->xsk_pool);
+		if (!bi->xdp) {
+			ok = false;
+			break;
+		}
+
+		dma = xsk_buff_xdp_get_dma(bi->xdp);
+		desc->read.pkt_addr = cpu_to_le64(dma);
+
+		desc++;
+		bi++;
+		i++;
+		if (unlikely(!i)) {
+			desc = IGC_RX_DESC(ring, 0);
+			bi = ring->rx_buffer_info;
+			i -= ring->count;
+		}
+
+		/* Clear the length for the next_to_use descriptor. */
+		desc->wb.upper.length = 0;
+
+		count--;
+	} while (count);
+
+	i += ring->count;
+
+	if (ring->next_to_use != i) {
+		ring->next_to_use = i;
+
+		/* Force memory writes to complete before letting h/w
+		 * know there are new descriptors to fetch.  (Only
+		 * applicable for weak-ordered memory model archs,
+		 * such as IA-64).
+		 */
+		wmb();
+		writel(i, ring->tail);
+	}
+
+	return ok;
+}
+
 static int igc_xdp_init_tx_buffer(struct igc_tx_buffer *buffer,
 				  struct xdp_frame *xdpf,
 				  struct igc_ring *ring)
@@ -2254,6 +2353,148 @@ static int igc_clean_rx_irq(struct igc_q_vector *q_vector, const int budget)
 	return total_packets;
 }
 
+static struct sk_buff *igc_construct_skb_zc(struct igc_ring *ring,
+					    struct xdp_buff *xdp)
+{
+	unsigned int metasize = xdp->data - xdp->data_meta;
+	unsigned int datasize = xdp->data_end - xdp->data;
+	unsigned int totalsize = metasize + datasize;
+	struct sk_buff *skb;
+
+	skb = __napi_alloc_skb(&ring->q_vector->napi,
+			       xdp->data_end - xdp->data_hard_start,
+			       GFP_ATOMIC | __GFP_NOWARN);
+	if (unlikely(!skb))
+		return NULL;
+
+	skb_reserve(skb, xdp->data_meta - xdp->data_hard_start);
+	memcpy(__skb_put(skb, totalsize), xdp->data_meta, totalsize);
+	if (metasize)
+		skb_metadata_set(skb, metasize);
+
+	return skb;
+}
+
+static void igc_dispatch_skb_zc(struct igc_q_vector *q_vector,
+				union igc_adv_rx_desc *desc,
+				struct xdp_buff *xdp,
+				ktime_t timestamp)
+{
+	struct igc_ring *ring = q_vector->rx.ring;
+	struct sk_buff *skb;
+
+	skb = igc_construct_skb_zc(ring, xdp);
+	if (!skb) {
+		ring->rx_stats.alloc_failed++;
+		return;
+	}
+
+	if (timestamp)
+		skb_hwtstamps(skb)->hwtstamp = timestamp;
+
+	if (igc_cleanup_headers(ring, desc, skb))
+		return;
+
+	igc_process_skb_fields(ring, desc, skb);
+	napi_gro_receive(&q_vector->napi, skb);
+}
+
+static int igc_clean_rx_irq_zc(struct igc_q_vector *q_vector, const int budget)
+{
+	struct igc_adapter *adapter = q_vector->adapter;
+	struct igc_ring *ring = q_vector->rx.ring;
+	u16 cleaned_count = igc_desc_unused(ring);
+	int total_bytes = 0, total_packets = 0;
+	u16 ntc = ring->next_to_clean;
+	struct bpf_prog *prog;
+	bool failure = false;
+	int xdp_status = 0;
+
+	rcu_read_lock();
+
+	prog = READ_ONCE(adapter->xdp_prog);
+
+	while (likely(total_packets < budget)) {
+		union igc_adv_rx_desc *desc;
+		struct igc_rx_buffer *bi;
+		ktime_t timestamp = 0;
+		unsigned int size;
+		int res;
+
+		desc = IGC_RX_DESC(ring, ntc);
+		size = le16_to_cpu(desc->wb.upper.length);
+		if (!size)
+			break;
+
+		/* This memory barrier is needed to keep us from reading
+		 * any other fields out of the rx_desc until we know the
+		 * descriptor has been written back
+		 */
+		dma_rmb();
+
+		bi = &ring->rx_buffer_info[ntc];
+
+		if (igc_test_staterr(desc, IGC_RXDADV_STAT_TSIP)) {
+			timestamp = igc_ptp_rx_pktstamp(q_vector->adapter,
+							bi->xdp->data);
+
+			bi->xdp->data += IGC_TS_HDR_LEN;
+
+			/* HW timestamp has been copied into local variable. Metadata
+			 * length when XDP program is called should be 0.
+			 */
+			bi->xdp->data_meta += IGC_TS_HDR_LEN;
+			size -= IGC_TS_HDR_LEN;
+		}
+
+		bi->xdp->data_end = bi->xdp->data + size;
+		xsk_buff_dma_sync_for_cpu(bi->xdp, ring->xsk_pool);
+
+		res = __igc_xdp_run_prog(adapter, prog, bi->xdp);
+		switch (res) {
+		case IGC_XDP_PASS:
+			igc_dispatch_skb_zc(q_vector, desc, bi->xdp, timestamp);
+			fallthrough;
+		case IGC_XDP_CONSUMED:
+			xsk_buff_free(bi->xdp);
+			break;
+		case IGC_XDP_TX:
+		case IGC_XDP_REDIRECT:
+			xdp_status |= res;
+			break;
+		}
+
+		bi->xdp = NULL;
+		total_bytes += size;
+		total_packets++;
+		cleaned_count++;
+		ntc++;
+		if (ntc == ring->count)
+			ntc = 0;
+	}
+
+	ring->next_to_clean = ntc;
+	rcu_read_unlock();
+
+	if (cleaned_count >= IGC_RX_BUFFER_WRITE)
+		failure = !igc_alloc_rx_buffers_zc(ring, cleaned_count);
+
+	if (xdp_status)
+		igc_finalize_xdp(adapter, xdp_status);
+
+	igc_update_rx_stats(q_vector, total_packets, total_bytes);
+
+	if (xsk_uses_need_wakeup(ring->xsk_pool)) {
+		if (failure || ring->next_to_clean == ring->next_to_use)
+			xsk_set_rx_need_wakeup(ring->xsk_pool);
+		else
+			xsk_clear_rx_need_wakeup(ring->xsk_pool);
+		return total_packets;
+	}
+
+	return failure ? budget : total_packets;
+}
+
 static void igc_update_tx_stats(struct igc_q_vector *q_vector,
 				unsigned int packets, unsigned int bytes)
 {
@@ -2946,7 +3187,10 @@ static void igc_configure(struct igc_adapter *adapter)
 	for (i = 0; i < adapter->num_rx_queues; i++) {
 		struct igc_ring *ring = adapter->rx_ring[i];
 
