diff mbox series

[net-next,v9,4/4] gve: Add support for raw addressing in the tx path

Message ID 20201202182413.232510-5-awogbemila@google.com
State Superseded
Headers show
Series GVE Raw Addressing | expand

Commit Message

David Awogbemila Dec. 2, 2020, 6:24 p.m. UTC
From: Catherine Sullivan <csully@google.com>

During TX, skbs' data addresses are dma_map'ed and passed to the NIC.
This means that the device can perform DMA directly from these addresses
and the driver does not have to copy the buffer content into
pre-allocated buffers/qpls (as in qpl mode).

Reviewed-by: Yangchun Fu <yangchun@google.com>
Signed-off-by: Catherine Sullivan <csully@google.com>
Signed-off-by: David Awogbemila <awogbemila@google.com>
---
 drivers/net/ethernet/google/gve/gve.h         |  16 +-
 drivers/net/ethernet/google/gve/gve_adminq.c  |   4 +-
 drivers/net/ethernet/google/gve/gve_desc.h    |   8 +-
 drivers/net/ethernet/google/gve/gve_ethtool.c |   2 +
 drivers/net/ethernet/google/gve/gve_tx.c      | 197 ++++++++++++++----
 5 files changed, 185 insertions(+), 42 deletions(-)

Comments

Alexander Duyck Dec. 3, 2020, 6:16 p.m. UTC | #1
On Wed, Dec 2, 2020 at 10:24 AM David Awogbemila <awogbemila@google.com> wrote:
>

> From: Catherine Sullivan <csully@google.com>

>

> During TX, skbs' data addresses are dma_map'ed and passed to the NIC.

> This means that the device can perform DMA directly from these addresses

> and the driver does not have to copy the buffer content into

> pre-allocated buffers/qpls (as in qpl mode).

>

> Reviewed-by: Yangchun Fu <yangchun@google.com>

> Signed-off-by: Catherine Sullivan <csully@google.com>

> Signed-off-by: David Awogbemila <awogbemila@google.com>

> ---

>  drivers/net/ethernet/google/gve/gve.h         |  16 +-

>  drivers/net/ethernet/google/gve/gve_adminq.c  |   4 +-

>  drivers/net/ethernet/google/gve/gve_desc.h    |   8 +-

>  drivers/net/ethernet/google/gve/gve_ethtool.c |   2 +

>  drivers/net/ethernet/google/gve/gve_tx.c      | 197 ++++++++++++++----

>  5 files changed, 185 insertions(+), 42 deletions(-)

>

> diff --git a/drivers/net/ethernet/google/gve/gve.h b/drivers/net/ethernet/google/gve/gve.h

> index 8aad4af2aa2b..9888fa92be86 100644

> --- a/drivers/net/ethernet/google/gve/gve.h

> +++ b/drivers/net/ethernet/google/gve/gve.h

> @@ -112,12 +112,20 @@ struct gve_tx_iovec {

>         u32 iov_padding; /* padding associated with this segment */

>  };

>

> +struct gve_tx_dma_buf {

> +       DEFINE_DMA_UNMAP_ADDR(dma);

> +       DEFINE_DMA_UNMAP_LEN(len);

> +};

> +

>  /* Tracks the memory in the fifo occupied by the skb. Mapped 1:1 to a desc

>   * ring entry but only used for a pkt_desc not a seg_desc

>   */

>  struct gve_tx_buffer_state {

>         struct sk_buff *skb; /* skb for this pkt */

> -       struct gve_tx_iovec iov[GVE_TX_MAX_IOVEC]; /* segments of this pkt */

> +       union {

> +               struct gve_tx_iovec iov[GVE_TX_MAX_IOVEC]; /* segments of this pkt */

> +               struct gve_tx_dma_buf buf;

> +       };

>  };

>

>  /* A TX buffer - each queue has one */

> @@ -140,19 +148,23 @@ struct gve_tx_ring {

>         __be32 last_nic_done ____cacheline_aligned; /* NIC tail pointer */

>         u64 pkt_done; /* free-running - total packets completed */

>         u64 bytes_done; /* free-running - total bytes completed */

> +       u32 dropped_pkt; /* free-running - total packets dropped */


Generally I would probably use a u64 for any count values. I'm not
sure what rate you will be moving packets at but if something goes
wrong you are better off with the counter not rolling over every few
minutes.

