#include <linux/dma-mapping.h>
#include <linux/dmapool.h>
+#include <linux/ethtool_netlink.h>
#include <linux/netdevice.h>
#include <linux/pci.h>
#include <linux/u64_stats_sync.h>
u32 mask; /* Mask for indices to the size of the ring */
};
+struct gve_header_buf {
+ u8 *data;
+ dma_addr_t addr;
+};
+
/* Stores state for tracking buffers posted to HW */
struct gve_rx_buf_state_dqo {
/* The page posted to HW. */
/* track number of used buffers */
u16 used_buf_states_cnt;
+
+ /* Address info of the buffers for header-split */
+ struct gve_header_buf hdr_bufs;
} dqo;
};
struct gve_qpl_config *qpl_cfg;
u16 ring_size;
+ u16 packet_buffer_size;
bool raw_addressing;
bool enable_header_split;
u32 rx_coalesce_usecs;
u16 header_buf_size; /* device configured, header-split supported if non-zero */
+ bool header_split_enabled; /* True if the header split is enabled by the user */
};
enum gve_service_task_flags_bit {
struct gve_rx_alloc_rings_cfg *cfg);
void gve_rx_start_ring_gqi(struct gve_priv *priv, int idx);
void gve_rx_stop_ring_gqi(struct gve_priv *priv, int idx);
+u16 gve_get_pkt_buf_size(const struct gve_priv *priv, bool enable_hplit);
+bool gve_header_split_supported(const struct gve_priv *priv);
+int gve_set_hsplit_config(struct gve_priv *priv, u8 tcp_data_split);
/* Reset */
void gve_schedule_reset(struct gve_priv *priv);
int gve_reset(struct gve_priv *priv, bool attempt_teardown);
cpu_to_be16(rx_buff_ring_entries);
cmd.create_rx_queue.enable_rsc =
!!(priv->dev->features & NETIF_F_LRO);
+ if (priv->header_split_enabled)
+ cmd.create_rx_queue.header_buffer_size =
+ cpu_to_be16(priv->header_buf_size);
}
return gve_adminq_issue_cmd(priv, &cmd);
* Copyright (C) 2015-2021 Google, Inc.
*/
-#include <linux/ethtool.h>
#include <linux/rtnetlink.h>
#include "gve.h"
#include "gve_adminq.h"
cmd->tx_max_pending = priv->tx_desc_cnt;
cmd->rx_pending = priv->rx_desc_cnt;
cmd->tx_pending = priv->tx_desc_cnt;
+
+ if (!gve_header_split_supported(priv))
+ kernel_cmd->tcp_data_split = ETHTOOL_TCP_DATA_SPLIT_UNKNOWN;
+ else if (priv->header_split_enabled)
+ kernel_cmd->tcp_data_split = ETHTOOL_TCP_DATA_SPLIT_ENABLED;
+ else
+ kernel_cmd->tcp_data_split = ETHTOOL_TCP_DATA_SPLIT_DISABLED;
+}
+
+static int gve_set_ringparam(struct net_device *netdev,
+ struct ethtool_ringparam *cmd,
+ struct kernel_ethtool_ringparam *kernel_cmd,
+ struct netlink_ext_ack *extack)
+{
+ struct gve_priv *priv = netdev_priv(netdev);
+
+ if (priv->tx_desc_cnt != cmd->tx_pending ||
+ priv->rx_desc_cnt != cmd->rx_pending) {
+ dev_info(&priv->pdev->dev, "Modify ring size is not supported.\n");
+ return -EOPNOTSUPP;
+ }
+
+ return gve_set_hsplit_config(priv, kernel_cmd->tcp_data_split);
}
static int gve_user_reset(struct net_device *netdev, u32 *flags)
const struct ethtool_ops gve_ethtool_ops = {
.supported_coalesce_params = ETHTOOL_COALESCE_USECS,
+ .supported_ring_params = ETHTOOL_RING_USE_TCP_DATA_SPLIT,
.get_drvinfo = gve_get_drvinfo,
.get_strings = gve_get_strings,
.get_sset_count = gve_get_sset_count,
.get_coalesce = gve_get_coalesce,
.set_coalesce = gve_set_coalesce,
.get_ringparam = gve_get_ringparam,
+ .set_ringparam = gve_set_ringparam,
.reset = gve_user_reset,
.get_tunable = gve_get_tunable,
.set_tunable = gve_set_tunable,
cfg->qcfg = &priv->rx_cfg;
cfg->qcfg_tx = &priv->tx_cfg;
cfg->raw_addressing = !gve_is_qpl(priv);
+ cfg->enable_header_split = priv->header_split_enabled;
cfg->qpls = priv->qpls;
cfg->qpl_cfg = &priv->qpl_cfg;
cfg->ring_size = priv->rx_desc_cnt;
+ cfg->packet_buffer_size = gve_is_gqi(priv) ?