-		igc_alloc_rx_buffers(ring, igc_desc_unused(ring));
+		if (ring->xsk_pool)
+			igc_alloc_rx_buffers_zc(ring, igc_desc_unused(ring));
+		else
+			igc_alloc_rx_buffers(ring, igc_desc_unused(ring));
 	}
 }
 
@@ -3561,14 +3805,17 @@ static int igc_poll(struct napi_struct *napi, int budget)
 	struct igc_q_vector *q_vector = container_of(napi,
 						     struct igc_q_vector,
 						     napi);
+	struct igc_ring *rx_ring = q_vector->rx.ring;
 	bool clean_complete = true;
 	int work_done = 0;
 
 	if (q_vector->tx.ring)
 		clean_complete = igc_clean_tx_irq(q_vector, budget);
 
-	if (q_vector->rx.ring) {
-		int cleaned = igc_clean_rx_irq(q_vector, budget);
+	if (rx_ring) {
+		int cleaned = rx_ring->xsk_pool ?
+			      igc_clean_rx_irq_zc(q_vector, budget) :
+			      igc_clean_rx_irq(q_vector, budget);
 
 		work_done += cleaned;
 		if (cleaned >= budget)
@@ -5206,6 +5453,9 @@ static int igc_bpf(struct net_device *dev, struct netdev_bpf *bpf)
 	switch (bpf->command) {
 	case XDP_SETUP_PROG:
 		return igc_xdp_set_prog(adapter, bpf->prog, bpf->extack);
+	case XDP_SETUP_XSK_POOL:
+		return igc_xdp_setup_pool(adapter, bpf->xsk.pool,
+					  bpf->xsk.queue_id);
 	default:
 		return -EOPNOTSUPP;
 	}
@@ -5251,6 +5501,43 @@ static int igc_xdp_xmit(struct net_device *dev, int num_frames,
 	return num_frames - drops;
 }
 