>         /* Cacheline 2 -- Read-mostly fields */

>         union gve_tx_desc *desc ____cacheline_aligned;

>         struct gve_tx_buffer_state *info; /* Maps 1:1 to a desc */

>         struct netdev_queue *netdev_txq;

>         struct gve_queue_resources *q_resources; /* head and tail pointer idx */

> +       struct device *dev;

>         u32 mask; /* masks req and done down to queue size */

> +       u8 raw_addressing; /* use raw_addressing? */

>

>         /* Slow-path fields */

>         u32 q_num ____cacheline_aligned; /* queue idx */

>         u32 stop_queue; /* count of queue stops */

>         u32 wake_queue; /* count of queue wakes */

>         u32 ntfy_id; /* notification block index */

> +       u32 dma_mapping_error; /* count of dma mapping errors */


Since this is a counter wouldn't it make more sense to place it up
with the other counters?

Looking over the rest of the patch it seems fine to me. The counters
were the only thing that had me a bit concerned.
David Awogbemila Dec. 7, 2020, 6:15 p.m. UTC | #2
On Thu, Dec 3, 2020 at 10:16 AM Alexander Duyck
<alexander.duyck@gmail.com> wrote:
>

> On Wed, Dec 2, 2020 at 10:24 AM David Awogbemila <awogbemila@google.com> wrote:

> >

> > From: Catherine Sullivan <csully@google.com>

> >

> > During TX, skbs' data addresses are dma_map'ed and passed to the NIC.

> > This means that the device can perform DMA directly from these addresses

> > and the driver does not have to copy the buffer content into

> > pre-allocated buffers/qpls (as in qpl mode).

> >

> > Reviewed-by: Yangchun Fu <yangchun@google.com>

> > Signed-off-by: Catherine Sullivan <csully@google.com>

> > Signed-off-by: David Awogbemila <awogbemila@google.com>

> > ---

> >  drivers/net/ethernet/google/gve/gve.h         |  16 +-

> >  drivers/net/ethernet/google/gve/gve_adminq.c  |   4 +-

> >  drivers/net/ethernet/google/gve/gve_desc.h    |   8 +-

> >  drivers/net/ethernet/google/gve/gve_ethtool.c |   2 +

> >  drivers/net/ethernet/google/gve/gve_tx.c      | 197 ++++++++++++++----

> >  5 files changed, 185 insertions(+), 42 deletions(-)

> >

> > diff --git a/drivers/net/ethernet/google/gve/gve.h b/drivers/net/ethernet/google/gve/gve.h

> > index 8aad4af2aa2b..9888fa92be86 100644

> > --- a/drivers/net/ethernet/google/gve/gve.h

> > +++ b/drivers/net/ethernet/google/gve/gve.h

> > @@ -112,12 +112,20 @@ struct gve_tx_iovec {

> >         u32 iov_padding; /* padding associated with this segment */

> >  };

> >

> > +struct gve_tx_dma_buf {

> > +       DEFINE_DMA_UNMAP_ADDR(dma);

> > +       DEFINE_DMA_UNMAP_LEN(len);

> > +};

> > +

> >  /* Tracks the memory in the fifo occupied by the skb. Mapped 1:1 to a desc

> >   * ring entry but only used for a pkt_desc not a seg_desc

> >   */

> >  struct gve_tx_buffer_state {

> >         struct sk_buff *skb; /* skb for this pkt */

> > -       struct gve_tx_iovec iov[GVE_TX_MAX_IOVEC]; /* segments of this pkt */