+ GVE_DEFAULT_RX_BUFFER_SIZE :
+ priv->data_buffer_size_dqo;
cfg->rx = priv->rx;
}
if (err)
goto reset;
+ priv->header_split_enabled = rx_alloc_cfg->enable_header_split;
+ priv->data_buffer_size_dqo = rx_alloc_cfg->packet_buffer_size;
+
err = gve_create_rings(priv);
if (err)
goto reset;
priv->tx_timeo_cnt++;
}
+u16 gve_get_pkt_buf_size(const struct gve_priv *priv, bool enable_hsplit)
+{
+ if (enable_hsplit && priv->max_rx_buffer_size >= GVE_MAX_RX_BUFFER_SIZE)
+ return GVE_MAX_RX_BUFFER_SIZE;
+ else
+ return GVE_DEFAULT_RX_BUFFER_SIZE;
+}
+
+/* header-split is not supported on non-DQO_RDA yet even if device advertises it */
+bool gve_header_split_supported(const struct gve_priv *priv)
+{
+ return priv->header_buf_size && priv->queue_format == GVE_DQO_RDA_FORMAT;
+}
+
+int gve_set_hsplit_config(struct gve_priv *priv, u8 tcp_data_split)
+{
+ struct gve_tx_alloc_rings_cfg tx_alloc_cfg = {0};
+ struct gve_rx_alloc_rings_cfg rx_alloc_cfg = {0};
+ struct gve_qpls_alloc_cfg qpls_alloc_cfg = {0};
+ bool enable_hdr_split;
+ int err = 0;
+
+ if (tcp_data_split == ETHTOOL_TCP_DATA_SPLIT_UNKNOWN)
+ return 0;
+
+ if (!gve_header_split_supported(priv)) {
+ dev_err(&priv->pdev->dev, "Header-split not supported\n");
+ return -EOPNOTSUPP;
+ }
+
+ if (tcp_data_split == ETHTOOL_TCP_DATA_SPLIT_ENABLED)
+ enable_hdr_split = true;
+ else
+ enable_hdr_split = false;
+
+ if (enable_hdr_split == priv->header_split_enabled)
+ return 0;
+
+ gve_get_curr_alloc_cfgs(priv, &qpls_alloc_cfg,
+ &tx_alloc_cfg, &rx_alloc_cfg);
+
+ rx_alloc_cfg.enable_header_split = enable_hdr_split;
+ rx_alloc_cfg.packet_buffer_size = gve_get_pkt_buf_size(priv, enable_hdr_split);
+
+ if (netif_running(priv->dev))
+ err = gve_adjust_config(priv, &qpls_alloc_cfg,
+ &tx_alloc_cfg, &rx_alloc_cfg);
+ return err;
+}
+
static int gve_set_features(struct net_device *netdev,
netdev_features_t features)
{
return 0;
}
+static void gve_rx_free_hdr_bufs(struct gve_priv *priv, struct gve_rx_ring *rx)
+{
+ struct device *hdev = &priv->pdev->dev;
+ int buf_count = rx->dqo.bufq.mask + 1;
+
+ if (rx->dqo.hdr_bufs.data) {
+ dma_free_coherent(hdev, priv->header_buf_size * buf_count,
+ rx->dqo.hdr_bufs.data, rx->dqo.hdr_bufs.addr);
+ rx->dqo.hdr_bufs.data = NULL;
+ }
+}
+
void gve_rx_stop_ring_dqo(struct gve_priv *priv, int idx)
{
int ntfy_idx = gve_rx_idx_to_ntfy(priv, idx);
kvfree(rx->dqo.buf_states);
rx->dqo.buf_states = NULL;
+ gve_rx_free_hdr_bufs(priv, rx);
+
netif_dbg(priv, drv, priv->dev, "freed rx ring %d\n", idx);
}
+static int gve_rx_alloc_hdr_bufs(struct gve_priv *priv, struct gve_rx_ring *rx)
+{
+ struct device *hdev = &priv->pdev->dev;
+ int buf_count = rx->dqo.bufq.mask + 1;
+
+ rx->dqo.hdr_bufs.data = dma_alloc_coherent(hdev, priv->header_buf_size * buf_count,
+ &rx->dqo.hdr_bufs.addr, GFP_KERNEL);
+ if (!rx->dqo.hdr_bufs.data)
+ return -ENOMEM;
+
+ return 0;
+}
+
void gve_rx_start_ring_dqo(struct gve_priv *priv, int idx)
{
int ntfy_idx = gve_rx_idx_to_ntfy(priv, idx);
if (!rx->dqo.buf_states)
return -ENOMEM;
+ /* Allocate header buffers for header-split */
+ if (cfg->enable_header_split)
+ if (gve_rx_alloc_hdr_bufs(priv, rx))
+ goto err;
+
/* Set up linked list of buffer IDs */
for (i = 0; i < rx->dqo.num_buf_states - 1; i++)