+static void igc_trigger_rxtxq_interrupt(struct igc_adapter *adapter,
+					struct igc_q_vector *q_vector)
+{
+	struct igc_hw *hw = &adapter->hw;
+	u32 eics = 0;
+
+	eics |= q_vector->eims_value;
+	wr32(IGC_EICS, eics);
+}
+
+int igc_xsk_wakeup(struct net_device *dev, u32 queue_id, u32 flags)
+{
+	struct igc_adapter *adapter = netdev_priv(dev);
+	struct igc_q_vector *q_vector;
+	struct igc_ring *ring;
+
+	if (test_bit(__IGC_DOWN, &adapter->state))
+		return -ENETDOWN;
+
+	if (!igc_xdp_is_enabled(adapter))
+		return -ENXIO;
+
+	if (queue_id >= adapter->num_rx_queues)
+		return -EINVAL;
+
+	ring = adapter->rx_ring[queue_id];
+
+	if (!ring->xsk_pool)
+		return -ENXIO;
+
+	q_vector = adapter->q_vector[queue_id];
+	if (!napi_if_scheduled_mark_missed(&q_vector->napi))
+		igc_trigger_rxtxq_interrupt(adapter, q_vector);
+
+	return 0;
+}
+
 static const struct net_device_ops igc_netdev_ops = {
 	.ndo_open		= igc_open,
 	.ndo_stop		= igc_close,
@@ -5266,6 +5553,7 @@ static const struct net_device_ops igc_netdev_ops = {
 	.ndo_setup_tc		= igc_setup_tc,
 	.ndo_bpf		= igc_bpf,
 	.ndo_xdp_xmit		= igc_xdp_xmit,
+	.ndo_xsk_wakeup		= igc_xsk_wakeup,
 };
 
 /* PCIe configuration access */
@@ -6018,6 +6306,36 @@ struct net_device *igc_get_hw_dev(struct igc_hw *hw)
 	return adapter->netdev;
 }
 
+static void igc_disable_rx_ring_hw(struct igc_ring *ring)
+{
+	struct igc_hw *hw = &ring->q_vector->adapter->hw;
+	u8 idx = ring->reg_idx;
+	u32 rxdctl;
+
+	rxdctl = rd32(IGC_RXDCTL(idx));
+	rxdctl &= ~IGC_RXDCTL_QUEUE_ENABLE;
+	rxdctl |= IGC_RXDCTL_SWFLUSH;
+	wr32(IGC_RXDCTL(idx), rxdctl);
+}
+
+void igc_disable_rx_ring(struct igc_ring *ring)
+{
+	igc_disable_rx_ring_hw(ring);
+	igc_clean_rx_ring(ring);
+}
+
+void igc_enable_rx_ring(struct igc_ring *ring)
+{
+	struct igc_adapter *adapter = ring->q_vector->adapter;
+
+	igc_configure_rx_ring(adapter, ring);
+
+	if (ring->xsk_pool)
+		igc_alloc_rx_buffers_zc(ring, igc_desc_unused(ring));
+	else
+		igc_alloc_rx_buffers(ring, igc_desc_unused(ring));
+}
+
 /**
  * igc_init_module - Driver Registration Routine
  *
diff --git a/drivers/net/ethernet/intel/igc/igc_xdp.c b/drivers/net/ethernet/intel/igc/igc_xdp.c
index 27c886a254f1..c65d690b75bf 100644
--- a/drivers/net/ethernet/intel/igc/igc_xdp.c
+++ b/drivers/net/ethernet/intel/igc/igc_xdp.c
@@ -1,6 +1,8 @@
 // SPDX-License-Identifier: GPL-2.0
 /* Copyright (c) 2020, Intel Corporation. */
 
+#include <net/xdp_sock_drv.h>
+
 #include "igc.h"
 #include "igc_xdp.h"
 
@@ -31,3 +33,101 @@ int igc_xdp_set_prog(struct igc_adapter *adapter, struct bpf_prog *prog,
 