> > +       union {

> > +               struct gve_tx_iovec iov[GVE_TX_MAX_IOVEC]; /* segments of this pkt */

> > +               struct gve_tx_dma_buf buf;

> > +       };

> >  };

> >

> >  /* A TX buffer - each queue has one */

> > @@ -140,19 +148,23 @@ struct gve_tx_ring {

> >         __be32 last_nic_done ____cacheline_aligned; /* NIC tail pointer */

> >         u64 pkt_done; /* free-running - total packets completed */

> >         u64 bytes_done; /* free-running - total bytes completed */

> > +       u32 dropped_pkt; /* free-running - total packets dropped */

>

> Generally I would probably use a u64 for any count values. I'm not

> sure what rate you will be moving packets at but if something goes

> wrong you are better off with the counter not rolling over every few

> minutes.


Ok, using a u64 here would be better - I'll do that, thanks.

>

> >         /* Cacheline 2 -- Read-mostly fields */

> >         union gve_tx_desc *desc ____cacheline_aligned;

> >         struct gve_tx_buffer_state *info; /* Maps 1:1 to a desc */

> >         struct netdev_queue *netdev_txq;

> >         struct gve_queue_resources *q_resources; /* head and tail pointer idx */

> > +       struct device *dev;

> >         u32 mask; /* masks req and done down to queue size */

> > +       u8 raw_addressing; /* use raw_addressing? */

> >

> >         /* Slow-path fields */

> >         u32 q_num ____cacheline_aligned; /* queue idx */

> >         u32 stop_queue; /* count of queue stops */

> >         u32 wake_queue; /* count of queue wakes */

> >         u32 ntfy_id; /* notification block index */

> > +       u32 dma_mapping_error; /* count of dma mapping errors */

>

> Since this is a counter wouldn't it make more sense to place it up

> with the other counters?

>

> Looking over the rest of the patch it seems fine to me. The counters

> were the only thing that had me a bit concerned.


Ok, I'll use a u64 here and place it with the counter above.
diff mbox series

Patch

diff --git a/drivers/net/ethernet/google/gve/gve.h b/drivers/net/ethernet/google/gve/gve.h
index 8aad4af2aa2b..9888fa92be86 100644
--- a/drivers/net/ethernet/google/gve/gve.h
+++ b/drivers/net/ethernet/google/gve/gve.h
@@ -112,12 +112,20 @@  struct gve_tx_iovec {
 	u32 iov_padding; /* padding associated with this segment */
 };
 
+struct gve_tx_dma_buf {
+	DEFINE_DMA_UNMAP_ADDR(dma);
+	DEFINE_DMA_UNMAP_LEN(len);
+};
+
 /* Tracks the memory in the fifo occupied by the skb. Mapped 1:1 to a desc
  * ring entry but only used for a pkt_desc not a seg_desc
  */
 struct gve_tx_buffer_state {
 	struct sk_buff *skb; /* skb for this pkt */
-	struct gve_tx_iovec iov[GVE_TX_MAX_IOVEC]; /* segments of this pkt */
+	union {
+		struct gve_tx_iovec iov[GVE_TX_MAX_IOVEC]; /* segments of this pkt */
+		struct gve_tx_dma_buf buf;
+	};
 };
 
 /* A TX buffer - each queue has one */
@@ -140,19 +148,23 @@  struct gve_tx_ring {
 	__be32 last_nic_done ____cacheline_aligned; /* NIC tail pointer */
 	u64 pkt_done; /* free-running - total packets completed */
 	u64 bytes_done; /* free-running - total bytes completed */
+	u32 dropped_pkt; /* free-running - total packets dropped */
 
 	/* Cacheline 2 -- Read-mostly fields */
 	union gve_tx_desc *desc ____cacheline_aligned;
 	struct gve_tx_buffer_state *info; /* Maps 1:1 to a desc */
 	struct netdev_queue *netdev_txq;
 	struct gve_queue_resources *q_resources; /* head and tail pointer idx */
+	struct device *dev;
 	u32 mask; /* masks req and done down to queue size */
+	u8 raw_addressing; /* use raw_addressing? */
 
