From 01573e231f18eb2d99162747186f59511f56b64d Mon Sep 17 00:00:00 2001
From: hc <hc@nodka.com>
Date: Fri, 08 Dec 2023 10:40:48 +0000
Subject: [PATCH] 移去rt
---
kernel/drivers/net/ethernet/intel/ice/ice_txrx.c | 1895 +++++++++++++++++++++++++++++++++++++++++-----------------
1 files changed, 1,337 insertions(+), 558 deletions(-)
diff --git a/kernel/drivers/net/ethernet/intel/ice/ice_txrx.c b/kernel/drivers/net/ethernet/intel/ice/ice_txrx.c
index 1d84fed..442a9bc 100644
--- a/kernel/drivers/net/ethernet/intel/ice/ice_txrx.c
+++ b/kernel/drivers/net/ethernet/intel/ice/ice_txrx.c
@@ -5,9 +5,99 @@
#include <linux/prefetch.h>
#include <linux/mm.h>
+#include <linux/bpf_trace.h>
+#include <net/xdp.h>
+#include "ice_txrx_lib.h"
+#include "ice_lib.h"
#include "ice.h"
+#include "ice_dcb_lib.h"
+#include "ice_xsk.h"
#define ICE_RX_HDR_SIZE 256
+
+#define FDIR_DESC_RXDID 0x40
+#define ICE_FDIR_CLEAN_DELAY 10
+
+/**
+ * ice_prgm_fdir_fltr - Program a Flow Director filter
+ * @vsi: VSI to send dummy packet
+ * @fdir_desc: flow director descriptor
+ * @raw_packet: allocated buffer for flow director
+ */
+int
+ice_prgm_fdir_fltr(struct ice_vsi *vsi, struct ice_fltr_desc *fdir_desc,
+ u8 *raw_packet)
+{
+ struct ice_tx_buf *tx_buf, *first;
+ struct ice_fltr_desc *f_desc;
+ struct ice_tx_desc *tx_desc;
+ struct ice_ring *tx_ring;
+ struct device *dev;
+ dma_addr_t dma;
+ u32 td_cmd;
+ u16 i;
+
+ /* VSI and Tx ring */
+ if (!vsi)
+ return -ENOENT;
+ tx_ring = vsi->tx_rings[0];
+ if (!tx_ring || !tx_ring->desc)
+ return -ENOENT;
+ dev = tx_ring->dev;
+
+ /* we are using two descriptors to add/del a filter and we can wait */
+ for (i = ICE_FDIR_CLEAN_DELAY; ICE_DESC_UNUSED(tx_ring) < 2; i--) {
+ if (!i)
+ return -EAGAIN;
+ msleep_interruptible(1);
+ }
+
+ dma = dma_map_single(dev, raw_packet, ICE_FDIR_MAX_RAW_PKT_SIZE,
+ DMA_TO_DEVICE);
+
+ if (dma_mapping_error(dev, dma))
+ return -EINVAL;
+
+ /* grab the next descriptor */
+ i = tx_ring->next_to_use;
+ first = &tx_ring->tx_buf[i];
+ f_desc = ICE_TX_FDIRDESC(tx_ring, i);
+ memcpy(f_desc, fdir_desc, sizeof(*f_desc));
+
+ i++;
+ i = (i < tx_ring->count) ? i : 0;
+ tx_desc = ICE_TX_DESC(tx_ring, i);
+ tx_buf = &tx_ring->tx_buf[i];
+
+ i++;
+ tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
+
+ memset(tx_buf, 0, sizeof(*tx_buf));
+ dma_unmap_len_set(tx_buf, len, ICE_FDIR_MAX_RAW_PKT_SIZE);
+ dma_unmap_addr_set(tx_buf, dma, dma);
+
+ tx_desc->buf_addr = cpu_to_le64(dma);
+ td_cmd = ICE_TXD_LAST_DESC_CMD | ICE_TX_DESC_CMD_DUMMY |
+ ICE_TX_DESC_CMD_RE;
+
+ tx_buf->tx_flags = ICE_TX_FLAGS_DUMMY_PKT;
+ tx_buf->raw_buf = raw_packet;
+
+ tx_desc->cmd_type_offset_bsz =
+ ice_build_ctob(td_cmd, 0, ICE_FDIR_MAX_RAW_PKT_SIZE, 0);
+
+ /* Force memory write to complete before letting h/w know
+ * there are new descriptors to fetch.
+ */
+ wmb();
+
+ /* mark the data descriptor to be watched */
+ first->next_to_watch = tx_desc;
+
+ writel(tx_ring->next_to_use, tx_ring->tail);
+
+ return 0;
+}
/**
* ice_unmap_and_free_tx_buf - Release a Tx buffer
@@ -18,7 +108,12 @@
ice_unmap_and_free_tx_buf(struct ice_ring *ring, struct ice_tx_buf *tx_buf)
{
if (tx_buf->skb) {
- dev_kfree_skb_any(tx_buf->skb);
+ if (tx_buf->tx_flags & ICE_TX_FLAGS_DUMMY_PKT)
+ devm_kfree(ring->dev, tx_buf->raw_buf);
+ else if (ice_ring_is_xdp(ring))
+ page_frag_free(tx_buf->raw_buf);
+ else
+ dev_kfree_skb_any(tx_buf->skb);
if (dma_unmap_len(tx_buf, len))
dma_unmap_single(ring->dev,
dma_unmap_addr(tx_buf, dma),
@@ -48,19 +143,23 @@
*/
void ice_clean_tx_ring(struct ice_ring *tx_ring)
{
- unsigned long size;
u16 i;
+
+ if (ice_ring_is_xdp(tx_ring) && tx_ring->xsk_pool) {
+ ice_xsk_clean_xdp_ring(tx_ring);
+ goto tx_skip_free;
+ }
/* ring already cleared, nothing to do */
if (!tx_ring->tx_buf)
return;
- /* Free all the Tx ring sk_bufss */
+ /* Free all the Tx ring sk_buffs */
for (i = 0; i < tx_ring->count; i++)
ice_unmap_and_free_tx_buf(tx_ring, &tx_ring->tx_buf[i]);
- size = sizeof(struct ice_tx_buf) * tx_ring->count;
- memset(tx_ring->tx_buf, 0, size);
+tx_skip_free:
+ memset(tx_ring->tx_buf, 0, sizeof(*tx_ring->tx_buf) * tx_ring->count);
/* Zero out the descriptor ring */
memset(tx_ring->desc, 0, tx_ring->size);
@@ -96,17 +195,16 @@
/**
* ice_clean_tx_irq - Reclaim resources after transmit completes
- * @vsi: the VSI we care about
* @tx_ring: Tx ring to clean
* @napi_budget: Used to determine if we are in netpoll
*
* Returns true if there's any budget left (e.g. the clean is finished)
*/
-static bool ice_clean_tx_irq(struct ice_vsi *vsi, struct ice_ring *tx_ring,
- int napi_budget)
+static bool ice_clean_tx_irq(struct ice_ring *tx_ring, int napi_budget)
{
unsigned int total_bytes = 0, total_pkts = 0;
- unsigned int budget = vsi->work_lmt;
+ unsigned int budget = ICE_DFLT_IRQ_WORK;
+ struct ice_vsi *vsi = tx_ring->vsi;
s16 i = tx_ring->next_to_clean;
struct ice_tx_desc *tx_desc;
struct ice_tx_buf *tx_buf;
@@ -114,6 +212,8 @@
tx_buf = &tx_ring->tx_buf[i];
tx_desc = ICE_TX_DESC(tx_ring, i);
i -= tx_ring->count;
+
+ prefetch(&vsi->state);
do {
struct ice_tx_desc *eop_desc = tx_buf->next_to_watch;
@@ -136,8 +236,11 @@
total_bytes += tx_buf->bytecount;
total_pkts += tx_buf->gso_segs;
- /* free the skb */
- napi_consume_skb(tx_buf->skb, napi_budget);
+ if (ice_ring_is_xdp(tx_ring))
+ page_frag_free(tx_buf->raw_buf);
+ else
+ /* free the skb */
+ napi_consume_skb(tx_buf->skb, napi_budget);
/* unmap skb header data */
dma_unmap_single(tx_ring->dev,
@@ -188,12 +291,11 @@
i += tx_ring->count;
tx_ring->next_to_clean = i;
- u64_stats_update_begin(&tx_ring->syncp);
- tx_ring->stats.bytes += total_bytes;
- tx_ring->stats.pkts += total_pkts;
- u64_stats_update_end(&tx_ring->syncp);
- tx_ring->q_vector->tx.total_bytes += total_bytes;
- tx_ring->q_vector->tx.total_pkts += total_pkts;
+
+ ice_update_tx_ring_stats(tx_ring, total_pkts, total_bytes);
+
+ if (ice_ring_is_xdp(tx_ring))
+ return !!budget;
netdev_tx_completed_queue(txring_txq(tx_ring), total_pkts,
total_bytes);
@@ -207,7 +309,7 @@
smp_mb();
if (__netif_subqueue_stopped(tx_ring->netdev,
tx_ring->q_index) &&
- !test_bit(__ICE_DOWN, vsi->state)) {
+ !test_bit(__ICE_DOWN, vsi->state)) {
netif_wake_subqueue(tx_ring->netdev,
tx_ring->q_index);
++tx_ring->tx_stats.restart_q;
@@ -219,28 +321,28 @@
/**
* ice_setup_tx_ring - Allocate the Tx descriptors
- * @tx_ring: the tx ring to set up
+ * @tx_ring: the Tx ring to set up
*
* Return 0 on success, negative on error
*/
int ice_setup_tx_ring(struct ice_ring *tx_ring)
{
struct device *dev = tx_ring->dev;
- int bi_size;
if (!dev)
return -ENOMEM;
/* warn if we are about to overwrite the pointer */
WARN_ON(tx_ring->tx_buf);
- bi_size = sizeof(struct ice_tx_buf) * tx_ring->count;
- tx_ring->tx_buf = devm_kzalloc(dev, bi_size, GFP_KERNEL);
+ tx_ring->tx_buf =
+ devm_kzalloc(dev, sizeof(*tx_ring->tx_buf) * tx_ring->count,
+ GFP_KERNEL);
if (!tx_ring->tx_buf)
return -ENOMEM;
- /* round up to nearest 4K */
- tx_ring->size = tx_ring->count * sizeof(struct ice_tx_desc);
- tx_ring->size = ALIGN(tx_ring->size, 4096);
+ /* round up to nearest page */
+ tx_ring->size = ALIGN(tx_ring->count * sizeof(struct ice_tx_desc),
+ PAGE_SIZE);
tx_ring->desc = dmam_alloc_coherent(dev, tx_ring->size, &tx_ring->dma,
GFP_KERNEL);
if (!tx_ring->desc) {
@@ -251,6 +353,7 @@
tx_ring->next_to_use = 0;
tx_ring->next_to_clean = 0;
+ tx_ring->tx_stats.prev_pkt = -1;
return 0;
err:
@@ -266,12 +369,16 @@
void ice_clean_rx_ring(struct ice_ring *rx_ring)
{
struct device *dev = rx_ring->dev;
- unsigned long size;
u16 i;
/* ring already cleared, nothing to do */
if (!rx_ring->rx_buf)
return;
+
+ if (rx_ring->xsk_pool) {
+ ice_xsk_clean_rx_ring(rx_ring);
+ goto rx_skip_free;
+ }
/* Free all the Rx ring sk_buffs */
for (i = 0; i < rx_ring->count; i++) {
@@ -284,15 +391,25 @@
if (!rx_buf->page)
continue;
- dma_unmap_page(dev, rx_buf->dma, PAGE_SIZE, DMA_FROM_DEVICE);
- __free_pages(rx_buf->page, 0);
+ /* Invalidate cache lines that may have been written to by
+ * device so that we avoid corrupting memory.