 	return 0;
 }
+
+static int igc_xdp_enable_pool(struct igc_adapter *adapter,
+			       struct xsk_buff_pool *pool, u16 queue_id)
+{
+	struct net_device *ndev = adapter->netdev;
+	struct device *dev = &adapter->pdev->dev;
+	struct igc_ring *rx_ring;
+	struct napi_struct *napi;
+	bool needs_reset;
+	u32 frame_size;
+	int err;
+
+	if (queue_id >= adapter->num_rx_queues)
+		return -EINVAL;
+
+	frame_size = xsk_pool_get_rx_frame_size(pool);
+	if (frame_size < ETH_FRAME_LEN + VLAN_HLEN * 2) {
+		/* When XDP is enabled, the driver doesn't support frames that
+		 * span over multiple buffers. To avoid that, we check if xsk
+		 * frame size is big enough to fit the max ethernet frame size
+		 * + vlan double tagging.
+		 */
+		return -EOPNOTSUPP;
+	}
+
+	err = xsk_pool_dma_map(pool, dev, IGC_RX_DMA_ATTR);
+	if (err) {
+		netdev_err(ndev, "Failed to map xsk pool\n");
+		return err;
+	}
+
+	needs_reset = netif_running(adapter->netdev) && igc_xdp_is_enabled(adapter);
+
+	rx_ring = adapter->rx_ring[queue_id];
+	napi = &rx_ring->q_vector->napi;
+
+	if (needs_reset) {
+		igc_disable_rx_ring(rx_ring);
+		napi_disable(napi);
+	}
+
+	set_bit(IGC_RING_FLAG_AF_XDP_ZC, &rx_ring->flags);
+
+	if (needs_reset) {
+		napi_enable(napi);
+		igc_enable_rx_ring(rx_ring);
+
+		err = igc_xsk_wakeup(ndev, queue_id, XDP_WAKEUP_RX);
+		if (err) {
+			xsk_pool_dma_unmap(pool, IGC_RX_DMA_ATTR);
+			return err;
+		}
+	}
+
+	return 0;
+}
+
+static int igc_xdp_disable_pool(struct igc_adapter *adapter, u16 queue_id)
+{
+	struct xsk_buff_pool *pool;
+	struct igc_ring *rx_ring;
+	struct napi_struct *napi;
+	bool needs_reset;
+
+	if (queue_id >= adapter->num_rx_queues)
+		return -EINVAL;
+
+	pool = xsk_get_pool_from_qid(adapter->netdev, queue_id);
+	if (!pool)
+		return -EINVAL;
+
+	needs_reset = netif_running(adapter->netdev) && igc_xdp_is_enabled(adapter);
+
+	rx_ring = adapter->rx_ring[queue_id];
+	napi = &rx_ring->q_vector->napi;
+
+	if (needs_reset) {
+		igc_disable_rx_ring(rx_ring);
+		napi_disable(napi);
+	}
+
+	xsk_pool_dma_unmap(pool, IGC_RX_DMA_ATTR);
+	clear_bit(IGC_RING_FLAG_AF_XDP_ZC, &rx_ring->flags);
+
+	if (needs_reset) {
+		napi_enable(napi);
+		igc_enable_rx_ring(rx_ring);
+	}
+
+	return 0;
+}
+
+int igc_xdp_setup_pool(struct igc_adapter *adapter, struct xsk_buff_pool *pool,
+		       u16 queue_id)
+{
+	return pool ? igc_xdp_enable_pool(adapter, pool, queue_id) :
+		      igc_xdp_disable_pool(adapter, queue_id);
+}
diff --git a/drivers/net/ethernet/intel/igc/igc_xdp.h b/drivers/net/ethernet/intel/igc/igc_xdp.h
index cdaa2c39b03a..a74e5487d199 100644
--- a/drivers/net/ethernet/intel/igc/igc_xdp.h
+++ b/drivers/net/ethernet/intel/igc/igc_xdp.h
@@ -6,6 +6,8 @@
 
 int igc_xdp_set_prog(struct igc_adapter *adapter, struct bpf_prog *prog,
 		     struct netlink_ext_ack *extack);
+int igc_xdp_setup_pool(struct igc_adapter *adapter, struct xsk_buff_pool *pool,
+		       u16 queue_id);
 
 static inline bool igc_xdp_is_enabled(struct igc_adapter *adapter)
 {
-- 
2.26.2

[PATCH net-next v2 9/9] igc: Enable TX via AF_XDP zero-copy

[Tony Nguyen]

From: Andre Guedes <andre.guedes@intel.com>

Add support for transmitting packets via AF_XDP zero-copy mechanism.

The packet transmission itself is implemented by igc_xdp_xmit_zc() which
is called from igc_clean_tx_irq() when the ring has AF_XDP zero-copy
enabled. Likewise i40e and ice drivers, the transmission budget used is
the number of descriptors available on the ring.

A new tx buffer type is introduced to 'enum igc_tx_buffer_type' to
indicate the tx buffer uses memory from xsk pool so it can be properly
cleaned after transmission or when the ring is cleaned.