 	/* Slow-path fields */
 	u32 q_num ____cacheline_aligned; /* queue idx */
 	u32 stop_queue; /* count of queue stops */
 	u32 wake_queue; /* count of queue wakes */
 	u32 ntfy_id; /* notification block index */
+	u32 dma_mapping_error; /* count of dma mapping errors */
 	dma_addr_t bus; /* dma address of the descr ring */
 	dma_addr_t q_resources_bus; /* dma address of the queue resources */
 	struct u64_stats_sync statss; /* sync stats for 32bit archs */
@@ -442,7 +454,7 @@  static inline u32 gve_rx_idx_to_ntfy(struct gve_priv *priv, u32 queue_idx)
  */
 static inline u32 gve_num_tx_qpls(struct gve_priv *priv)
 {
-	return priv->tx_cfg.num_queues;
+	return priv->raw_addressing ? 0 : priv->tx_cfg.num_queues;
 }
 
 /* Returns the number of rx queue page lists
diff --git a/drivers/net/ethernet/google/gve/gve_adminq.c b/drivers/net/ethernet/google/gve/gve_adminq.c
index a1b9370db1e4..53864f200599 100644
--- a/drivers/net/ethernet/google/gve/gve_adminq.c
+++ b/drivers/net/ethernet/google/gve/gve_adminq.c
@@ -369,8 +369,10 @@  static int gve_adminq_create_tx_queue(struct gve_priv *priv, u32 queue_index)
 {
 	struct gve_tx_ring *tx = &priv->tx[queue_index];
 	union gve_adminq_command cmd;
+	u32 qpl_id;
 	int err;
 
+	qpl_id = priv->raw_addressing ? GVE_RAW_ADDRESSING_QPL_ID : tx->tx_fifo.qpl->id;
 	memset(&cmd, 0, sizeof(cmd));
 	cmd.opcode = cpu_to_be32(GVE_ADMINQ_CREATE_TX_QUEUE);
 	cmd.create_tx_queue = (struct gve_adminq_create_tx_queue) {
@@ -379,7 +381,7 @@  static int gve_adminq_create_tx_queue(struct gve_priv *priv, u32 queue_index)
 		.queue_resources_addr =
 			cpu_to_be64(tx->q_resources_bus),
 		.tx_ring_addr = cpu_to_be64(tx->bus),
-		.queue_page_list_id = cpu_to_be32(tx->tx_fifo.qpl->id),
+		.queue_page_list_id = cpu_to_be32(qpl_id),
 		.ntfy_id = cpu_to_be32(tx->ntfy_id),
 	};
 
diff --git a/drivers/net/ethernet/google/gve/gve_desc.h b/drivers/net/ethernet/google/gve/gve_desc.h
index a1c0aaa60139..05ae6300e984 100644
--- a/drivers/net/ethernet/google/gve/gve_desc.h
+++ b/drivers/net/ethernet/google/gve/gve_desc.h
@@ -16,9 +16,11 @@ 
  * Base addresses encoded in seg_addr are not assumed to be physical
  * addresses. The ring format assumes these come from some linear address
  * space. This could be physical memory, kernel virtual memory, user virtual
- * memory. gVNIC uses lists of registered pages. Each queue is assumed
- * to be associated with a single such linear address space to ensure a
- * consistent meaning for seg_addrs posted to its rings.
+ * memory.
+ * If raw dma addressing is not supported then gVNIC uses lists of registered
+ * pages. Each queue is assumed to be associated with a single such linear
+ * address space to ensure a consistent meaning for seg_addrs posted to its
+ * rings.
  */
 