+ */
+ dma_sync_single_range_for_cpu(dev, rx_buf->dma,
+ rx_buf->page_offset,
+ rx_ring->rx_buf_len,
+ DMA_FROM_DEVICE);
+
+ /* free resources associated with mapping */
+ dma_unmap_page_attrs(dev, rx_buf->dma, ice_rx_pg_size(rx_ring),
+ DMA_FROM_DEVICE, ICE_RX_DMA_ATTR);
+ __page_frag_cache_drain(rx_buf->page, rx_buf->pagecnt_bias);
rx_buf->page = NULL;
rx_buf->page_offset = 0;
}
- size = sizeof(struct ice_rx_buf) * rx_ring->count;
- memset(rx_ring->rx_buf, 0, size);
+rx_skip_free:
+ memset(rx_ring->rx_buf, 0, sizeof(*rx_ring->rx_buf) * rx_ring->count);
/* Zero out the descriptor ring */
memset(rx_ring->desc, 0, rx_ring->size);
@@ -311,6 +428,10 @@
void ice_free_rx_ring(struct ice_ring *rx_ring)
{
ice_clean_rx_ring(rx_ring);
+ if (rx_ring->vsi->type == ICE_VSI_PF)
+ if (xdp_rxq_info_is_reg(&rx_ring->xdp_rxq))
+ xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
+ rx_ring->xdp_prog = NULL;
devm_kfree(rx_ring->dev, rx_ring->rx_buf);
rx_ring->rx_buf = NULL;
@@ -323,28 +444,28 @@
/**
* ice_setup_rx_ring - Allocate the Rx descriptors
- * @rx_ring: the rx ring to set up
+ * @rx_ring: the Rx ring to set up
*
* Return 0 on success, negative on error
*/
int ice_setup_rx_ring(struct ice_ring *rx_ring)
{
struct device *dev = rx_ring->dev;
- int bi_size;
if (!dev)
return -ENOMEM;
/* warn if we are about to overwrite the pointer */
WARN_ON(rx_ring->rx_buf);
- bi_size = sizeof(struct ice_rx_buf) * rx_ring->count;
- rx_ring->rx_buf = devm_kzalloc(dev, bi_size, GFP_KERNEL);
+ rx_ring->rx_buf =
+ devm_kzalloc(dev, sizeof(*rx_ring->rx_buf) * rx_ring->count,
+ GFP_KERNEL);
if (!rx_ring->rx_buf)
return -ENOMEM;
- /* round up to nearest 4K */
- rx_ring->size = rx_ring->count * sizeof(union ice_32byte_rx_desc);
- rx_ring->size = ALIGN(rx_ring->size, 4096);
+ /* round up to nearest page */
+ rx_ring->size = ALIGN(rx_ring->count * sizeof(union ice_32byte_rx_desc),
+ PAGE_SIZE);
rx_ring->desc = dmam_alloc_coherent(dev, rx_ring->size, &rx_ring->dma,
GFP_KERNEL);
if (!rx_ring->desc) {
@@ -355,6 +476,15 @@
rx_ring->next_to_use = 0;
rx_ring->next_to_clean = 0;
+
+ if (ice_is_xdp_ena_vsi(rx_ring->vsi))
+ WRITE_ONCE(rx_ring->xdp_prog, rx_ring->vsi->xdp_prog);
+
+ if (rx_ring->vsi->type == ICE_VSI_PF &&
+ !xdp_rxq_info_is_reg(&rx_ring->xdp_rxq))
+ if (xdp_rxq_info_reg(&rx_ring->xdp_rxq, rx_ring->netdev,
+ rx_ring->q_index))
+ goto err;
return 0;
err:
@@ -364,24 +494,127 @@
}
/**
- * ice_release_rx_desc - Store the new tail and head values
- * @rx_ring: ring to bump
- * @val: new head index
+ * ice_rx_offset - Return expected offset into page to access data
+ * @rx_ring: Ring we are requesting offset of
+ *
+ * Returns the offset value for ring into the data buffer.
*/
-static void ice_release_rx_desc(struct ice_ring *rx_ring, u32 val)
+static unsigned int ice_rx_offset(struct ice_ring *rx_ring)
{
- rx_ring->next_to_use = val;
+ if (ice_ring_uses_build_skb(rx_ring))
+ return ICE_SKB_PAD;
+ else if (ice_is_xdp_ena_vsi(rx_ring->vsi))
+ return XDP_PACKET_HEADROOM;
- /* update next to alloc since we have filled the ring */
- rx_ring->next_to_alloc = val;
+ return 0;
+}
- /* 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(val, rx_ring->tail);
+static unsigned int
+ice_rx_frame_truesize(struct ice_ring *rx_ring, unsigned int __maybe_unused size)
+{
+ unsigned int truesize;
+
+#if (PAGE_SIZE < 8192)
+ truesize = ice_rx_pg_size(rx_ring) / 2; /* Must be power-of-2 */
+#else
+ truesize = ice_rx_offset(rx_ring) ?
+ SKB_DATA_ALIGN(ice_rx_offset(rx_ring) + size) +
+ SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) :
+ SKB_DATA_ALIGN(size);
+#endif
+ return truesize;
+}
+
+/**
+ * ice_run_xdp - Executes an XDP program on initialized xdp_buff
+ * @rx_ring: Rx ring
+ * @xdp: xdp_buff used as input to the XDP program
+ * @xdp_prog: XDP program to run
+ *
+ * Returns any of ICE_XDP_{PASS, CONSUMED, TX, REDIR}
+ */
+static int
+ice_run_xdp(struct ice_ring *rx_ring, struct xdp_buff *xdp,
+ struct bpf_prog *xdp_prog)
+{
+ struct ice_ring *xdp_ring;
+ int err, result;
+ u32 act;
+
+ act = bpf_prog_run_xdp(xdp_prog, xdp);
+ switch (act) {
+ case XDP_PASS:
+ return ICE_XDP_PASS;
+ case XDP_TX:
+ xdp_ring = rx_ring->vsi->xdp_rings[smp_processor_id()];
+ result = ice_xmit_xdp_buff(xdp, xdp_ring);
+ if (result == ICE_XDP_CONSUMED)
+ goto out_failure;
+ return result;
+ case XDP_REDIRECT:
+ err = xdp_do_redirect(rx_ring->netdev, xdp, xdp_prog);
+ if (err)
+ goto out_failure;
+ return ICE_XDP_REDIR;
+ default:
+ bpf_warn_invalid_xdp_action(act);
+ fallthrough;
+ case XDP_ABORTED:
+out_failure:
+ trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
+ fallthrough;
+ case XDP_DROP:
+ return ICE_XDP_CONSUMED;
+ }
+}
+
+/**
+ * ice_xdp_xmit - submit packets to XDP ring for transmission
+ * @dev: netdev
+ * @n: number of XDP frames to be transmitted
+ * @frames: XDP frames to be transmitted
+ * @flags: transmit flags
+ *
+ * Returns number of frames successfully sent. Frames that fail are
+ * free'ed via XDP return API.
+ * For error cases, a negative errno code is returned and no-frames
+ * are transmitted (caller must handle freeing frames).
+ */
+int
+ice_xdp_xmit(struct net_device *dev, int n, struct xdp_frame **frames,
+ u32 flags)
+{
+ struct ice_netdev_priv *np = netdev_priv(dev);
+ unsigned int queue_index = smp_processor_id();
+ struct ice_vsi *vsi = np->vsi;
+ struct ice_ring *xdp_ring;
+ int drops = 0, i;
+
+ if (test_bit(__ICE_DOWN, vsi->state))
+ return -ENETDOWN;
+
+ if (!ice_is_xdp_ena_vsi(vsi) || queue_index >= vsi->num_xdp_txq)
+ return -ENXIO;
+
+ if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
+ return -EINVAL;
+
+ xdp_ring = vsi->xdp_rings[queue_index];
+ for (i = 0; i < n; i++) {
+ struct xdp_frame *xdpf = frames[i];
+ int err;
+
+ err = ice_xmit_xdp_ring(xdpf->data, xdpf->len, xdp_ring);
+ if (err != ICE_XDP_TX) {
+ xdp_return_frame_rx_napi(xdpf);
+ drops++;
+ }
+ }
+
+ if (unlikely(flags & XDP_XMIT_FLUSH))
+ ice_xdp_ring_update_tail(xdp_ring);
+
+ return n - drops;
}
/**
@@ -392,40 +625,41 @@
* Returns true if the page was successfully allocated or
* reused.
*/
-static bool ice_alloc_mapped_page(struct ice_ring *rx_ring,
- struct ice_rx_buf *bi)
+static bool
+ice_alloc_mapped_page(struct ice_ring *rx_ring, struct ice_rx_buf *bi)
{
struct page *page = bi->page;
dma_addr_t dma;
/* since we are recycling buffers we should seldom need to alloc */
- if (likely(page)) {
- rx_ring->rx_stats.page_reuse_count++;
+ if (likely(page))
return true;
- }
/* alloc new page for storage */
- page = alloc_page(GFP_ATOMIC | __GFP_NOWARN);
+ page = dev_alloc_pages(ice_rx_pg_order(rx_ring));
if (unlikely(!page)) {
rx_ring->rx_stats.alloc_page_failed++;
return false;
}
/* map page for use */
- dma = dma_map_page(rx_ring->dev, page, 0, PAGE_SIZE, DMA_FROM_DEVICE);
+ dma = dma_map_page_attrs(rx_ring->dev, page, 0, ice_rx_pg_size(rx_ring),
+ DMA_FROM_DEVICE, ICE_RX_DMA_ATTR);
/* if mapping failed free memory back to system since
* there isn't much point in holding memory we can't use
*/
if (dma_mapping_error(rx_ring->dev, dma)) {
- __free_pages(page, 0);
+ __free_pages(page, ice_rx_pg_order(rx_ring));
rx_ring->rx_stats.alloc_page_failed++;
return false;
}
bi->dma = dma;
bi->page = page;
- bi->page_offset = 0;
+ bi->page_offset = ice_rx_offset(rx_ring);
+ page_ref_add(page, USHRT_MAX - 1);
+ bi->pagecnt_bias = USHRT_MAX;
return true;
}
@@ -435,7 +669,13 @@
* @rx_ring: ring to place buffers on
* @cleaned_count: number of buffers to replace
*
- * Returns false if all allocations were successful, true if any fail
+ * Returns false if all allocations were successful, true if any fail. Returning
+ * true signals to the caller that we didn't replace cleaned_count buffers and
+ * there is more work to do.
+ *
+ * First, try to clean "cleaned_count" Rx buffers. Then refill the cleaned Rx
+ * buffers. Then bump tail at most one time. Grouping like this lets us avoid
+ * multiple tail writes per call.
*/
bool ice_alloc_rx_bufs(struct ice_ring *rx_ring, u16 cleaned_count)
{
@@ -444,16 +684,24 @@
struct ice_rx_buf *bi;
/* do nothing if no valid netdev defined */
- if (!rx_ring->netdev || !cleaned_count)
+ if ((!rx_ring->netdev && rx_ring->vsi->type != ICE_VSI_CTRL) ||
+ !cleaned_count)
return false;
- /* get the RX descriptor and buffer based on next_to_use */
+ /* get the Rx descriptor and buffer based on next_to_use */
rx_desc = ICE_RX_DESC(rx_ring, ntu);
bi = &rx_ring->rx_buf[ntu];
do {
+ /* if we fail here, we have work remaining */
if (!ice_alloc_mapped_page(rx_ring, bi))
- goto no_bufs;
+ break;
+
+ /* sync the buffer for use by the device */
+ dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
+ bi->page_offset,
+ rx_ring->rx_buf_len,
+ DMA_FROM_DEVICE);
/* Refresh the desc even if buffer_addrs didn't change
* because each write-back erases this info.
@@ -478,16 +726,7 @@
if (rx_ring->next_to_use != ntu)
ice_release_rx_desc(rx_ring, ntu);
- return false;
-
-no_bufs:
- if (rx_ring->next_to_use != ntu)
- ice_release_rx_desc(rx_ring, ntu);
-
- /* make sure to come back via polling to try again after
- * allocation failure
- */
- return true;
+ return !!cleaned_count;
}
/**
@@ -500,61 +739,42 @@
}
/**
- * ice_add_rx_frag - Add contents of Rx buffer to sk_buff
- * @rx_buf: buffer containing page to add
- * @rx_desc: descriptor containing length of buffer written by hardware
- * @skb: sk_buf to place the data into
+ * ice_rx_buf_adjust_pg_offset - Prepare Rx buffer for reuse
+ * @rx_buf: Rx buffer to adjust
+ * @size: Size of adjustment
*
- * This function will add the data contained in rx_buf->page to the skb.
- * This is done either through a direct copy if the data in the buffer is
- * less than the skb header size, otherwise it will just attach the page as
- * a frag to the skb.
- *
- * The function will then update the page offset if necessary and return
- * true if the buffer can be reused by the adapter.
+ * Update the offset within page so that Rx buf will be ready to be reused.