The I225 controller has only 4 Tx hardware queues so the main difference
between igc and other Intel drivers that support AF_XDP zero-copy is
that there is no tx ring dedicated exclusively to XDP. Instead, tx
rings are shared between the network stack and XDP, and netdev queue
lock is used to ensure mutual exclusion. This is the same approach
implemented to support XDP_TX and XDP_REDIRECT actions.

Signed-off-by: Andre Guedes <andre.guedes@intel.com>
Signed-off-by: Vedang Patel <vedang.patel@intel.com>
Signed-off-by: Jithu Joseph <jithu.joseph@intel.com>
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Dvora Fuxbrumer <dvorax.fuxbrumer@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
---
 drivers/net/ethernet/intel/igc/igc.h      |   3 +
 drivers/net/ethernet/intel/igc/igc_base.h |   1 +
 drivers/net/ethernet/intel/igc/igc_main.c | 113 +++++++++++++++++++++-
 drivers/net/ethernet/intel/igc/igc_xdp.c  |  20 +++-
 4 files changed, 129 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/intel/igc/igc.h b/drivers/net/ethernet/intel/igc/igc.h
index cd6f4c94c4dd..b6d3277c6f52 100644
--- a/drivers/net/ethernet/intel/igc/igc.h
+++ b/drivers/net/ethernet/intel/igc/igc.h
@@ -258,6 +258,8 @@ int igc_set_spd_dplx(struct igc_adapter *adapter, u32 spd, u8 dplx);
 void igc_update_stats(struct igc_adapter *adapter);
 void igc_disable_rx_ring(struct igc_ring *ring);
 void igc_enable_rx_ring(struct igc_ring *ring);
+void igc_disable_tx_ring(struct igc_ring *ring);
+void igc_enable_tx_ring(struct igc_ring *ring);
 int igc_xsk_wakeup(struct net_device *dev, u32 queue_id, u32 flags);
 
 /* igc_dump declarations */
@@ -413,6 +415,7 @@ enum igc_boards {
 enum igc_tx_buffer_type {
 	IGC_TX_BUFFER_TYPE_SKB,
 	IGC_TX_BUFFER_TYPE_XDP,
+	IGC_TX_BUFFER_TYPE_XSK,
 };
 
 /* wrapper around a pointer to a socket buffer,
diff --git a/drivers/net/ethernet/intel/igc/igc_base.h b/drivers/net/ethernet/intel/igc/igc_base.h
index 2ca028c1919f..ce530f5fd7bd 100644
--- a/drivers/net/ethernet/intel/igc/igc_base.h
+++ b/drivers/net/ethernet/intel/igc/igc_base.h
@@ -78,6 +78,7 @@ union igc_adv_rx_desc {
 
 /* Additional Transmit Descriptor Control definitions */
 #define IGC_TXDCTL_QUEUE_ENABLE	0x02000000 /* Ena specific Tx Queue */
+#define IGC_TXDCTL_SWFLUSH	0x04000000 /* Transmit Software Flush */
 
 /* Additional Receive Descriptor Control definitions */
 #define IGC_RXDCTL_QUEUE_ENABLE	0x02000000 /* Ena specific Rx Queue */
diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethernet/intel/igc/igc_main.c
index 3ffc20fae4c6..ea998d2defa4 100644
--- a/drivers/net/ethernet/intel/igc/igc_main.c
+++ b/drivers/net/ethernet/intel/igc/igc_main.c
@@ -187,24 +187,28 @@ static void igc_clean_tx_ring(struct igc_ring *tx_ring)
 {
 	u16 i = tx_ring->next_to_clean;
 	struct igc_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
+	u32 xsk_frames = 0;
 
 	while (i != tx_ring->next_to_use) {
 		union igc_adv_tx_desc *eop_desc, *tx_desc;
 
 		switch (tx_buffer->type) {
+		case IGC_TX_BUFFER_TYPE_XSK:
+			xsk_frames++;
+			break;
 		case IGC_TX_BUFFER_TYPE_XDP:
 			xdp_return_frame(tx_buffer->xdpf);
+			igc_unmap_tx_buffer(tx_ring->dev, tx_buffer);
 			break;
 		case IGC_TX_BUFFER_TYPE_SKB:
 			dev_kfree_skb_any(tx_buffer->skb);
+			igc_unmap_tx_buffer(tx_ring->dev, tx_buffer);
 			break;
 		default:
 			netdev_warn_once(tx_ring->netdev, "Unknown Tx buffer type\n");
 			break;
 		}
 