 struct gve_tx_pkt_desc {
diff --git a/drivers/net/ethernet/google/gve/gve_ethtool.c b/drivers/net/ethernet/google/gve/gve_ethtool.c
index 7b44769bd87c..07ec957ae091 100644
--- a/drivers/net/ethernet/google/gve/gve_ethtool.c
+++ b/drivers/net/ethernet/google/gve/gve_ethtool.c
@@ -50,6 +50,7 @@  static const char gve_gstrings_rx_stats[][ETH_GSTRING_LEN] = {
 static const char gve_gstrings_tx_stats[][ETH_GSTRING_LEN] = {
 	"tx_posted_desc[%u]", "tx_completed_desc[%u]", "tx_bytes[%u]",
 	"tx_wake[%u]", "tx_stop[%u]", "tx_event_counter[%u]",
+	"tx_dma_mapping_error[%u]",
 };
 
 static const char gve_gstrings_adminq_stats[][ETH_GSTRING_LEN] = {
@@ -322,6 +323,7 @@  gve_get_ethtool_stats(struct net_device *netdev,
 			data[i++] = tx->stop_queue;
 			data[i++] = be32_to_cpu(gve_tx_load_event_counter(priv,
 									  tx));
+			data[i++] = tx->dma_mapping_error;
 			/* stats from NIC */
 			if (skip_nic_stats) {
 				/* skip NIC tx stats */
diff --git a/drivers/net/ethernet/google/gve/gve_tx.c b/drivers/net/ethernet/google/gve/gve_tx.c
index d0244feb0301..6938f3a939d6 100644
--- a/drivers/net/ethernet/google/gve/gve_tx.c
+++ b/drivers/net/ethernet/google/gve/gve_tx.c
@@ -158,9 +158,11 @@  static void gve_tx_free_ring(struct gve_priv *priv, int idx)
 			  tx->q_resources, tx->q_resources_bus);
 	tx->q_resources = NULL;
 
-	gve_tx_fifo_release(priv, &tx->tx_fifo);
-	gve_unassign_qpl(priv, tx->tx_fifo.qpl->id);
-	tx->tx_fifo.qpl = NULL;
+	if (!tx->raw_addressing) {
+		gve_tx_fifo_release(priv, &tx->tx_fifo);
+		gve_unassign_qpl(priv, tx->tx_fifo.qpl->id);
+		tx->tx_fifo.qpl = NULL;
+	}
 
 	bytes = sizeof(*tx->desc) * slots;
 	dma_free_coherent(hdev, bytes, tx->desc, tx->bus);
@@ -206,11 +208,15 @@  static int gve_tx_alloc_ring(struct gve_priv *priv, int idx)
 	if (!tx->desc)
 		goto abort_with_info;
 
-	tx->tx_fifo.qpl = gve_assign_tx_qpl(priv);
+	tx->raw_addressing = priv->raw_addressing;
+	tx->dev = &priv->pdev->dev;
+	if (!tx->raw_addressing) {
+		tx->tx_fifo.qpl = gve_assign_tx_qpl(priv);
 
-	/* map Tx FIFO */
-	if (gve_tx_fifo_init(priv, &tx->tx_fifo))
-		goto abort_with_desc;
+		/* map Tx FIFO */
+		if (gve_tx_fifo_init(priv, &tx->tx_fifo))
+			goto abort_with_desc;
+	}
 
 	tx->q_resources =
 		dma_alloc_coherent(hdev,
@@ -228,7 +234,8 @@  static int gve_tx_alloc_ring(struct gve_priv *priv, int idx)
 	return 0;
 
 abort_with_fifo:
-	gve_tx_fifo_release(priv, &tx->tx_fifo);
+	if (!tx->raw_addressing)
+		gve_tx_fifo_release(priv, &tx->tx_fifo);
 abort_with_desc:
 	dma_free_coherent(hdev, bytes, tx->desc, tx->bus);
 	tx->desc = NULL;
@@ -301,27 +308,47 @@  static inline int gve_skb_fifo_bytes_required(struct gve_tx_ring *tx,
 	return bytes;
 }
 