+ * For systems with PAGE_SIZE < 8192 this function will flip the page offset
+ * so the second half of page assigned to Rx buffer will be used, otherwise
+ * the offset is moved by "size" bytes
*/
-static bool ice_add_rx_frag(struct ice_rx_buf *rx_buf,
- union ice_32b_rx_flex_desc *rx_desc,
- struct sk_buff *skb)
+static void
+ice_rx_buf_adjust_pg_offset(struct ice_rx_buf *rx_buf, unsigned int size)
{
#if (PAGE_SIZE < 8192)
- unsigned int truesize = ICE_RXBUF_2048;
+ /* flip page offset to other buffer */
+ rx_buf->page_offset ^= size;
#else
- unsigned int last_offset = PAGE_SIZE - ICE_RXBUF_2048;
- unsigned int truesize;
-#endif /* PAGE_SIZE < 8192) */
+ /* move offset up to the next cache line */
+ rx_buf->page_offset += size;
+#endif
+}
- struct page *page;
- unsigned int size;
-
- size = le16_to_cpu(rx_desc->wb.pkt_len) &
- ICE_RX_FLX_DESC_PKT_LEN_M;
-
- page = rx_buf->page;
-
-#if (PAGE_SIZE >= 8192)
- truesize = ALIGN(size, L1_CACHE_BYTES);
-#endif /* PAGE_SIZE >= 8192) */
-
- /* will the data fit in the skb we allocated? if so, just
- * copy it as it is pretty small anyway
- */
- if (size <= ICE_RX_HDR_SIZE && !skb_is_nonlinear(skb)) {
- unsigned char *va = page_address(page) + rx_buf->page_offset;
-
- memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
-
- /* page is not reserved, we can reuse buffer as-is */
- if (likely(!ice_page_is_reserved(page)))
- return true;
-
- /* this page cannot be reused so discard it */
- __free_pages(page, 0);
- return false;
- }
-
- skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, page,
- rx_buf->page_offset, size, truesize);
+/**
+ * ice_can_reuse_rx_page - Determine if page can be reused for another Rx
+ * @rx_buf: buffer containing the page
+ * @rx_buf_pgcnt: rx_buf page refcount pre xdp_do_redirect() call
+ *
+ * If page is reusable, we have a green light for calling ice_reuse_rx_page,
+ * which will assign the current buffer to the buffer that next_to_alloc is
+ * pointing to; otherwise, the DMA mapping needs to be destroyed and
+ * page freed
+ */
+static bool
+ice_can_reuse_rx_page(struct ice_rx_buf *rx_buf, int rx_buf_pgcnt)
+{
+ unsigned int pagecnt_bias = rx_buf->pagecnt_bias;
+ struct page *page = rx_buf->page;
/* avoid re-using remote pages */
if (unlikely(ice_page_is_reserved(page)))
@@ -562,36 +782,66 @@
#if (PAGE_SIZE < 8192)
/* if we are only owner of page we can reuse it */
- if (unlikely(page_count(page) != 1))
+ if (unlikely((rx_buf_pgcnt - pagecnt_bias) > 1))
return false;
-
- /* flip page offset to other buffer */
- rx_buf->page_offset ^= truesize;
#else
- /* move offset up to the next cache line */
- rx_buf->page_offset += truesize;
-
- if (rx_buf->page_offset > last_offset)
+#define ICE_LAST_OFFSET \
+ (SKB_WITH_OVERHEAD(PAGE_SIZE) - ICE_RXBUF_2048)
+ if (rx_buf->page_offset > ICE_LAST_OFFSET)
return false;
#endif /* PAGE_SIZE < 8192) */
- /* Even if we own the page, we are not allowed to use atomic_set()
- * This would break get_page_unless_zero() users.
+ /* If we have drained the page fragment pool we need to update
+ * the pagecnt_bias and page count so that we fully restock the
+ * number of references the driver holds.
*/
- get_page(rx_buf->page);
+ if (unlikely(pagecnt_bias == 1)) {
+ page_ref_add(page, USHRT_MAX - 1);
+ rx_buf->pagecnt_bias = USHRT_MAX;
+ }
return true;
}
/**
+ * ice_add_rx_frag - Add contents of Rx buffer to sk_buff as a frag
+ * @rx_ring: Rx descriptor ring to transact packets on
+ * @rx_buf: buffer containing page to add
+ * @skb: sk_buff to place the data into
+ * @size: packet length from rx_desc
+ *
+ * This function will add the data contained in rx_buf->page to the skb.
+ * It will just attach the page as a frag to the skb.
+ * The function will then update the page offset.
+ */
+static void
+ice_add_rx_frag(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf,
+ struct sk_buff *skb, unsigned int size)
+{
+#if (PAGE_SIZE >= 8192)
+ unsigned int truesize = SKB_DATA_ALIGN(size + ice_rx_offset(rx_ring));
+#else
+ unsigned int truesize = ice_rx_pg_size(rx_ring) / 2;
+#endif
+
+ if (!size)
+ return;
+ skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buf->page,
+ rx_buf->page_offset, size, truesize);
+
+ /* page is being used so we must update the page offset */
+ ice_rx_buf_adjust_pg_offset(rx_buf, truesize);
+}
+
+/**
* ice_reuse_rx_page - page flip buffer and store it back on the ring
- * @rx_ring: rx descriptor ring to store buffers on
+ * @rx_ring: Rx descriptor ring to store buffers on
* @old_buf: donor buffer to have page reused
*
* Synchronizes page for reuse by the adapter
*/
-static void ice_reuse_rx_page(struct ice_ring *rx_ring,
- struct ice_rx_buf *old_buf)
+static void
+ice_reuse_rx_page(struct ice_ring *rx_ring, struct ice_rx_buf *old_buf)
{
u16 nta = rx_ring->next_to_alloc;
struct ice_rx_buf *new_buf;
@@ -602,163 +852,203 @@
nta++;
rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
- /* transfer page from old buffer to new buffer */
- *new_buf = *old_buf;
+ /* Transfer page from old buffer to new buffer.
+ * Move each member individually to avoid possible store
+ * forwarding stalls and unnecessary copy of skb.
+ */
+ new_buf->dma = old_buf->dma;
+ new_buf->page = old_buf->page;
+ new_buf->page_offset = old_buf->page_offset;
+ new_buf->pagecnt_bias = old_buf->pagecnt_bias;
}
/**
- * ice_fetch_rx_buf - Allocate skb and populate it
- * @rx_ring: rx descriptor ring to transact packets on
- * @rx_desc: descriptor containing info written by hardware
+ * ice_get_rx_buf - Fetch Rx buffer and synchronize data for use
+ * @rx_ring: Rx descriptor ring to transact packets on
+ * @skb: skb to be used
+ * @size: size of buffer to add to skb
+ * @rx_buf_pgcnt: rx_buf page refcount
*
- * This function allocates an skb on the fly, and populates it with the page
- * data from the current receive descriptor, taking care to set up the skb
- * correctly, as well as handling calling the page recycle function if
- * necessary.
+ * This function will pull an Rx buffer from the ring and synchronize it
+ * for use by the CPU.
*/
-static struct sk_buff *ice_fetch_rx_buf(struct ice_ring *rx_ring,
- union ice_32b_rx_flex_desc *rx_desc)
+static struct ice_rx_buf *
+ice_get_rx_buf(struct ice_ring *rx_ring, struct sk_buff **skb,
+ const unsigned int size, int *rx_buf_pgcnt)
{
struct ice_rx_buf *rx_buf;
- struct sk_buff *skb;
- struct page *page;
rx_buf = &rx_ring->rx_buf[rx_ring->next_to_clean];
- page = rx_buf->page;
- prefetchw(page);
+ *rx_buf_pgcnt =
+#if (PAGE_SIZE < 8192)
+ page_count(rx_buf->page);
+#else
+ 0;
+#endif
+ prefetchw(rx_buf->page);
+ *skb = rx_buf->skb;
- skb = rx_buf->skb;
+ if (!size)
+ return rx_buf;
+ /* we are reusing so sync this buffer for CPU use */
+ dma_sync_single_range_for_cpu(rx_ring->dev, rx_buf->dma,
+ rx_buf->page_offset, size,
+ DMA_FROM_DEVICE);
- if (likely(!skb)) {
- u8 *page_addr = page_address(page) + rx_buf->page_offset;
+ /* We have pulled a buffer for use, so decrement pagecnt_bias */
+ rx_buf->pagecnt_bias--;
- /* prefetch first cache line of first page */
- prefetch(page_addr);
-#if L1_CACHE_BYTES < 128
- prefetch((void *)(page_addr + L1_CACHE_BYTES));
-#endif /* L1_CACHE_BYTES */
+ return rx_buf;
+}
- /* allocate a skb to store the frags */
- skb = __napi_alloc_skb(&rx_ring->q_vector->napi,
- ICE_RX_HDR_SIZE,
- GFP_ATOMIC | __GFP_NOWARN);
- if (unlikely(!skb)) {
- rx_ring->rx_stats.alloc_buf_failed++;
- return NULL;
- }
+/**
+ * ice_build_skb - Build skb around an existing buffer
+ * @rx_ring: Rx descriptor ring to transact packets on
+ * @rx_buf: Rx buffer to pull data from
+ * @xdp: xdp_buff pointing to the data
+ *
+ * This function builds an skb around an existing Rx buffer, taking care
+ * to set up the skb correctly and avoid any memcpy overhead.
+ */
+static struct sk_buff *
+ice_build_skb(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf,
+ struct xdp_buff *xdp)
+{
+ u8 metasize = xdp->data - xdp->data_meta;
+#if (PAGE_SIZE < 8192)
+ unsigned int truesize = ice_rx_pg_size(rx_ring) / 2;
+#else
+ unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
+ SKB_DATA_ALIGN(xdp->data_end -
+ xdp->data_hard_start);
+#endif
+ struct sk_buff *skb;
- /* we will be copying header into skb->data in
- * pskb_may_pull so it is in our interest to prefetch
- * it now to avoid a possible cache miss
- */
- prefetchw(skb->data);
+ /* Prefetch first cache line of first page. If xdp->data_meta
+ * is unused, this points exactly as xdp->data, otherwise we
+ * likely have a consumer accessing first few bytes of meta
+ * data, and then actual data.
+ */
+ net_prefetch(xdp->data_meta);
+ /* build an skb around the page buffer */
+ skb = build_skb(xdp->data_hard_start, truesize);
+ if (unlikely(!skb))
+ return NULL;
- skb_record_rx_queue(skb, rx_ring->q_index);
- } else {
- /* we are reusing so sync this buffer for CPU use */
- dma_sync_single_range_for_cpu(rx_ring->dev, rx_buf->dma,
- rx_buf->page_offset,
- ICE_RXBUF_2048,
- DMA_FROM_DEVICE);
+ /* must to record Rx queue, otherwise OS features such as
+ * symmetric queue won't work
+ */
+ skb_record_rx_queue(skb, rx_ring->q_index);
- rx_buf->skb = NULL;
- }
+ /* update pointers within the skb to store the data */
+ skb_reserve(skb, xdp->data - xdp->data_hard_start);
+ __skb_put(skb, xdp->data_end - xdp->data);
+ if (metasize)
+ skb_metadata_set(skb, metasize);
- /* pull page into skb */
- if (ice_add_rx_frag(rx_buf, rx_desc, skb)) {
- /* hand second half of page back to the ring */
- ice_reuse_rx_page(rx_ring, rx_buf);
- rx_ring->rx_stats.page_reuse_count++;
- } else {
- /* we are not reusing the buffer so unmap it */
- dma_unmap_page(rx_ring->dev, rx_buf->dma, PAGE_SIZE,
- DMA_FROM_DEVICE);
- }
-
- /* clear contents of buffer_info */
- rx_buf->page = NULL;
+ /* buffer is used by skb, update page_offset */
+ ice_rx_buf_adjust_pg_offset(rx_buf, truesize);
return skb;
}
/**
- * ice_pull_tail - ice specific version of skb_pull_tail
- * @skb: pointer to current skb being adjusted
+ * ice_construct_skb - Allocate skb and populate it
+ * @rx_ring: Rx descriptor ring to transact packets on
+ * @rx_buf: Rx buffer to pull data from
+ * @xdp: xdp_buff pointing to the data
*
- * This function is an ice specific version of __pskb_pull_tail. The
- * main difference between this version and the original function is that
- * this function can make several assumptions about the state of things
- * that allow for significant optimizations versus the standard function.
- * As a result we can do things like drop a frag and maintain an accurate
- * truesize for the skb.
+ * This function allocates an skb. It then populates it with the page
+ * data from the current receive descriptor, taking care to set up the
+ * skb correctly.
*/
-static void ice_pull_tail(struct sk_buff *skb)
+static struct sk_buff *
+ice_construct_skb(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf,
+ struct xdp_buff *xdp)
{
- struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
- unsigned int pull_len;
- unsigned char *va;
+ unsigned int size = xdp->data_end - xdp->data;
+ unsigned int headlen;
+ struct sk_buff *skb;
- /* it is valid to use page_address instead of kmap since we are
- * working with pages allocated out of the lomem pool per
- * alloc_page(GFP_ATOMIC)
- */
- va = skb_frag_address(frag);
+ /* prefetch first cache line of first page */
+ net_prefetch(xdp->data);
- /* we need the header to contain the greater of either ETH_HLEN or
- * 60 bytes if the skb->len is less than 60 for skb_pad.