-		igc_unmap_tx_buffer(tx_ring->dev, tx_buffer);
-
 		/* check for eop_desc to determine the end of the packet */
 		eop_desc = tx_buffer->next_to_watch;
 		tx_desc = IGC_TX_DESC(tx_ring, i);
@@ -234,6 +238,9 @@ static void igc_clean_tx_ring(struct igc_ring *tx_ring)
 		}
 	}
 
+	if (tx_ring->xsk_pool && xsk_frames)
+		xsk_tx_completed(tx_ring->xsk_pool, xsk_frames);
+
 	/* reset BQL for queue */
 	netdev_tx_reset_queue(txring_txq(tx_ring));
 
@@ -676,6 +683,8 @@ static void igc_configure_tx_ring(struct igc_adapter *adapter,
 	u64 tdba = ring->dma;
 	u32 txdctl = 0;
 
+	ring->xsk_pool = igc_get_xsk_pool(adapter, ring);
+
 	/* disable the queue */
 	wr32(IGC_TXDCTL(reg_idx), 0);
 	wrfl();
@@ -2509,6 +2518,65 @@ static void igc_update_tx_stats(struct igc_q_vector *q_vector,
 	q_vector->tx.total_packets += packets;
 }
 
+static void igc_xdp_xmit_zc(struct igc_ring *ring)
+{
+	struct xsk_buff_pool *pool = ring->xsk_pool;
+	struct netdev_queue *nq = txring_txq(ring);
+	union igc_adv_tx_desc *tx_desc = NULL;
+	int cpu = smp_processor_id();
+	u16 ntu = ring->next_to_use;
+	struct xdp_desc xdp_desc;
+	u16 budget;
+
+	if (!netif_carrier_ok(ring->netdev))
+		return;
+
+	__netif_tx_lock(nq, cpu);
+
+	budget = igc_desc_unused(ring);
+
+	while (xsk_tx_peek_desc(pool, &xdp_desc) && budget--) {
+		u32 cmd_type, olinfo_status;
+		struct igc_tx_buffer *bi;
+		dma_addr_t dma;
+
+		cmd_type = IGC_ADVTXD_DTYP_DATA | IGC_ADVTXD_DCMD_DEXT |
+			   IGC_ADVTXD_DCMD_IFCS | IGC_TXD_DCMD |
+			   xdp_desc.len;
+		olinfo_status = xdp_desc.len << IGC_ADVTXD_PAYLEN_SHIFT;
+
+		dma = xsk_buff_raw_get_dma(pool, xdp_desc.addr);
+		xsk_buff_raw_dma_sync_for_device(pool, dma, xdp_desc.len);
+
+		tx_desc = IGC_TX_DESC(ring, ntu);
+		tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
+		tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
+		tx_desc->read.buffer_addr = cpu_to_le64(dma);
+
+		bi = &ring->tx_buffer_info[ntu];
+		bi->type = IGC_TX_BUFFER_TYPE_XSK;
+		bi->protocol = 0;
+		bi->bytecount = xdp_desc.len;
+		bi->gso_segs = 1;
+		bi->time_stamp = jiffies;
+		bi->next_to_watch = tx_desc;
+
+		netdev_tx_sent_queue(txring_txq(ring), xdp_desc.len);
+
+		ntu++;
+		if (ntu == ring->count)
+			ntu = 0;
+	}
+
+	ring->next_to_use = ntu;
+	if (tx_desc) {
+		igc_flush_tx_descriptors(ring);
+		xsk_tx_release(pool);
+	}
+
+	__netif_tx_unlock(nq);
+}
+
 /**
  * igc_clean_tx_irq - Reclaim resources after transmit completes
  * @q_vector: pointer to q_vector containing needed info
@@ -2525,6 +2593,7 @@ static bool igc_clean_tx_irq(struct igc_q_vector *q_vector, int napi_budget)
 	unsigned int i = tx_ring->next_to_clean;
 	struct igc_tx_buffer *tx_buffer;
 	union igc_adv_tx_desc *tx_desc;
+	u32 xsk_frames = 0;
 
 	if (test_bit(__IGC_DOWN, &adapter->state))
 		return true;
@@ -2555,19 +2624,22 @@ static bool igc_clean_tx_irq(struct igc_q_vector *q_vector, int napi_budget)
 		total_packets += tx_buffer->gso_segs;
 
 		switch (tx_buffer->type) {
+		case IGC_TX_BUFFER_TYPE_XSK:
+			xsk_frames++;
+			break;
 		case IGC_TX_BUFFER_TYPE_XDP:
 			xdp_return_frame(tx_buffer->xdpf);
+			igc_unmap_tx_buffer(tx_ring->dev, tx_buffer);
 			break;
 		case IGC_TX_BUFFER_TYPE_SKB:
 			napi_consume_skb(tx_buffer->skb, napi_budget);
+			igc_unmap_tx_buffer(tx_ring->dev, tx_buffer);
 			break;
 		default:
 			netdev_warn_once(tx_ring->netdev, "Unknown Tx buffer type\n");
 			break;
 		}
 