-/* The most descriptors we could need are 3 - 1 for the headers, 1 for
- * the beginning of the payload at the end of the FIFO, and 1 if the
- * payload wraps to the beginning of the FIFO.
+/* The most descriptors we could need is MAX_SKB_FRAGS + 3 : 1 for each skb frag,
+ * +1 for the skb linear portion, +1 for when tcp hdr needs to be in separate descriptor,
+ * and +1 if the payload wraps to the beginning of the FIFO.
  */
-#define MAX_TX_DESC_NEEDED	3
+#define MAX_TX_DESC_NEEDED	(MAX_SKB_FRAGS + 3)
+static void gve_tx_unmap_buf(struct device *dev, struct gve_tx_buffer_state *info)
+{
+	if (info->skb) {
+		dma_unmap_single(dev, dma_unmap_addr(&info->buf, dma),
+				 dma_unmap_len(&info->buf, len),
+				 DMA_TO_DEVICE);
+		dma_unmap_len_set(&info->buf, len, 0);
+	} else {
+		dma_unmap_page(dev, dma_unmap_addr(&info->buf, dma),
+			       dma_unmap_len(&info->buf, len),
+			       DMA_TO_DEVICE);
+		dma_unmap_len_set(&info->buf, len, 0);
+	}
+}
 
 /* Check if sufficient resources (descriptor ring space, FIFO space) are
  * available to transmit the given number of bytes.
  */
 static inline bool gve_can_tx(struct gve_tx_ring *tx, int bytes_required)
 {
-	return (gve_tx_avail(tx) >= MAX_TX_DESC_NEEDED &&
-		gve_tx_fifo_can_alloc(&tx->tx_fifo, bytes_required));
+	bool can_alloc = true;
+
+	if (!tx->raw_addressing)
+		can_alloc = gve_tx_fifo_can_alloc(&tx->tx_fifo, bytes_required);
+
+	return (gve_tx_avail(tx) >= MAX_TX_DESC_NEEDED && can_alloc);
 }
 
 /* Stops the queue if the skb cannot be transmitted. */
 static int gve_maybe_stop_tx(struct gve_tx_ring *tx, struct sk_buff *skb)
 {
-	int bytes_required;
+	int bytes_required = 0;
+
+	if (!tx->raw_addressing)
+		bytes_required = gve_skb_fifo_bytes_required(tx, skb);
 
-	bytes_required = gve_skb_fifo_bytes_required(tx, skb);
 	if (likely(gve_can_tx(tx, bytes_required)))
 		return 0;
 
@@ -395,17 +422,13 @@  static void gve_dma_sync_for_device(struct device *dev, dma_addr_t *page_buses,
 {
 	u64 last_page = (iov_offset + iov_len - 1) / PAGE_SIZE;
 	u64 first_page = iov_offset / PAGE_SIZE;
-	dma_addr_t dma;
 	u64 page;
 
-	for (page = first_page; page <= last_page; page++) {
-		dma = page_buses[page];
-		dma_sync_single_for_device(dev, dma, PAGE_SIZE, DMA_TO_DEVICE);
-	}
+	for (page = first_page; page <= last_page; page++)
+		dma_sync_single_for_device(dev, page_buses[page], PAGE_SIZE, DMA_TO_DEVICE);
 }
 