- */
- pull_len = eth_get_headlen(va, ICE_RX_HDR_SIZE);
+ /* allocate a skb to store the frags */
+ skb = __napi_alloc_skb(&rx_ring->q_vector->napi, ICE_RX_HDR_SIZE,
+ GFP_ATOMIC | __GFP_NOWARN);
+ if (unlikely(!skb))
+ return NULL;
+
+ skb_record_rx_queue(skb, rx_ring->q_index);
+ /* Determine available headroom for copy */
+ headlen = size;
+ if (headlen > ICE_RX_HDR_SIZE)
+ headlen = eth_get_headlen(skb->dev, xdp->data, ICE_RX_HDR_SIZE);
/* align pull length to size of long to optimize memcpy performance */
- skb_copy_to_linear_data(skb, va, ALIGN(pull_len, sizeof(long)));
+ memcpy(__skb_put(skb, headlen), xdp->data, ALIGN(headlen,
+ sizeof(long)));
- /* update all of the pointers */
- skb_frag_size_sub(frag, pull_len);
- frag->page_offset += pull_len;
- skb->data_len -= pull_len;
- skb->tail += pull_len;
+ /* if we exhaust the linear part then add what is left as a frag */
+ size -= headlen;
+ if (size) {
+#if (PAGE_SIZE >= 8192)
+ unsigned int truesize = SKB_DATA_ALIGN(size);
+#else
+ unsigned int truesize = ice_rx_pg_size(rx_ring) / 2;
+#endif
+ skb_add_rx_frag(skb, 0, rx_buf->page,
+ rx_buf->page_offset + headlen, size, truesize);
+ /* buffer is used by skb, update page_offset */
+ ice_rx_buf_adjust_pg_offset(rx_buf, truesize);
+ } else {
+ /* buffer is unused, reset bias back to rx_buf; data was copied
+ * onto skb's linear part so there's no need for adjusting
+ * page offset and we can reuse this buffer as-is
+ */
+ rx_buf->pagecnt_bias++;
+ }
+
+ return skb;
}
/**
- * ice_cleanup_headers - Correct empty headers
- * @skb: pointer to current skb being fixed
+ * ice_put_rx_buf - Clean up used buffer and either recycle or free
+ * @rx_ring: Rx descriptor ring to transact packets on
+ * @rx_buf: Rx buffer to pull data from
+ * @rx_buf_pgcnt: Rx buffer page count pre xdp_do_redirect()
*
- * Also address the case where we are pulling data in on pages only
- * and as such no data is present in the skb header.
- *
- * In addition if skb is not at least 60 bytes we need to pad it so that
- * it is large enough to qualify as a valid Ethernet frame.
- *
- * Returns true if an error was encountered and skb was freed.
+ * This function will update next_to_clean and then clean up the contents
+ * of the rx_buf. It will either recycle the buffer or unmap it and free
+ * the associated resources.
*/
-static bool ice_cleanup_headers(struct sk_buff *skb)
+static void
+ice_put_rx_buf(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf,
+ int rx_buf_pgcnt)
{
- /* place header in linear portion of buffer */
- if (skb_is_nonlinear(skb))
- ice_pull_tail(skb);
+ u16 ntc = rx_ring->next_to_clean + 1;
- /* if eth_skb_pad returns an error the skb was freed */
- if (eth_skb_pad(skb))
- return true;
+ /* fetch, update, and store next to clean */
+ ntc = (ntc < rx_ring->count) ? ntc : 0;
+ rx_ring->next_to_clean = ntc;
- return false;
-}
+ if (!rx_buf)
+ return;
-/**
- * ice_test_staterr - tests bits in Rx descriptor status and error fields
- * @rx_desc: pointer to receive descriptor (in le64 format)
- * @stat_err_bits: value to mask
- *
- * This function does some fast chicanery in order to return the
- * value of the mask which is really only used for boolean tests.
- * The status_error_len doesn't need to be shifted because it begins
- * at offset zero.
- */
-static bool ice_test_staterr(union ice_32b_rx_flex_desc *rx_desc,
- const u16 stat_err_bits)
-{
- return !!(rx_desc->wb.status_error0 &
- cpu_to_le16(stat_err_bits));
+ if (ice_can_reuse_rx_page(rx_buf, rx_buf_pgcnt)) {
+ /* hand second half of page back to the ring */
+ ice_reuse_rx_page(rx_ring, rx_buf);
+ } else {
+ /* we are not reusing the buffer so unmap it */
+ dma_unmap_page_attrs(rx_ring->dev, rx_buf->dma,
+ ice_rx_pg_size(rx_ring), DMA_FROM_DEVICE,
+ ICE_RX_DMA_ATTR);
+ __page_frag_cache_drain(rx_buf->page, rx_buf->pagecnt_bias);
+ }
+
+ /* clear contents of buffer_info */
+ rx_buf->page = NULL;
+ rx_buf->skb = NULL;
}
/**
@@ -767,216 +1057,64 @@
* @rx_desc: Rx descriptor for current buffer
* @skb: Current socket buffer containing buffer in progress
*
- * This function updates next to clean. If the buffer is an EOP buffer
- * this function exits returning false, otherwise it will place the
- * sk_buff in the next buffer to be chained and return true indicating
- * that this is in fact a non-EOP buffer.
+ * If the buffer is an EOP buffer, this function exits returning false,
+ * otherwise return true indicating that this is in fact a non-EOP buffer.
*/
-static bool ice_is_non_eop(struct ice_ring *rx_ring,
- union ice_32b_rx_flex_desc *rx_desc,
- struct sk_buff *skb)
+static bool
+ice_is_non_eop(struct ice_ring *rx_ring, union ice_32b_rx_flex_desc *rx_desc,
+ struct sk_buff *skb)
{
- u32 ntc = rx_ring->next_to_clean + 1;
-
- /* fetch, update, and store next to clean */
- ntc = (ntc < rx_ring->count) ? ntc : 0;
- rx_ring->next_to_clean = ntc;
-
- prefetch(ICE_RX_DESC(rx_ring, ntc));
-
/* if we are the last buffer then there is nothing else to do */
#define ICE_RXD_EOF BIT(ICE_RX_FLEX_DESC_STATUS0_EOF_S)
if (likely(ice_test_staterr(rx_desc, ICE_RXD_EOF)))
return false;
/* place skb in next buffer to be received */
- rx_ring->rx_buf[ntc].skb = skb;
+ rx_ring->rx_buf[rx_ring->next_to_clean].skb = skb;
rx_ring->rx_stats.non_eop_descs++;
return true;
}
/**
- * ice_ptype_to_htype - get a hash type
- * @ptype: the ptype value from the descriptor
- *
- * Returns a hash type to be used by skb_set_hash
- */
-static enum pkt_hash_types ice_ptype_to_htype(u8 __always_unused ptype)
-{
- return PKT_HASH_TYPE_NONE;
-}
-
-/**
- * ice_rx_hash - set the hash value in the skb
- * @rx_ring: descriptor ring
- * @rx_desc: specific descriptor
- * @skb: pointer to current skb
- * @rx_ptype: the ptype value from the descriptor
- */
-static void
-ice_rx_hash(struct ice_ring *rx_ring, union ice_32b_rx_flex_desc *rx_desc,
- struct sk_buff *skb, u8 rx_ptype)
-{
- struct ice_32b_rx_flex_desc_nic *nic_mdid;
- u32 hash;
-
- if (!(rx_ring->netdev->features & NETIF_F_RXHASH))
- return;
-
- if (rx_desc->wb.rxdid != ICE_RXDID_FLEX_NIC)
- return;
-
- nic_mdid = (struct ice_32b_rx_flex_desc_nic *)rx_desc;
- hash = le32_to_cpu(nic_mdid->rss_hash);
- skb_set_hash(skb, hash, ice_ptype_to_htype(rx_ptype));
-}
-
-/**
- * ice_rx_csum - Indicate in skb if checksum is good
- * @vsi: the VSI we care about
- * @skb: skb currently being received and modified
- * @rx_desc: the receive descriptor
- * @ptype: the packet type decoded by hardware
- *
- * skb->protocol must be set before this function is called
- */
-static void ice_rx_csum(struct ice_vsi *vsi, struct sk_buff *skb,
- union ice_32b_rx_flex_desc *rx_desc, u8 ptype)
-{
- struct ice_rx_ptype_decoded decoded;
- u32 rx_error, rx_status;
- bool ipv4, ipv6;
-
- rx_status = le16_to_cpu(rx_desc->wb.status_error0);
- rx_error = rx_status;
-
- decoded = ice_decode_rx_desc_ptype(ptype);
-
- /* Start with CHECKSUM_NONE and by default csum_level = 0 */
- skb->ip_summed = CHECKSUM_NONE;
- skb_checksum_none_assert(skb);
-
- /* check if Rx checksum is enabled */
- if (!(vsi->netdev->features & NETIF_F_RXCSUM))
- return;
-
- /* check if HW has decoded the packet and checksum */
- if (!(rx_status & BIT(ICE_RX_FLEX_DESC_STATUS0_L3L4P_S)))
- return;
-
- if (!(decoded.known && decoded.outer_ip))
- return;
-
- ipv4 = (decoded.outer_ip == ICE_RX_PTYPE_OUTER_IP) &&
- (decoded.outer_ip_ver == ICE_RX_PTYPE_OUTER_IPV4);
- ipv6 = (decoded.outer_ip == ICE_RX_PTYPE_OUTER_IP) &&
- (decoded.outer_ip_ver == ICE_RX_PTYPE_OUTER_IPV6);
-
- if (ipv4 && (rx_error & (BIT(ICE_RX_FLEX_DESC_STATUS0_XSUM_IPE_S) |
- BIT(ICE_RX_FLEX_DESC_STATUS0_XSUM_EIPE_S))))
- goto checksum_fail;
- else if (ipv6 && (rx_status &
- (BIT(ICE_RX_FLEX_DESC_STATUS0_IPV6EXADD_S))))
- goto checksum_fail;
-
- /* check for L4 errors and handle packets that were not able to be
- * checksummed due to arrival speed
- */
- if (rx_error & BIT(ICE_RX_FLEX_DESC_STATUS0_XSUM_L4E_S))
- goto checksum_fail;
-
- /* Only report checksum unnecessary for TCP, UDP, or SCTP */
- switch (decoded.inner_prot) {
- case ICE_RX_PTYPE_INNER_PROT_TCP:
- case ICE_RX_PTYPE_INNER_PROT_UDP:
- case ICE_RX_PTYPE_INNER_PROT_SCTP:
- skb->ip_summed = CHECKSUM_UNNECESSARY;
- default:
- break;
- }
- return;
-
-checksum_fail:
- vsi->back->hw_csum_rx_error++;
-}
-
-/**
- * ice_process_skb_fields - Populate skb header fields from Rx descriptor
- * @rx_ring: rx descriptor ring packet is being transacted on
- * @rx_desc: pointer to the EOP Rx descriptor
- * @skb: pointer to current skb being populated
- * @ptype: the packet type decoded by hardware
- *
- * This function checks the ring, descriptor, and packet information in
- * order to populate the hash, checksum, VLAN, protocol, and
- * other fields within the skb.
- */
-static void ice_process_skb_fields(struct ice_ring *rx_ring,
- union ice_32b_rx_flex_desc *rx_desc,
- struct sk_buff *skb, u8 ptype)
-{
- ice_rx_hash(rx_ring, rx_desc, skb, ptype);
-
- /* modifies the skb - consumes the enet header */
- skb->protocol = eth_type_trans(skb, rx_ring->netdev);
-
- ice_rx_csum(rx_ring->vsi, skb, rx_desc, ptype);
-}
-
-/**
- * ice_receive_skb - Send a completed packet up the stack
- * @rx_ring: rx ring in play
- * @skb: packet to send up
- * @vlan_tag: vlan tag for packet
- *
- * This function sends the completed packet (via. skb) up the stack using
- * gro receive functions (with/without vlan tag)
- */
-static void ice_receive_skb(struct ice_ring *rx_ring, struct sk_buff *skb,
- u16 vlan_tag)
-{
- if ((rx_ring->netdev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
- (vlan_tag & VLAN_VID_MASK)) {
- __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vlan_tag);
- }
- napi_gro_receive(&rx_ring->q_vector->napi, skb);
-}
-
-/**
* ice_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
- * @rx_ring: rx descriptor ring to transact packets on
+ * @rx_ring: Rx descriptor ring to transact packets on
* @budget: Total limit on number of packets to process
*
* This function provides a "bounce buffer" approach to Rx interrupt
- * processing. The advantage to this is that on systems that have
+ * processing. The advantage to this is that on systems that have
* expensive overhead for IOMMU access this provides a means of avoiding
* it by maintaining the mapping of the page to the system.