-		igc_unmap_tx_buffer(tx_ring->dev, tx_buffer);
-
 		/* clear last DMA location and unmap remaining buffers */
 		while (tx_desc != eop_desc) {
 			tx_buffer++;
@@ -2609,6 +2681,14 @@ static bool igc_clean_tx_irq(struct igc_q_vector *q_vector, int napi_budget)
 
 	igc_update_tx_stats(q_vector, total_packets, total_bytes);
 
+	if (tx_ring->xsk_pool) {
+		if (xsk_frames)
+			xsk_tx_completed(tx_ring->xsk_pool, xsk_frames);
+		if (xsk_uses_need_wakeup(tx_ring->xsk_pool))
+			xsk_set_tx_need_wakeup(tx_ring->xsk_pool);
+		igc_xdp_xmit_zc(tx_ring);
+	}
+
 	if (test_bit(IGC_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags)) {
 		struct igc_hw *hw = &adapter->hw;
 
@@ -6336,6 +6416,31 @@ void igc_enable_rx_ring(struct igc_ring *ring)
 		igc_alloc_rx_buffers(ring, igc_desc_unused(ring));
 }
 
+static void igc_disable_tx_ring_hw(struct igc_ring *ring)
+{
+	struct igc_hw *hw = &ring->q_vector->adapter->hw;
+	u8 idx = ring->reg_idx;
+	u32 txdctl;
+
+	txdctl = rd32(IGC_TXDCTL(idx));
+	txdctl &= ~IGC_TXDCTL_QUEUE_ENABLE;
+	txdctl |= IGC_TXDCTL_SWFLUSH;
+	wr32(IGC_TXDCTL(idx), txdctl);
+}
+
+void igc_disable_tx_ring(struct igc_ring *ring)
+{
+	igc_disable_tx_ring_hw(ring);
+	igc_clean_tx_ring(ring);
+}
+
+void igc_enable_tx_ring(struct igc_ring *ring)
+{
+	struct igc_adapter *adapter = ring->q_vector->adapter;
+
+	igc_configure_tx_ring(adapter, ring);
+}
+
 /**
  * igc_init_module - Driver Registration Routine
  *
diff --git a/drivers/net/ethernet/intel/igc/igc_xdp.c b/drivers/net/ethernet/intel/igc/igc_xdp.c
index c65d690b75bf..a8cf5374be47 100644
--- a/drivers/net/ethernet/intel/igc/igc_xdp.c
+++ b/drivers/net/ethernet/intel/igc/igc_xdp.c
@@ -39,13 +39,14 @@ static int igc_xdp_enable_pool(struct igc_adapter *adapter,
 {
 	struct net_device *ndev = adapter->netdev;
 	struct device *dev = &adapter->pdev->dev;
-	struct igc_ring *rx_ring;
+	struct igc_ring *rx_ring, *tx_ring;
 	struct napi_struct *napi;
 	bool needs_reset;
 	u32 frame_size;
 	int err;
 
-	if (queue_id >= adapter->num_rx_queues)
+	if (queue_id >= adapter->num_rx_queues ||
+	    queue_id >= adapter->num_tx_queues)
 		return -EINVAL;
 
 	frame_size = xsk_pool_get_rx_frame_size(pool);
@@ -67,18 +68,23 @@ static int igc_xdp_enable_pool(struct igc_adapter *adapter,
 	needs_reset = netif_running(adapter->netdev) && igc_xdp_is_enabled(adapter);
 
 	rx_ring = adapter->rx_ring[queue_id];
+	tx_ring = adapter->tx_ring[queue_id];
+	/* Rx and Tx rings share the same napi context. */
 	napi = &rx_ring->q_vector->napi;
 
 	if (needs_reset) {
 		igc_disable_rx_ring(rx_ring);
+		igc_disable_tx_ring(tx_ring);
 		napi_disable(napi);
 	}
 