-static int gve_tx_add_skb(struct gve_tx_ring *tx, struct sk_buff *skb,
-			  struct device *dev)
+static int gve_tx_add_skb_copy(struct gve_priv *priv, struct gve_tx_ring *tx, struct sk_buff *skb)
 {
 	int pad_bytes, hlen, hdr_nfrags, payload_nfrags, l4_hdr_offset;
 	union gve_tx_desc *pkt_desc, *seg_desc;
@@ -447,7 +470,7 @@  static int gve_tx_add_skb(struct gve_tx_ring *tx, struct sk_buff *skb,
 	skb_copy_bits(skb, 0,
 		      tx->tx_fifo.base + info->iov[hdr_nfrags - 1].iov_offset,
 		      hlen);
-	gve_dma_sync_for_device(dev, tx->tx_fifo.qpl->page_buses,
+	gve_dma_sync_for_device(&priv->pdev->dev, tx->tx_fifo.qpl->page_buses,
 				info->iov[hdr_nfrags - 1].iov_offset,
 				info->iov[hdr_nfrags - 1].iov_len);
 	copy_offset = hlen;
@@ -463,7 +486,7 @@  static int gve_tx_add_skb(struct gve_tx_ring *tx, struct sk_buff *skb,
 		skb_copy_bits(skb, copy_offset,
 			      tx->tx_fifo.base + info->iov[i].iov_offset,
 			      info->iov[i].iov_len);
-		gve_dma_sync_for_device(dev, tx->tx_fifo.qpl->page_buses,
+		gve_dma_sync_for_device(&priv->pdev->dev, tx->tx_fifo.qpl->page_buses,
 					info->iov[i].iov_offset,
 					info->iov[i].iov_len);
 		copy_offset += info->iov[i].iov_len;
@@ -472,6 +495,94 @@  static int gve_tx_add_skb(struct gve_tx_ring *tx, struct sk_buff *skb,
 	return 1 + payload_nfrags;
 }
 
+static int gve_tx_add_skb_no_copy(struct gve_priv *priv, struct gve_tx_ring *tx,
+				  struct sk_buff *skb)
+{
+	const struct skb_shared_info *shinfo = skb_shinfo(skb);
+	int hlen, payload_nfrags, l4_hdr_offset;
+	union gve_tx_desc *pkt_desc, *seg_desc;
+	struct gve_tx_buffer_state *info;
+	bool is_gso = skb_is_gso(skb);
+	u32 idx = tx->req & tx->mask;
+	struct gve_tx_dma_buf *buf;
+	u64 addr;
+	u32 len;
+	int i;
+
+	info = &tx->info[idx];
+	pkt_desc = &tx->desc[idx];
+
+	l4_hdr_offset = skb_checksum_start_offset(skb);
+	/* If the skb is gso, then we want only up to the tcp header in the first segment
+	 * to efficiently replicate on each segment otherwise we want the linear portion
+	 * of the skb (which will contain the checksum because skb->csum_start and
+	 * skb->csum_offset are given relative to skb->head) in the first segment.
+	 */
+	hlen = is_gso ? l4_hdr_offset + tcp_hdrlen(skb) : skb_headlen(skb);
+	len = skb_headlen(skb);
+
+	info->skb =  skb;
+
+	addr = dma_map_single(tx->dev, skb->data, len, DMA_TO_DEVICE);
+	if (unlikely(dma_mapping_error(tx->dev, addr))) {
+		tx->dma_mapping_error++;
+		goto drop;
+	}
+	buf = &info->buf;
+	dma_unmap_len_set(buf, len, len);
+	dma_unmap_addr_set(buf, dma, addr);
+
+	payload_nfrags = shinfo->nr_frags;
+	if (hlen < len) {
+		/* For gso the rest of the linear portion of the skb needs to
+		 * be in its own descriptor.
+		 */
+		payload_nfrags++;
+		gve_tx_fill_pkt_desc(pkt_desc, skb, is_gso, l4_hdr_offset,
+				     1 + payload_nfrags, hlen, addr);
+
+		len -= hlen;
+		addr += hlen;
+		idx = (tx->req + 1) & tx->mask;
+		seg_desc = &tx->desc[idx];
+		gve_tx_fill_seg_desc(seg_desc, skb, is_gso, len, addr);
+	} else {
+		gve_tx_fill_pkt_desc(pkt_desc, skb, is_gso, l4_hdr_offset,
+				     1 + payload_nfrags, hlen, addr);
+	}
+
+	for (i = 0; i < shinfo->nr_frags; i++) {
+		const skb_frag_t *frag = &shinfo->frags[i];
+
+		idx = (idx + 1) & tx->mask;
+		seg_desc = &tx->desc[idx];
+		len = skb_frag_size(frag);
+		addr = skb_frag_dma_map(tx->dev, frag, 0, len, DMA_TO_DEVICE);
+		if (unlikely(dma_mapping_error(tx->dev, addr))) {
+			tx->dma_mapping_error++;
+			goto unmap_drop;
+		}
+		buf = &tx->info[idx].buf;
+		tx->info[idx].skb = NULL;
+		dma_unmap_len_set(buf, len, len);
+		dma_unmap_addr_set(buf, dma, addr);
+
+		gve_tx_fill_seg_desc(seg_desc, skb, is_gso, len, addr);
+	}
+
+	return 1 + payload_nfrags;
+
+unmap_drop:
+	i += (payload_nfrags == shinfo->nr_frags ? 1 : 2);
+	while (i--) {
+		idx--;
+		gve_tx_unmap_buf(tx->dev, &tx->info[idx & tx->mask]);
+	}
+drop:
+	tx->dropped_pkt++;
+	return 0;
+}
+
 netdev_tx_t gve_tx(struct sk_buff *skb, struct net_device *dev)
 {
 	struct gve_priv *priv = netdev_priv(dev);
@@ -490,17 +601,26 @@  netdev_tx_t gve_tx(struct sk_buff *skb, struct net_device *dev)
 		gve_tx_put_doorbell(priv, tx->q_resources, tx->req);
 		return NETDEV_TX_BUSY;
 	}
-	nsegs = gve_tx_add_skb(tx, skb, &priv->pdev->dev);
-
-	netdev_tx_sent_queue(tx->netdev_txq, skb->len);
-	skb_tx_timestamp(skb);
-
-	/* give packets to NIC */
-	tx->req += nsegs;
+	if (tx->raw_addressing)
+		nsegs = gve_tx_add_skb_no_copy(priv, tx, skb);
+	else
+		nsegs = gve_tx_add_skb_copy(priv, tx, skb);
+
+	/* If the packet is getting sent, we need to update the skb */
+	if (nsegs) {
+		netdev_tx_sent_queue(tx->netdev_txq, skb->len);
+		skb_tx_timestamp(skb);
+		tx->req += nsegs;
+	} else {
+		dev_kfree_skb_any(skb);
+	}
 