*
* Returns amount of work completed
*/
-static int ice_clean_rx_irq(struct ice_ring *rx_ring, int budget)
+int ice_clean_rx_irq(struct ice_ring *rx_ring, int budget)
{
unsigned int total_rx_bytes = 0, total_rx_pkts = 0;
u16 cleaned_count = ICE_DESC_UNUSED(rx_ring);
- bool failure = false;
+ unsigned int xdp_res, xdp_xmit = 0;
+ struct bpf_prog *xdp_prog = NULL;
+ struct xdp_buff xdp;
+ bool failure;
- /* start the loop to process RX packets bounded by 'budget' */
+ xdp.rxq = &rx_ring->xdp_rxq;
+ /* Frame size depend on rx_ring setup when PAGE_SIZE=4K */
+#if (PAGE_SIZE < 8192)
+ xdp.frame_sz = ice_rx_frame_truesize(rx_ring, 0);
+#endif
+
+ /* start the loop to process Rx packets bounded by 'budget' */
while (likely(total_rx_pkts < (unsigned int)budget)) {
union ice_32b_rx_flex_desc *rx_desc;
+ struct ice_rx_buf *rx_buf;
struct sk_buff *skb;
+ unsigned int size;
u16 stat_err_bits;
+ int rx_buf_pgcnt;
u16 vlan_tag = 0;
u8 rx_ptype;
- /* return some buffers to hardware, one at a time is too slow */
- if (cleaned_count >= ICE_RX_BUF_WRITE) {
- failure = failure ||
- ice_alloc_rx_bufs(rx_ring, cleaned_count);
- cleaned_count = 0;
- }
-
- /* get the RX desc from RX ring based on 'next_to_clean' */
+ /* get the Rx desc from Rx ring based on 'next_to_clean' */
rx_desc = ICE_RX_DESC(rx_ring, rx_ring->next_to_clean);
/* status_error_len will always be zero for unused descriptors
@@ -994,11 +1132,76 @@
*/
dma_rmb();
- /* allocate (if needed) and populate skb */
- skb = ice_fetch_rx_buf(rx_ring, rx_desc);
- if (!skb)
- break;
+ if (rx_desc->wb.rxdid == FDIR_DESC_RXDID || !rx_ring->netdev) {
+ ice_put_rx_buf(rx_ring, NULL, 0);
+ cleaned_count++;
+ continue;
+ }
+ size = le16_to_cpu(rx_desc->wb.pkt_len) &
+ ICE_RX_FLX_DESC_PKT_LEN_M;
+
+ /* retrieve a buffer from the ring */
+ rx_buf = ice_get_rx_buf(rx_ring, &skb, size, &rx_buf_pgcnt);
+
+ if (!size) {
+ xdp.data = NULL;
+ xdp.data_end = NULL;
+ xdp.data_hard_start = NULL;
+ xdp.data_meta = NULL;
+ goto construct_skb;
+ }
+
+ xdp.data = page_address(rx_buf->page) + rx_buf->page_offset;
+ xdp.data_hard_start = xdp.data - ice_rx_offset(rx_ring);
+ xdp.data_meta = xdp.data;
+ xdp.data_end = xdp.data + size;
+#if (PAGE_SIZE > 4096)
+ /* At larger PAGE_SIZE, frame_sz depend on len size */
+ xdp.frame_sz = ice_rx_frame_truesize(rx_ring, size);
+#endif
+
+ rcu_read_lock();
+ xdp_prog = READ_ONCE(rx_ring->xdp_prog);
+ if (!xdp_prog) {
+ rcu_read_unlock();
+ goto construct_skb;
+ }
+
+ xdp_res = ice_run_xdp(rx_ring, &xdp, xdp_prog);
+ rcu_read_unlock();
+ if (!xdp_res)
+ goto construct_skb;
+ if (xdp_res & (ICE_XDP_TX | ICE_XDP_REDIR)) {
+ xdp_xmit |= xdp_res;
+ ice_rx_buf_adjust_pg_offset(rx_buf, xdp.frame_sz);
+ } else {
+ rx_buf->pagecnt_bias++;
+ }
+ total_rx_bytes += size;
+ total_rx_pkts++;
+
+ cleaned_count++;
+ ice_put_rx_buf(rx_ring, rx_buf, rx_buf_pgcnt);
+ continue;
+construct_skb:
+ if (skb) {
+ ice_add_rx_frag(rx_ring, rx_buf, skb, size);
+ } else if (likely(xdp.data)) {
+ if (ice_ring_uses_build_skb(rx_ring))
+ skb = ice_build_skb(rx_ring, rx_buf, &xdp);
+ else
+ skb = ice_construct_skb(rx_ring, rx_buf, &xdp);
+ }
+ /* exit if we failed to retrieve a buffer */
+ if (!skb) {
+ rx_ring->rx_stats.alloc_buf_failed++;
+ if (rx_buf)
+ rx_buf->pagecnt_bias++;
+ break;
+ }
+
+ ice_put_rx_buf(rx_ring, rx_buf, rx_buf_pgcnt);
cleaned_count++;
/* skip if it is NOP desc */
@@ -1011,17 +1214,12 @@
continue;
}
- rx_ptype = le16_to_cpu(rx_desc->wb.ptype_flex_flags0) &
- ICE_RX_FLEX_DESC_PTYPE_M;
-
stat_err_bits = BIT(ICE_RX_FLEX_DESC_STATUS0_L2TAG1P_S);
if (ice_test_staterr(rx_desc, stat_err_bits))
vlan_tag = le16_to_cpu(rx_desc->wb.l2tag1);
- /* correct empty headers and pad skb if needed (to make valid
- * ethernet frame
- */
- if (ice_cleanup_headers(skb)) {
+ /* pad the skb if needed, to make a valid ethernet frame */
+ if (eth_skb_pad(skb)) {
skb = NULL;
continue;
}
@@ -1030,6 +1228,9 @@
total_rx_bytes += skb->len;
/* populate checksum, VLAN, and protocol */
+ rx_ptype = le16_to_cpu(rx_desc->wb.ptype_flex_flags0) &
+ ICE_RX_FLEX_DESC_PTYPE_M;
+
ice_process_skb_fields(rx_ring, rx_desc, skb, rx_ptype);
/* send completed skb up the stack */
@@ -1039,16 +1240,366 @@
total_rx_pkts++;
}
- /* update queue and vector specific stats */
- u64_stats_update_begin(&rx_ring->syncp);
- rx_ring->stats.pkts += total_rx_pkts;
- rx_ring->stats.bytes += total_rx_bytes;
- u64_stats_update_end(&rx_ring->syncp);
- rx_ring->q_vector->rx.total_pkts += total_rx_pkts;
- rx_ring->q_vector->rx.total_bytes += total_rx_bytes;
+ /* return up to cleaned_count buffers to hardware */
+ failure = ice_alloc_rx_bufs(rx_ring, cleaned_count);
+
+ if (xdp_prog)
+ ice_finalize_xdp_rx(rx_ring, xdp_xmit);
+
+ ice_update_rx_ring_stats(rx_ring, total_rx_pkts, total_rx_bytes);
/* guarantee a trip back through this routine if there was a failure */
return failure ? budget : (int)total_rx_pkts;
+}
+
+/**
+ * ice_adjust_itr_by_size_and_speed - Adjust ITR based on current traffic
+ * @port_info: port_info structure containing the current link speed
+ * @avg_pkt_size: average size of Tx or Rx packets based on clean routine
+ * @itr: ITR value to update
+ *
+ * Calculate how big of an increment should be applied to the ITR value passed
+ * in based on wmem_default, SKB overhead, ethernet overhead, and the current
+ * link speed.
+ *
+ * The following is a calculation derived from:
+ * wmem_default / (size + overhead) = desired_pkts_per_int
+ * rate / bits_per_byte / (size + ethernet overhead) = pkt_rate
+ * (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value
+ *
+ * Assuming wmem_default is 212992 and overhead is 640 bytes per
+ * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the
+ * formula down to:
+ *
+ * wmem_default * bits_per_byte * usecs_per_sec pkt_size + 24
+ * ITR = -------------------------------------------- * --------------
+ * rate pkt_size + 640
+ */
+static unsigned int
+ice_adjust_itr_by_size_and_speed(struct ice_port_info *port_info,
+ unsigned int avg_pkt_size,
+ unsigned int itr)
+{
+ switch (port_info->phy.link_info.link_speed) {
+ case ICE_AQ_LINK_SPEED_100GB:
+ itr += DIV_ROUND_UP(17 * (avg_pkt_size + 24),
+ avg_pkt_size + 640);
+ break;
+ case ICE_AQ_LINK_SPEED_50GB:
+ itr += DIV_ROUND_UP(34 * (avg_pkt_size + 24),
+ avg_pkt_size + 640);
+ break;
+ case ICE_AQ_LINK_SPEED_40GB:
+ itr += DIV_ROUND_UP(43 * (avg_pkt_size + 24),
+ avg_pkt_size + 640);
+ break;
+ case ICE_AQ_LINK_SPEED_25GB:
+ itr += DIV_ROUND_UP(68 * (avg_pkt_size + 24),
+ avg_pkt_size + 640);
+ break;
+ case ICE_AQ_LINK_SPEED_20GB:
+ itr += DIV_ROUND_UP(85 * (avg_pkt_size + 24),
+ avg_pkt_size + 640);
+ break;
+ case ICE_AQ_LINK_SPEED_10GB:
+ default:
+ itr += DIV_ROUND_UP(170 * (avg_pkt_size + 24),
+ avg_pkt_size + 640);
+ break;
+ }
+
+ if ((itr & ICE_ITR_MASK) > ICE_ITR_ADAPTIVE_MAX_USECS) {
+ itr &= ICE_ITR_ADAPTIVE_LATENCY;
+ itr += ICE_ITR_ADAPTIVE_MAX_USECS;
+ }
+
+ return itr;
+}
+
+/**
+ * ice_update_itr - update the adaptive ITR value based on statistics
+ * @q_vector: structure containing interrupt and ring information
+ * @rc: structure containing ring performance data
+ *
+ * Stores a new ITR value based on packets and byte
+ * counts during the last interrupt. The advantage of per interrupt
+ * computation is faster updates and more accurate ITR for the current
+ * traffic pattern. Constants in this function were computed
+ * based on theoretical maximum wire speed and thresholds were set based
+ * on testing data as well as attempting to minimize response time
+ * while increasing bulk throughput.
+ */
+static void
+ice_update_itr(struct ice_q_vector *q_vector, struct ice_ring_container *rc)
+{
+ unsigned long next_update = jiffies;
+ unsigned int packets, bytes, itr;
+ bool container_is_rx;
+
+ if (!rc->ring || !ITR_IS_DYNAMIC(rc->itr_setting))
+ return;
+
+ /* If itr_countdown is set it means we programmed an ITR within
+ * the last 4 interrupt cycles. This has a side effect of us
+ * potentially firing an early interrupt. In order to work around
+ * this we need to throw out any data received for a few
+ * interrupts following the update.
+ */
+ if (q_vector->itr_countdown) {
+ itr = rc->target_itr;
+ goto clear_counts;
+ }
+
+ container_is_rx = (&q_vector->rx == rc);
+ /* For Rx we want to push the delay up and default to low latency.
+ * for Tx we want to pull the delay down and default to high latency.
+ */
+ itr = container_is_rx ?
+ ICE_ITR_ADAPTIVE_MIN_USECS | ICE_ITR_ADAPTIVE_LATENCY :
+ ICE_ITR_ADAPTIVE_MAX_USECS | ICE_ITR_ADAPTIVE_LATENCY;
+
+ /* If we didn't update within up to 1 - 2 jiffies we can assume
+ * that either packets are coming in so slow there hasn't been
+ * any work, or that there is so much work that NAPI is dealing
+ * with interrupt moderation and we don't need to do anything.
+ */
+ if (time_after(next_update, rc->next_update))
+ goto clear_counts;
+
+ prefetch(q_vector->vsi->port_info);
+
+ packets = rc->total_pkts;
+ bytes = rc->total_bytes;
+
+ if (container_is_rx) {
+ /* If Rx there are 1 to 4 packets and bytes are less than
+ * 9000 assume insufficient data to use bulk rate limiting
+ * approach unless Tx is already in bulk rate limiting. We
+ * are likely latency driven.
+ */
+ if (packets && packets < 4 && bytes < 9000 &&
+ (q_vector->tx.target_itr & ICE_ITR_ADAPTIVE_LATENCY)) {
+ itr = ICE_ITR_ADAPTIVE_LATENCY;
+ goto adjust_by_size_and_speed;
+ }
+ } else if (packets < 4) {
+ /* If we have Tx and Rx ITR maxed and Tx ITR is running in
+ * bulk mode and we are receiving 4 or fewer packets just
+ * reset the ITR_ADAPTIVE_LATENCY bit for latency mode so
+ * that the Rx can relax.