rx->dqo.buf_states[i].next = i + 1;
desc->buf_id = cpu_to_le16(buf_state - rx->dqo.buf_states);
desc->buf_addr = cpu_to_le64(buf_state->addr +
buf_state->page_info.page_offset);
+ if (rx->dqo.hdr_bufs.data)
+ desc->header_buf_addr =
+ cpu_to_le64(rx->dqo.hdr_bufs.addr +
+ priv->header_buf_size * bufq->tail);
bufq->tail = (bufq->tail + 1) & bufq->mask;
complq->num_free_slots--;
*/
static int gve_rx_dqo(struct napi_struct *napi, struct gve_rx_ring *rx,
const struct gve_rx_compl_desc_dqo *compl_desc,
- int queue_idx)
+ u32 desc_idx, int queue_idx)
{
const u16 buffer_id = le16_to_cpu(compl_desc->buf_id);
+ const bool hbo = compl_desc->header_buffer_overflow;
const bool eop = compl_desc->end_of_packet != 0;
+ const bool hsplit = compl_desc->split_header;
struct gve_rx_buf_state_dqo *buf_state;
struct gve_priv *priv = rx->gve;
u16 buf_len;
+ u16 hdr_len;
if (unlikely(buffer_id >= rx->dqo.num_buf_states)) {
net_err_ratelimited("%s: Invalid RX buffer_id=%u\n",
}
buf_len = compl_desc->packet_len;
+ hdr_len = compl_desc->header_len;
/* Page might have not been used for awhile and was likely last written
* by a different thread.
*/
prefetch(buf_state->page_info.page);
+ /* Copy the header into the skb in the case of header split */
+ if (hsplit) {
+ if (hdr_len && !hbo) {
+ rx->ctx.skb_head = gve_rx_copy_data(priv->dev, napi,
+ rx->dqo.hdr_bufs.data +
+ desc_idx * priv->header_buf_size,
+ hdr_len);
+ if (unlikely(!rx->ctx.skb_head))
+ goto error;
+ rx->ctx.skb_tail = rx->ctx.skb_head;
+ }
+ }
+
/* Sync the portion of dma buffer for CPU to read. */
dma_sync_single_range_for_cpu(&priv->pdev->dev, buf_state->addr,
buf_state->page_info.page_offset,
/* Do not read data until we own the descriptor */
dma_rmb();
- err = gve_rx_dqo(napi, rx, compl_desc, rx->q_num);
+ err = gve_rx_dqo(napi, rx, compl_desc, complq->head, rx->q_num);
if (err < 0) {
gve_rx_free_skb(rx);
u64_stats_update_begin(&rx->statss);
rx->ntfy_id = ntfy_idx;
}
-struct sk_buff *gve_rx_copy(struct net_device *dev, struct napi_struct *napi,
- struct gve_rx_slot_page_info *page_info, u16 len)
+struct sk_buff *gve_rx_copy_data(struct net_device *dev, struct napi_struct *napi,
+ u8 *data, u16 len)
{
- void *va = page_info->page_address + page_info->page_offset +
- page_info->pad;
struct sk_buff *skb;
skb = napi_alloc_skb(napi, len);
return NULL;
__skb_put(skb, len);
- skb_copy_to_linear_data_offset(skb, 0, va, len);
+ skb_copy_to_linear_data_offset(skb, 0, data, len);
skb->protocol = eth_type_trans(skb, dev);
return skb;
}
+struct sk_buff *gve_rx_copy(struct net_device *dev, struct napi_struct *napi,
+ struct gve_rx_slot_page_info *page_info, u16 len)
+{
+ void *va = page_info->page_address + page_info->page_offset +
+ page_info->pad;
+
+ return gve_rx_copy_data(dev, napi, va, len);
+}
+
void gve_dec_pagecnt_bias(struct gve_rx_slot_page_info *page_info)
{
page_info->pagecnt_bias--;
void gve_rx_remove_from_block(struct gve_priv *priv, int queue_idx);
void gve_rx_add_to_block(struct gve_priv *priv, int queue_idx);
+struct sk_buff *gve_rx_copy_data(struct net_device *dev, struct napi_struct *napi,
+ u8 *data, u16 len);
+
struct sk_buff *gve_rx_copy(struct net_device *dev, struct napi_struct *napi,
struct gve_rx_slot_page_info *page_info, u16 len);