 	if (!netif_xmit_stopped(tx->netdev_txq) && netdev_xmit_more())
 		return NETDEV_TX_OK;
 
+	/* Give packets to NIC. Even if this packet failed to send the doorbell
+	 * might need to be rung because of xmit_more.
+	 */
 	gve_tx_put_doorbell(priv, tx->q_resources, tx->req);
 	return NETDEV_TX_OK;
 }
@@ -525,24 +645,29 @@  static int gve_clean_tx_done(struct gve_priv *priv, struct gve_tx_ring *tx,
 		info = &tx->info[idx];
 		skb = info->skb;
 
+		/* Unmap the buffer */
+		if (tx->raw_addressing)
+			gve_tx_unmap_buf(tx->dev, info);
+		tx->done++;
 		/* Mark as free */
 		if (skb) {
 			info->skb = NULL;
 			bytes += skb->len;
 			pkts++;
 			dev_consume_skb_any(skb);
+			if (tx->raw_addressing)
+				continue;
 			/* FIFO free */
 			for (i = 0; i < ARRAY_SIZE(info->iov); i++) {
-				space_freed += info->iov[i].iov_len +
-					       info->iov[i].iov_padding;
+				space_freed += info->iov[i].iov_len + info->iov[i].iov_padding;
 				info->iov[i].iov_len = 0;
 				info->iov[i].iov_padding = 0;
 			}
 		}
-		tx->done++;
 	}
 
-	gve_tx_free_fifo(&tx->tx_fifo, space_freed);
+	if (!tx->raw_addressing)
+		gve_tx_free_fifo(&tx->tx_fifo, space_freed);
 	u64_stats_update_begin(&tx->statss);
 	tx->bytes_done += bytes;
 	tx->pkt_done += pkts;