+ */
+ if (rc->target_itr == ICE_ITR_ADAPTIVE_MAX_USECS &&
+ (q_vector->rx.target_itr & ICE_ITR_MASK) ==
+ ICE_ITR_ADAPTIVE_MAX_USECS)
+ goto clear_counts;
+ } else if (packets > 32) {
+ /* If we have processed over 32 packets in a single interrupt
+ * for Tx assume we need to switch over to "bulk" mode.
+ */
+ rc->target_itr &= ~ICE_ITR_ADAPTIVE_LATENCY;
+ }
+
+ /* We have no packets to actually measure against. This means
+ * either one of the other queues on this vector is active or
+ * we are a Tx queue doing TSO with too high of an interrupt rate.
+ *
+ * Between 4 and 56 we can assume that our current interrupt delay
+ * is only slightly too low. As such we should increase it by a small
+ * fixed amount.
+ */
+ if (packets < 56) {
+ itr = rc->target_itr + ICE_ITR_ADAPTIVE_MIN_INC;
+ if ((itr & ICE_ITR_MASK) > ICE_ITR_ADAPTIVE_MAX_USECS) {
+ itr &= ICE_ITR_ADAPTIVE_LATENCY;
+ itr += ICE_ITR_ADAPTIVE_MAX_USECS;
+ }
+ goto clear_counts;
+ }
+
+ if (packets <= 256) {
+ itr = min(q_vector->tx.current_itr, q_vector->rx.current_itr);
+ itr &= ICE_ITR_MASK;
+
+ /* Between 56 and 112 is our "goldilocks" zone where we are
+ * working out "just right". Just report that our current
+ * ITR is good for us.
+ */
+ if (packets <= 112)
+ goto clear_counts;
+
+ /* If packet count is 128 or greater we are likely looking
+ * at a slight overrun of the delay we want. Try halving
+ * our delay to see if that will cut the number of packets
+ * in half per interrupt.
+ */
+ itr >>= 1;
+ itr &= ICE_ITR_MASK;
+ if (itr < ICE_ITR_ADAPTIVE_MIN_USECS)
+ itr = ICE_ITR_ADAPTIVE_MIN_USECS;
+
+ goto clear_counts;
+ }
+
+ /* The paths below assume we are dealing with a bulk ITR since
+ * number of packets is greater than 256. We are just going to have
+ * to compute a value and try to bring the count under control,
+ * though for smaller packet sizes there isn't much we can do as
+ * NAPI polling will likely be kicking in sooner rather than later.
+ */
+ itr = ICE_ITR_ADAPTIVE_BULK;
+
+adjust_by_size_and_speed:
+
+ /* based on checks above packets cannot be 0 so division is safe */
+ itr = ice_adjust_itr_by_size_and_speed(q_vector->vsi->port_info,
+ bytes / packets, itr);
+
+clear_counts:
+ /* write back value */
+ rc->target_itr = itr;
+
+ /* next update should occur within next jiffy */
+ rc->next_update = next_update + 1;
+
+ rc->total_bytes = 0;
+ rc->total_pkts = 0;
+}
+
+/**
+ * ice_buildreg_itr - build value for writing to the GLINT_DYN_CTL register
+ * @itr_idx: interrupt throttling index
+ * @itr: interrupt throttling value in usecs
+ */
+static u32 ice_buildreg_itr(u16 itr_idx, u16 itr)
+{
+ /* The ITR value is reported in microseconds, and the register value is
+ * recorded in 2 microsecond units. For this reason we only need to
+ * shift by the GLINT_DYN_CTL_INTERVAL_S - ICE_ITR_GRAN_S to apply this
+ * granularity as a shift instead of division. The mask makes sure the
+ * ITR value is never odd so we don't accidentally write into the field
+ * prior to the ITR field.
+ */
+ itr &= ICE_ITR_MASK;
+
+ return GLINT_DYN_CTL_INTENA_M | GLINT_DYN_CTL_CLEARPBA_M |
+ (itr_idx << GLINT_DYN_CTL_ITR_INDX_S) |
+ (itr << (GLINT_DYN_CTL_INTERVAL_S - ICE_ITR_GRAN_S));
+}
+
+/* The act of updating the ITR will cause it to immediately trigger. In order
+ * to prevent this from throwing off adaptive update statistics we defer the
+ * update so that it can only happen so often. So after either Tx or Rx are
+ * updated we make the adaptive scheme wait until either the ITR completely
+ * expires via the next_update expiration or we have been through at least
+ * 3 interrupts.
+ */
+#define ITR_COUNTDOWN_START 3
+
+/**
+ * ice_update_ena_itr - Update ITR and re-enable MSIX interrupt
+ * @q_vector: q_vector for which ITR is being updated and interrupt enabled
+ */
+static void ice_update_ena_itr(struct ice_q_vector *q_vector)
+{
+ struct ice_ring_container *tx = &q_vector->tx;
+ struct ice_ring_container *rx = &q_vector->rx;
+ struct ice_vsi *vsi = q_vector->vsi;
+ u32 itr_val;
+
+ /* when exiting WB_ON_ITR lets set a low ITR value and trigger
+ * interrupts to expire right away in case we have more work ready to go
+ * already
+ */
+ if (q_vector->itr_countdown == ICE_IN_WB_ON_ITR_MODE) {
+ itr_val = ice_buildreg_itr(rx->itr_idx, ICE_WB_ON_ITR_USECS);
+ wr32(&vsi->back->hw, GLINT_DYN_CTL(q_vector->reg_idx), itr_val);
+ /* set target back to last user set value */
+ rx->target_itr = rx->itr_setting;
+ /* set current to what we just wrote and dynamic if needed */
+ rx->current_itr = ICE_WB_ON_ITR_USECS |
+ (rx->itr_setting & ICE_ITR_DYNAMIC);
+ /* allow normal interrupt flow to start */
+ q_vector->itr_countdown = 0;
+ return;
+ }
+
+ /* This will do nothing if dynamic updates are not enabled */
+ ice_update_itr(q_vector, tx);
+ ice_update_itr(q_vector, rx);
+
+ /* This block of logic allows us to get away with only updating
+ * one ITR value with each interrupt. The idea is to perform a
+ * pseudo-lazy update with the following criteria.
+ *
+ * 1. Rx is given higher priority than Tx if both are in same state
+ * 2. If we must reduce an ITR that is given highest priority.
+ * 3. We then give priority to increasing ITR based on amount.
+ */
+ if (rx->target_itr < rx->current_itr) {
+ /* Rx ITR needs to be reduced, this is highest priority */
+ itr_val = ice_buildreg_itr(rx->itr_idx, rx->target_itr);
+ rx->current_itr = rx->target_itr;
+ q_vector->itr_countdown = ITR_COUNTDOWN_START;
+ } else if ((tx->target_itr < tx->current_itr) ||
+ ((rx->target_itr - rx->current_itr) <
+ (tx->target_itr - tx->current_itr))) {
+ /* Tx ITR needs to be reduced, this is second priority
+ * Tx ITR needs to be increased more than Rx, fourth priority
+ */
+ itr_val = ice_buildreg_itr(tx->itr_idx, tx->target_itr);
+ tx->current_itr = tx->target_itr;
+ q_vector->itr_countdown = ITR_COUNTDOWN_START;
+ } else if (rx->current_itr != rx->target_itr) {
+ /* Rx ITR needs to be increased, third priority */
+ itr_val = ice_buildreg_itr(rx->itr_idx, rx->target_itr);
+ rx->current_itr = rx->target_itr;
+ q_vector->itr_countdown = ITR_COUNTDOWN_START;
+ } else {
+ /* Still have to re-enable the interrupts */
+ itr_val = ice_buildreg_itr(ICE_ITR_NONE, 0);
+ if (q_vector->itr_countdown)
+ q_vector->itr_countdown--;
+ }
+
+ if (!test_bit(__ICE_DOWN, q_vector->vsi->state))
+ wr32(&q_vector->vsi->back->hw,
+ GLINT_DYN_CTL(q_vector->reg_idx),
+ itr_val);
+}
+
+/**
+ * ice_set_wb_on_itr - set WB_ON_ITR for this q_vector
+ * @q_vector: q_vector to set WB_ON_ITR on
+ *
+ * We need to tell hardware to write-back completed descriptors even when
+ * interrupts are disabled. Descriptors will be written back on cache line
+ * boundaries without WB_ON_ITR enabled, but if we don't enable WB_ON_ITR
+ * descriptors may not be written back if they don't fill a cache line until the
+ * next interrupt.
+ *
+ * This sets the write-back frequency to 2 microseconds as that is the minimum
+ * value that's not 0 due to ITR granularity. Also, set the INTENA_MSK bit to
+ * make sure hardware knows we aren't meddling with the INTENA_M bit.
+ */
+static void ice_set_wb_on_itr(struct ice_q_vector *q_vector)
+{
+ struct ice_vsi *vsi = q_vector->vsi;
+
+ /* already in WB_ON_ITR mode no need to change it */
+ if (q_vector->itr_countdown == ICE_IN_WB_ON_ITR_MODE)
+ return;
+
+ if (q_vector->num_ring_rx)
+ wr32(&vsi->back->hw, GLINT_DYN_CTL(q_vector->reg_idx),
+ ICE_GLINT_DYN_CTL_WB_ON_ITR(ICE_WB_ON_ITR_USECS,
+ ICE_RX_ITR));
+
+ if (q_vector->num_ring_tx)
+ wr32(&vsi->back->hw, GLINT_DYN_CTL(q_vector->reg_idx),
+ ICE_GLINT_DYN_CTL_WB_ON_ITR(ICE_WB_ON_ITR_USECS,
+ ICE_TX_ITR));
+
+ q_vector->itr_countdown = ICE_IN_WB_ON_ITR_MODE;
}
/**
@@ -1064,34 +1615,48 @@
{
struct ice_q_vector *q_vector =
container_of(napi, struct ice_q_vector, napi);
- struct ice_vsi *vsi = q_vector->vsi;
- struct ice_pf *pf = vsi->back;
bool clean_complete = true;
- int budget_per_ring = 0;
struct ice_ring *ring;
+ int budget_per_ring;
int work_done = 0;
/* Since the actual Tx work is minimal, we can give the Tx a larger
* budget and be more aggressive about cleaning up the Tx descriptors.
*/
- ice_for_each_ring(ring, q_vector->tx)
- if (!ice_clean_tx_irq(vsi, ring, budget))
+ ice_for_each_ring(ring, q_vector->tx) {
+ bool wd = ring->xsk_pool ?
+ ice_clean_tx_irq_zc(ring, budget) :
+ ice_clean_tx_irq(ring, budget);
+
+ if (!wd)
clean_complete = false;
+ }
/* Handle case where we are called by netpoll with a budget of 0 */
- if (budget <= 0)
+ if (unlikely(budget <= 0))
return budget;
- /* We attempt to distribute budget to each Rx queue fairly, but don't
- * allow the budget to go below 1 because that would exit polling early.
- */
- if (q_vector->num_ring_rx)
- budget_per_ring = max(budget / q_vector->num_ring_rx, 1);
+ /* normally we have 1 Rx ring per q_vector */
+ if (unlikely(q_vector->num_ring_rx > 1))
+ /* We attempt to distribute budget to each Rx queue fairly, but
+ * don't allow the budget to go below 1 because that would exit
+ * polling early.
+ */
+ budget_per_ring = max_t(int, budget / q_vector->num_ring_rx, 1);
+ else
+ /* Max of 1 Rx ring in this q_vector so give it the budget */
+ budget_per_ring = budget;
ice_for_each_ring(ring, q_vector->rx) {
int cleaned;
- cleaned = ice_clean_rx_irq(ring, budget_per_ring);
+ /* A dedicated path for zero-copy allows making a single
+ * comparison in the irq context instead of many inside the
+ * ice_clean_rx_irq function and makes the codebase cleaner.
+ */
+ cleaned = ring->xsk_pool ?