 	set_bit(IGC_RING_FLAG_AF_XDP_ZC, &rx_ring->flags);
+	set_bit(IGC_RING_FLAG_AF_XDP_ZC, &tx_ring->flags);
 
 	if (needs_reset) {
 		napi_enable(napi);
 		igc_enable_rx_ring(rx_ring);
+		igc_enable_tx_ring(tx_ring);
 
 		err = igc_xsk_wakeup(ndev, queue_id, XDP_WAKEUP_RX);
 		if (err) {
@@ -92,12 +98,13 @@ static int igc_xdp_enable_pool(struct igc_adapter *adapter,
 
 static int igc_xdp_disable_pool(struct igc_adapter *adapter, u16 queue_id)
 {
+	struct igc_ring *rx_ring, *tx_ring;
 	struct xsk_buff_pool *pool;
-	struct igc_ring *rx_ring;
 	struct napi_struct *napi;
 	bool needs_reset;
 
-	if (queue_id >= adapter->num_rx_queues)
+	if (queue_id >= adapter->num_rx_queues ||
+	    queue_id >= adapter->num_tx_queues)
 		return -EINVAL;
 
 	pool = xsk_get_pool_from_qid(adapter->netdev, queue_id);
@@ -107,19 +114,24 @@ static int igc_xdp_disable_pool(struct igc_adapter *adapter, u16 queue_id)
 	needs_reset = netif_running(adapter->netdev) && igc_xdp_is_enabled(adapter);
 
 	rx_ring = adapter->rx_ring[queue_id];
+	tx_ring = adapter->tx_ring[queue_id];
+	/* Rx and Tx rings share the same napi context. */
 	napi = &rx_ring->q_vector->napi;
 
 	if (needs_reset) {
 		igc_disable_rx_ring(rx_ring);
+		igc_disable_tx_ring(tx_ring);
 		napi_disable(napi);
 	}
 
 	xsk_pool_dma_unmap(pool, IGC_RX_DMA_ATTR);
 	clear_bit(IGC_RING_FLAG_AF_XDP_ZC, &rx_ring->flags);
+	clear_bit(IGC_RING_FLAG_AF_XDP_ZC, &tx_ring->flags);
 
 	if (needs_reset) {
 		napi_enable(napi);
 		igc_enable_rx_ring(rx_ring);
+		igc_enable_tx_ring(tx_ring);
 	}
 
 	return 0;
-- 
2.26.2

Re: [PATCH net-next v2 0/9][pull request] 1GbE Intel Wired LAN Driver Updates 2021-05-20

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net-next.git (refs/heads/master):

On Thu, 20 May 2021 11:17:35 -0700 you wrote:
> This series contains updates to igc driver only.
> 
> Andre Guedes says:
> 
> This series adds AF_XDP zero-copy feature to igc driver.
> 
> The initial patches do some code refactoring, preparing the code base to
> land the AF_XDP zero-copy feature, avoiding code duplications. The last
> patches of the series are the ones implementing the feature.
> 
> [...]

Here is the summary with links:
  - [net-next,v2,1/9] igc: Move igc_xdp_is_enabled()
    https://git.kernel.org/netdev/net-next/c/0c20f2d29fff
  - [net-next,v2,2/9] igc: Refactor __igc_xdp_run_prog()
    https://git.kernel.org/netdev/net-next/c/73a6e3721261
  - [net-next,v2,3/9] igc: Refactor igc_clean_rx_ring()
    https://git.kernel.org/netdev/net-next/c/f485164867d3
  - [net-next,v2,4/9] igc: Refactor XDP rxq info registration
    https://git.kernel.org/netdev/net-next/c/4609ffb9f615
  - [net-next,v2,5/9] igc: Introduce TX/RX stats helpers
    https://git.kernel.org/netdev/net-next/c/a27e6e73e550
  - [net-next,v2,6/9] igc: Introduce igc_unmap_tx_buffer() helper
    https://git.kernel.org/netdev/net-next/c/6123429516c7
  - [net-next,v2,7/9] igc: Replace IGC_TX_FLAGS_XDP flag by an enum
    https://git.kernel.org/netdev/net-next/c/859b4dfa4115
  - [net-next,v2,8/9] igc: Enable RX via AF_XDP zero-copy
    https://git.kernel.org/netdev/net-next/c/fc9df2a0b520
  - [net-next,v2,9/9] igc: Enable TX via AF_XDP zero-copy
    https://git.kernel.org/netdev/net-next/c/9acf59a752d4

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html