+ ice_clean_rx_irq_zc(ring, budget_per_ring) :
+ ice_clean_rx_irq(ring, budget_per_ring);
work_done += cleaned;
/* if we clean as many as budgeted, we must not be done */
if (cleaned >= budget_per_ring)
@@ -1102,27 +1667,19 @@
if (!clean_complete)
return budget;
- /* Work is done so exit the polling mode and re-enable the interrupt */
- napi_complete_done(napi, work_done);
- if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
- ice_irq_dynamic_ena(&vsi->back->hw, vsi, q_vector);
+ /* Exit the polling mode, but don't re-enable interrupts if stack might
+ * poll us due to busy-polling
+ */
+ if (likely(napi_complete_done(napi, work_done)))
+ ice_update_ena_itr(q_vector);
+ else
+ ice_set_wb_on_itr(q_vector);
- return min(work_done, budget - 1);
-}
-
-/* helper function for building cmd/type/offset */
-static __le64
-build_ctob(u64 td_cmd, u64 td_offset, unsigned int size, u64 td_tag)
-{
- return cpu_to_le64(ICE_TX_DESC_DTYPE_DATA |
- (td_cmd << ICE_TXD_QW1_CMD_S) |
- (td_offset << ICE_TXD_QW1_OFFSET_S) |
- ((u64)size << ICE_TXD_QW1_TX_BUF_SZ_S) |
- (td_tag << ICE_TXD_QW1_L2TAG1_S));
+ return min_t(int, work_done, budget - 1);
}
/**
- * __ice_maybe_stop_tx - 2nd level check for tx stop conditions
+ * __ice_maybe_stop_tx - 2nd level check for Tx stop conditions
* @tx_ring: the ring to be checked
* @size: the size buffer we want to assure is available
*
@@ -1145,7 +1702,7 @@
}
/**
- * ice_maybe_stop_tx - 1st level check for tx stop conditions
+ * ice_maybe_stop_tx - 1st level check for Tx stop conditions
* @tx_ring: the ring to be checked
* @size: the size buffer we want to assure is available
*
@@ -1155,6 +1712,7 @@
{
if (likely(ICE_DESC_UNUSED(tx_ring) >= size))
return 0;
+
return __ice_maybe_stop_tx(tx_ring, size);
}
@@ -1174,11 +1732,11 @@
{
u64 td_offset, td_tag, td_cmd;
u16 i = tx_ring->next_to_use;
- struct skb_frag_struct *frag;
unsigned int data_len, size;
struct ice_tx_desc *tx_desc;
struct ice_tx_buf *tx_buf;
struct sk_buff *skb;
+ skb_frag_t *frag;
dma_addr_t dma;
td_tag = off->td_l2tag1;
@@ -1220,7 +1778,8 @@
*/
while (unlikely(size > ICE_MAX_DATA_PER_TXD)) {
tx_desc->cmd_type_offset_bsz =
- build_ctob(td_cmd, td_offset, max_data, td_tag);
+ ice_build_ctob(td_cmd, td_offset, max_data,
+ td_tag);
tx_desc++;
i++;
@@ -1240,8 +1799,8 @@
if (likely(!data_len))
break;
- tx_desc->cmd_type_offset_bsz = build_ctob(td_cmd, td_offset,
- size, td_tag);
+ tx_desc->cmd_type_offset_bsz = ice_build_ctob(td_cmd, td_offset,
+ size, td_tag);
tx_desc++;
i++;
@@ -1271,9 +1830,9 @@
i = 0;
/* write last descriptor with RS and EOP bits */
- td_cmd |= (u64)(ICE_TX_DESC_CMD_EOP | ICE_TX_DESC_CMD_RS);
+ td_cmd |= (u64)ICE_TXD_LAST_DESC_CMD;
tx_desc->cmd_type_offset_bsz =
- build_ctob(td_cmd, td_offset, size, td_tag);
+ ice_build_ctob(td_cmd, td_offset, size, td_tag);
/* Force memory writes to complete before letting h/w know there
* are new descriptors to fetch.
@@ -1291,19 +1850,13 @@
ice_maybe_stop_tx(tx_ring, DESC_NEEDED);
/* notify HW of packet */
- if (netif_xmit_stopped(txring_txq(tx_ring)) || !skb->xmit_more) {
+ if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more())
writel(i, tx_ring->tail);
-
- /* we need this if more than one processor can write to our tail
- * at a time, it synchronizes IO on IA64/Altix systems
- */
- mmiowb();
- }
return;
dma_error:
- /* clear dma mappings for failed tx_buf map */
+ /* clear DMA mappings for failed tx_buf map */
for (;;) {
tx_buf = &tx_ring->tx_buf[i];
ice_unmap_and_free_tx_buf(tx_ring, tx_buf);
@@ -1353,12 +1906,97 @@
l2_len = ip.hdr - skb->data;
offset = (l2_len / 2) << ICE_TX_DESC_LEN_MACLEN_S;
- if (skb->encapsulation)
- return -1;
+ protocol = vlan_get_protocol(skb);
+
+ if (protocol == htons(ETH_P_IP))
+ first->tx_flags |= ICE_TX_FLAGS_IPV4;
+ else if (protocol == htons(ETH_P_IPV6))
+ first->tx_flags |= ICE_TX_FLAGS_IPV6;
+
+ if (skb->encapsulation) {
+ bool gso_ena = false;
+ u32 tunnel = 0;
+
+ /* define outer network header type */
+ if (first->tx_flags & ICE_TX_FLAGS_IPV4) {
+ tunnel |= (first->tx_flags & ICE_TX_FLAGS_TSO) ?
+ ICE_TX_CTX_EIPT_IPV4 :
+ ICE_TX_CTX_EIPT_IPV4_NO_CSUM;
+ l4_proto = ip.v4->protocol;
+ } else if (first->tx_flags & ICE_TX_FLAGS_IPV6) {
+ int ret;
+
+ tunnel |= ICE_TX_CTX_EIPT_IPV6;
+ exthdr = ip.hdr + sizeof(*ip.v6);
+ l4_proto = ip.v6->nexthdr;
+ ret = ipv6_skip_exthdr(skb, exthdr - skb->data,
+ &l4_proto, &frag_off);
+ if (ret < 0)
+ return -1;
+ }
+
+ /* define outer transport */
+ switch (l4_proto) {
+ case IPPROTO_UDP:
+ tunnel |= ICE_TXD_CTX_UDP_TUNNELING;
+ first->tx_flags |= ICE_TX_FLAGS_TUNNEL;
+ break;
+ case IPPROTO_GRE:
+ tunnel |= ICE_TXD_CTX_GRE_TUNNELING;
+ first->tx_flags |= ICE_TX_FLAGS_TUNNEL;
+ break;
+ case IPPROTO_IPIP:
+ case IPPROTO_IPV6:
+ first->tx_flags |= ICE_TX_FLAGS_TUNNEL;
+ l4.hdr = skb_inner_network_header(skb);
+ break;
+ default:
+ if (first->tx_flags & ICE_TX_FLAGS_TSO)
+ return -1;
+
+ skb_checksum_help(skb);
+ return 0;
+ }
+
+ /* compute outer L3 header size */
+ tunnel |= ((l4.hdr - ip.hdr) / 4) <<
+ ICE_TXD_CTX_QW0_EIPLEN_S;
+
+ /* switch IP header pointer from outer to inner header */
+ ip.hdr = skb_inner_network_header(skb);
+
+ /* compute tunnel header size */
+ tunnel |= ((ip.hdr - l4.hdr) / 2) <<
+ ICE_TXD_CTX_QW0_NATLEN_S;
+
+ gso_ena = skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL;
+ /* indicate if we need to offload outer UDP header */
+ if ((first->tx_flags & ICE_TX_FLAGS_TSO) && !gso_ena &&
+ (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL_CSUM))
+ tunnel |= ICE_TXD_CTX_QW0_L4T_CS_M;
+
+ /* record tunnel offload values */
+ off->cd_tunnel_params |= tunnel;
+
+ /* set DTYP=1 to indicate that it's an Tx context descriptor
+ * in IPsec tunnel mode with Tx offloads in Quad word 1
+ */
+ off->cd_qw1 |= (u64)ICE_TX_DESC_DTYPE_CTX;
+
+ /* switch L4 header pointer from outer to inner */
+ l4.hdr = skb_inner_transport_header(skb);
+ l4_proto = 0;
+
+ /* reset type as we transition from outer to inner headers */
+ first->tx_flags &= ~(ICE_TX_FLAGS_IPV4 | ICE_TX_FLAGS_IPV6);
+ if (ip.v4->version == 4)
+ first->tx_flags |= ICE_TX_FLAGS_IPV4;
+ if (ip.v6->version == 6)
+ first->tx_flags |= ICE_TX_FLAGS_IPV6;
+ }
/* Enable IP checksum offloads */
- protocol = vlan_get_protocol(skb);
- if (protocol == htons(ETH_P_IP)) {
+ if (first->tx_flags & ICE_TX_FLAGS_IPV4) {
l4_proto = ip.v4->protocol;
/* the stack computes the IP header already, the only time we
* need the hardware to recompute it is in the case of TSO.
@@ -1368,7 +2006,7 @@
else
cmd |= ICE_TX_DESC_CMD_IIPT_IPV4;
- } else if (protocol == htons(ETH_P_IPV6)) {
+ } else if (first->tx_flags & ICE_TX_FLAGS_IPV6) {
cmd |= ICE_TX_DESC_CMD_IIPT_IPV6;
exthdr = ip.hdr + sizeof(*ip.v6);
l4_proto = ip.v6->nexthdr;
@@ -1398,6 +2036,12 @@
offset |= l4_len << ICE_TX_DESC_LEN_L4_LEN_S;
break;
case IPPROTO_SCTP:
+ /* enable SCTP checksum offload */
+ cmd |= ICE_TX_DESC_CMD_L4T_EOFT_SCTP;
+ l4_len = sizeof(struct sctphdr) >> 2;
+ offset |= l4_len << ICE_TX_DESC_LEN_L4_LEN_S;
+ break;
+
default:
if (first->tx_flags & ICE_TX_FLAGS_TSO)
return -1;
@@ -1411,56 +2055,31 @@
}
/**
- * ice_tx_prepare_vlan_flags - prepare generic TX VLAN tagging flags for HW
+ * ice_tx_prepare_vlan_flags - prepare generic Tx VLAN tagging flags for HW
* @tx_ring: ring to send buffer on
* @first: pointer to struct ice_tx_buf
*
* Checks the skb and set up correspondingly several generic transmit flags
* related to VLAN tagging for the HW, such as VLAN, DCB, etc.
- *
- * Returns error code indicate the frame should be dropped upon error and the
- * otherwise returns 0 to indicate the flags has been set properly.
*/
-static int
+static void
ice_tx_prepare_vlan_flags(struct ice_ring *tx_ring, struct ice_tx_buf *first)
{
struct sk_buff *skb = first->skb;
- __be16 protocol = skb->protocol;
- if (protocol == htons(ETH_P_8021Q) &&
- !(tx_ring->netdev->features & NETIF_F_HW_VLAN_CTAG_TX)) {
- /* when HW VLAN acceleration is turned off by the user the
- * stack sets the protocol to 8021q so that the driver
- * can take any steps required to support the SW only
- * VLAN handling. In our case the driver doesn't need
- * to take any further steps so just set the protocol
- * to the encapsulated ethertype.
- */
- skb->protocol = vlan_get_protocol(skb);
- goto out;
- }
+ /* nothing left to do, software offloaded VLAN */
+ if (!skb_vlan_tag_present(skb) && eth_type_vlan(skb->protocol))
+ return;
- /* if we have a HW VLAN tag being added, default to the HW one */
+ /* currently, we always assume 802.1Q for VLAN insertion as VLAN
+ * insertion for 802.1AD is not supported
+ */
if (skb_vlan_tag_present(skb)) {
first->tx_flags |= skb_vlan_tag_get(skb) << ICE_TX_FLAGS_VLAN_S;
first->tx_flags |= ICE_TX_FLAGS_HW_VLAN;
- } else if (protocol == htons(ETH_P_8021Q)) {
- struct vlan_hdr *vhdr, _vhdr;
-
- /* for SW VLAN, check the next protocol and store the tag */
- vhdr = (struct vlan_hdr *)skb_header_pointer(skb, ETH_HLEN,
- sizeof(_vhdr),
- &_vhdr);
- if (!vhdr)
- return -EINVAL;
-
- first->tx_flags |= ntohs(vhdr->h_vlan_TCI) <<
- ICE_TX_FLAGS_VLAN_S;
- first->tx_flags |= ICE_TX_FLAGS_SW_VLAN;
}
-out:
- return 0;
+ ice_tx_prepare_vlan_flags_dcb(tx_ring, first);
}
/**
@@ -1481,10 +2100,12 @@
} ip;
union {
struct tcphdr *tcp;
+ struct udphdr *udp;
unsigned char *hdr;
} l4;
u64 cd_mss, cd_tso_len;
- u32 paylen, l4_start;
+ u32 paylen;
+ u8 l4_start;
int err;
if (skb->ip_summed != CHECKSUM_PARTIAL)
@@ -1497,6 +2118,7 @@
if (err < 0)
return err;
+ /* cppcheck-suppress unreadVariable */
ip.hdr = skb_network_header(skb);
l4.hdr = skb_transport_header(skb);
@@ -1508,15 +2130,57 @@
ip.v6->payload_len = 0;
}
+ if (skb_shinfo(skb)->gso_type & (SKB_GSO_GRE |
+ SKB_GSO_GRE_CSUM |
+ SKB_GSO_IPXIP4 |
+ SKB_GSO_IPXIP6 |
+ SKB_GSO_UDP_TUNNEL |
+ SKB_GSO_UDP_TUNNEL_CSUM)) {
+ if (!(skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL) &&
+ (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL_CSUM)) {
+ l4.udp->len = 0;
+
+ /* determine offset of outer transport header */
+ l4_start = (u8)(l4.hdr - skb->data);
+
+ /* remove payload length from outer checksum */
+ paylen = skb->len - l4_start;
+ csum_replace_by_diff(&l4.udp->check,
+ (__force __wsum)htonl(paylen));
+ }
+
+ /* reset pointers to inner headers */
+
+ /* cppcheck-suppress unreadVariable */
+ ip.hdr = skb_inner_network_header(skb);
+ l4.hdr = skb_inner_transport_header(skb);
+
+ /* initialize inner IP header fields */
+ if (ip.v4->version == 4) {
+ ip.v4->tot_len = 0;
+ ip.v4->check = 0;
+ } else {
+ ip.v6->payload_len = 0;
+ }
+ }
+
/* determine offset of transport header */
- l4_start = l4.hdr - skb->data;
+ l4_start = (u8)(l4.hdr - skb->data);
/* remove payload length from checksum */
paylen = skb->len - l4_start;
- csum_replace_by_diff(&l4.tcp->check, (__force __wsum)htonl(paylen));
- /* compute length of segmentation header */
- off->header_len = (l4.tcp->doff * 4) + l4_start;
+ if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4) {
+ csum_replace_by_diff(&l4.udp->check,
+ (__force __wsum)htonl(paylen));
+ /* compute length of UDP segmentation header */
+ off->header_len = (u8)sizeof(l4.udp) + l4_start;
+ } else {
+ csum_replace_by_diff(&l4.tcp->check,
+ (__force __wsum)htonl(paylen));
+ /* compute length of TCP segmentation header */
+ off->header_len = (u8)((l4.tcp->doff * 4) + l4_start);
+ }
/* update gso_segs and bytecount */
first->gso_segs = skb_shinfo(skb)->gso_segs;
@@ -1526,10 +2190,10 @@
cd_mss = skb_shinfo(skb)->gso_size;
/* record cdesc_qw1 with TSO parameters */
- off->cd_qw1 |= ICE_TX_DESC_DTYPE_CTX |
- (ICE_TX_CTX_DESC_TSO << ICE_TXD_CTX_QW1_CMD_S) |
- (cd_tso_len << ICE_TXD_CTX_QW1_TSO_LEN_S) |
- (cd_mss << ICE_TXD_CTX_QW1_MSS_S);
+ off->cd_qw1 |= (u64)(ICE_TX_DESC_DTYPE_CTX |
+ (ICE_TX_CTX_DESC_TSO << ICE_TXD_CTX_QW1_CMD_S) |
+ (cd_tso_len << ICE_TXD_CTX_QW1_TSO_LEN_S) |
+ (cd_mss << ICE_TXD_CTX_QW1_MSS_S));
first->tx_flags |= ICE_TX_FLAGS_TSO;
return 1;
}
@@ -1552,30 +2216,30 @@
* Finally, we add one to round up. Because 256 isn't an exact multiple of
* 3, we'll underestimate near each multiple of 12K. This is actually more
* accurate as we have 4K - 1 of wiggle room that we can fit into the last
- * segment. For our purposes this is accurate out to 1M which is orders of
+ * segment. For our purposes this is accurate out to 1M which is orders of
* magnitude greater than our largest possible GSO size.
*
* This would then be implemented as:
- * return (((size >> 12) * 85) >> 8) + 1;
+ * return (((size >> 12) * 85) >> 8) + ICE_DESCS_FOR_SKB_DATA_PTR;
*
* Since multiplication and division are commutative, we can reorder
* operations into:
- * return ((size * 85) >> 20) + 1;
+ * return ((size * 85) >> 20) + ICE_DESCS_FOR_SKB_DATA_PTR;
*/
static unsigned int ice_txd_use_count(unsigned int size)
{
- return ((size * 85) >> 20) + 1;
+ return ((size * 85) >> 20) + ICE_DESCS_FOR_SKB_DATA_PTR;
}
/**
- * ice_xmit_desc_count - calculate number of tx descriptors needed
+ * ice_xmit_desc_count - calculate number of Tx descriptors needed
* @skb: send buffer
*
* Returns number of data descriptors needed for this skb.
*/
static unsigned int ice_xmit_desc_count(struct sk_buff *skb)
{
- const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
+ const skb_frag_t *frag = &skb_shinfo(skb)->frags[0];
unsigned int nr_frags = skb_shinfo(skb)->nr_frags;
unsigned int count = 0, size = skb_headlen(skb);
@@ -1606,7 +2270,7 @@
*/
static bool __ice_chk_linearize(struct sk_buff *skb)
{
- const struct skb_frag_struct *frag, *stale;
+ const skb_frag_t *frag, *stale;
int nr_frags, sum;
/* no need to check if number of frags is less than 7 */
@@ -1620,8 +2284,8 @@
nr_frags -= ICE_MAX_BUF_TXD - 2;
frag = &skb_shinfo(skb)->frags[0];
- /* Initialize size to the negative value of gso_size minus 1. We
- * use this as the worst case scenerio in which the frag ahead
+ /* Initialize size to the negative value of gso_size minus 1. We
+ * use this as the worst case scenario in which the frag ahead
* of us only provides one byte which is why we are limited to 6
* descriptors for a single transmit as the header and previous
* fragment are already consuming 2 descriptors.
@@ -1638,9 +2302,29 @@
/* Walk through fragments adding latest fragment, testing it, and
* then removing stale fragments from the sum.
*/
- stale = &skb_shinfo(skb)->frags[0];
- for (;;) {
+ for (stale = &skb_shinfo(skb)->frags[0];; stale++) {
+ int stale_size = skb_frag_size(stale);
+
sum += skb_frag_size(frag++);
+
+ /* The stale fragment may present us with a smaller
+ * descriptor than the actual fragment size. To account
+ * for that we need to remove all the data on the front and
+ * figure out what the remainder would be in the last
+ * descriptor associated with the fragment.
+ */
+ if (stale_size > ICE_MAX_DATA_PER_TXD) {
+ int align_pad = -(skb_frag_off(stale)) &
+ (ICE_MAX_READ_REQ_SIZE - 1);
+
+ sum -= align_pad;
+ stale_size -= align_pad;
+
+ do {
+ sum -= ICE_MAX_DATA_PER_TXD_ALIGNED;
+ stale_size -= ICE_MAX_DATA_PER_TXD_ALIGNED;
+ } while (stale_size > ICE_MAX_DATA_PER_TXD);
+ }
/* if sum is negative we failed to make sufficient progress */
if (sum < 0)
@@ -1649,7 +2333,7 @@
if (!nr_frags--)
break;
- sum -= skb_frag_size(stale++);
+ sum -= stale_size;
}
return false;
@@ -1688,7 +2372,9 @@
ice_xmit_frame_ring(struct sk_buff *skb, struct ice_ring *tx_ring)
{
struct ice_tx_offload_params offload = { 0 };
+ struct ice_vsi *vsi = tx_ring->vsi;
struct ice_tx_buf *first;
+ struct ethhdr *eth;
unsigned int count;
int tso, csum;
@@ -1706,7 +2392,8 @@
* + 1 desc for context descriptor,
* otherwise try next time
*/
- if (ice_maybe_stop_tx(tx_ring, count + 4 + 1)) {
+ if (ice_maybe_stop_tx(tx_ring, count + ICE_DESCS_PER_CACHE_LINE +
+ ICE_DESCS_FOR_CTX_DESC)) {
tx_ring->tx_stats.tx_busy++;
return NETDEV_TX_BUSY;
}
@@ -1721,8 +2408,7 @@
first->tx_flags = 0;
/* prepare the VLAN tagging flags for Tx */
- if (ice_tx_prepare_vlan_flags(tx_ring, first))
- goto out_drop;
+ ice_tx_prepare_vlan_flags(tx_ring, first);
/* set up TSO offload */
tso = ice_tso(first, &offload);
@@ -1734,9 +2420,19 @@
if (csum < 0)
goto out_drop;
- if (tso || offload.cd_tunnel_params) {
+ /* allow CONTROL frames egress from main VSI if FW LLDP disabled */
+ eth = (struct ethhdr *)skb_mac_header(skb);
+ if (unlikely((skb->priority == TC_PRIO_CONTROL ||
+ eth->h_proto == htons(ETH_P_LLDP)) &&
+ vsi->type == ICE_VSI_PF &&
+ vsi->port_info->qos_cfg.is_sw_lldp))
+ offload.cd_qw1 |= (u64)(ICE_TX_DESC_DTYPE_CTX |
+ ICE_TX_CTX_DESC_SWTCH_UPLINK <<
+ ICE_TXD_CTX_QW1_CMD_S);
+
+ if (offload.cd_qw1 & ICE_TX_DESC_DTYPE_CTX) {
struct ice_tx_ctx_desc *cdesc;
- int i = tx_ring->next_to_use;
+ u16 i = tx_ring->next_to_use;
/* grab the next descriptor */
cdesc = ICE_TX_CTX_DESC(tx_ring, i);
@@ -1781,3 +2477,86 @@
return ice_xmit_frame_ring(skb, tx_ring);
}
+
+/**
+ * ice_clean_ctrl_tx_irq - interrupt handler for flow director Tx queue
+ * @tx_ring: tx_ring to clean
+ */
+void ice_clean_ctrl_tx_irq(struct ice_ring *tx_ring)
+{
+ struct ice_vsi *vsi = tx_ring->vsi;
+ s16 i = tx_ring->next_to_clean;
+ int budget = ICE_DFLT_IRQ_WORK;
+ struct ice_tx_desc *tx_desc;
+ struct ice_tx_buf *tx_buf;
+
+ tx_buf = &tx_ring->tx_buf[i];
+ tx_desc = ICE_TX_DESC(tx_ring, i);
+ i -= tx_ring->count;
+
+ do {
+ struct ice_tx_desc *eop_desc = tx_buf->next_to_watch;
+
+ /* if next_to_watch is not set then there is no pending work */
+ if (!eop_desc)
+ break;
+
+ /* prevent any other reads prior to eop_desc */
+ smp_rmb();
+
+ /* if the descriptor isn't done, no work to do */
+ if (!(eop_desc->cmd_type_offset_bsz &
+ cpu_to_le64(ICE_TX_DESC_DTYPE_DESC_DONE)))
+ break;
+
+ /* clear next_to_watch to prevent false hangs */
+ tx_buf->next_to_watch = NULL;
+ tx_desc->buf_addr = 0;
+ tx_desc->cmd_type_offset_bsz = 0;
+
+ /* move past filter desc */
+ tx_buf++;
+ tx_desc++;
+ i++;
+ if (unlikely(!i)) {
+ i -= tx_ring->count;
+ tx_buf = tx_ring->tx_buf;
+ tx_desc = ICE_TX_DESC(tx_ring, 0);
+ }
+
+ /* unmap the data header */
+ if (dma_unmap_len(tx_buf, len))
+ dma_unmap_single(tx_ring->dev,
+ dma_unmap_addr(tx_buf, dma),
+ dma_unmap_len(tx_buf, len),
+ DMA_TO_DEVICE);
+ if (tx_buf->tx_flags & ICE_TX_FLAGS_DUMMY_PKT)
+ devm_kfree(tx_ring->dev, tx_buf->raw_buf);
+
+ /* clear next_to_watch to prevent false hangs */
+ tx_buf->raw_buf = NULL;
+ tx_buf->tx_flags = 0;
+ tx_buf->next_to_watch = NULL;
+ dma_unmap_len_set(tx_buf, len, 0);
+ tx_desc->buf_addr = 0;
+ tx_desc->cmd_type_offset_bsz = 0;
+
+ /* move past eop_desc for start of next FD desc */
+ tx_buf++;
+ tx_desc++;
+ i++;
+ if (unlikely(!i)) {
+ i -= tx_ring->count;
+ tx_buf = tx_ring->tx_buf;
+ tx_desc = ICE_TX_DESC(tx_ring, 0);
+ }
+
+ budget--;
+ } while (likely(budget));
+
+ i += tx_ring->count;
+ tx_ring->next_to_clean = i;
+
+ /* re-enable interrupt if needed */
+ ice_irq_dynamic_ena(&vsi->back->hw, vsi, vsi->q_vectors[0]);
+}
--
Gitblit v1.6.2