Skip to content
Permalink
master
Switch branches/tags
Go to file
 
 
Cannot retrieve contributors at this time
// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
/*
* Copyright 2015-2020 Amazon.com, Inc. or its affiliates. All rights reserved.
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
#ifdef CONFIG_RFS_ACCEL
#include <linux/cpu_rmap.h>
#endif /* CONFIG_RFS_ACCEL */
#include <linux/ethtool.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/moduleparam.h>
#include <linux/numa.h>
#include <linux/pci.h>
#include <linux/utsname.h>
#include <linux/version.h>
#include <linux/vmalloc.h>
#if defined(CONFIG_NET_RX_BUSY_POLL) && (LINUX_VERSION_CODE < KERNEL_VERSION(4,5,0))
#include <net/busy_poll.h>
#endif
#include <net/ip.h>
#include "ena_netdev.h"
#include "ena_pci_id_tbl.h"
#include "ena_sysfs.h"
#include "ena_xdp.h"
#include "ena_lpc.h"
#include "ena_devlink.h"
static char version[] = DEVICE_NAME " v" DRV_MODULE_GENERATION "\n";
MODULE_AUTHOR("Amazon.com, Inc. or its affiliates");
MODULE_DESCRIPTION(DEVICE_NAME);
MODULE_LICENSE("GPL");
MODULE_VERSION(DRV_MODULE_GENERATION);
/* Time in jiffies before concluding the transmitter is hung. */
#define TX_TIMEOUT (5 * HZ)
#define ENA_MAX_RINGS min_t(unsigned int, ENA_MAX_NUM_IO_QUEUES, num_possible_cpus())
#define ENA_NAPI_BUDGET 64
#define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV | NETIF_MSG_PROBE | NETIF_MSG_IFUP | \
NETIF_MSG_TX_DONE | NETIF_MSG_TX_ERR | NETIF_MSG_RX_ERR)
#ifndef ENA_LINEAR_FRAG_SUPPORTED
#define ENA_SKB_PULL_MIN_LEN 64
#endif
static int debug = -1;
module_param(debug, int, 0444);
MODULE_PARM_DESC(debug, "Debug level (-1=default,0=none,...,16=all)");
static int rx_queue_size = ENA_DEFAULT_RING_SIZE;
module_param(rx_queue_size, int, 0444);
MODULE_PARM_DESC(rx_queue_size, "Rx queue size. The size should be a power of 2. Max value is 8K\n");
static int force_large_llq_header = 0;
module_param(force_large_llq_header, int, 0444);
MODULE_PARM_DESC(force_large_llq_header, "Increases maximum supported header size in LLQ mode to 224 bytes, while reducing the maximum TX queue size by half.\n");
static int num_io_queues = ENA_MAX_NUM_IO_QUEUES;
module_param(num_io_queues, int, 0444);
MODULE_PARM_DESC(num_io_queues, "Sets number of RX/TX queues to allocate to device. The maximum value depends on the device and number of online CPUs.\n");
static int enable_bql = 0;
module_param(enable_bql, int, 0444);
MODULE_PARM_DESC(enable_bql, "Enable BQL.\n");
static int lpc_size = ENA_LPC_DEFAULT_MULTIPLIER;
module_param(lpc_size, uint, 0444);
MODULE_PARM_DESC(lpc_size, "Each local page cache (lpc) holds N * 1024 pages. This parameter sets N which is rounded up to a multiplier of 2. If zero, the page cache is disabled. Max: 32\n");
static struct ena_aenq_handlers aenq_handlers;
static struct workqueue_struct *ena_wq;
MODULE_DEVICE_TABLE(pci, ena_pci_tbl);
static int ena_rss_init_default(struct ena_adapter *adapter);
static void check_for_admin_com_state(struct ena_adapter *adapter);
static void ena_calc_io_queue_size(struct ena_adapter *adapter,
struct ena_com_dev_get_features_ctx *get_feat_ctx);
static void ena_set_dev_offloads(struct ena_com_dev_get_features_ctx *feat,
struct net_device *netdev);
#ifdef HAVE_NDO_TX_TIMEOUT_STUCK_QUEUE_PARAMETER
static void ena_tx_timeout(struct net_device *dev, unsigned int txqueue)
#else
static void ena_tx_timeout(struct net_device *dev)
#endif
{
struct ena_adapter *adapter = netdev_priv(dev);
/* Change the state of the device to trigger reset
* Check that we are not in the middle or a trigger already
*/
if (test_and_set_bit(ENA_FLAG_TRIGGER_RESET, &adapter->flags))
return;
ena_reset_device(adapter, ENA_REGS_RESET_OS_NETDEV_WD);
ena_increase_stat(&adapter->dev_stats.tx_timeout, 1, &adapter->syncp);
netif_err(adapter, tx_err, dev, "Transmit time out\n");
}
static void update_rx_ring_mtu(struct ena_adapter *adapter, int mtu)
{
int i;
for (i = 0; i < adapter->num_io_queues; i++)
adapter->rx_ring[i].mtu = mtu;
}
static int ena_change_mtu(struct net_device *dev, int new_mtu)
{
struct ena_adapter *adapter = netdev_priv(dev);
int ret;
#ifndef HAVE_MTU_MIN_MAX_IN_NET_DEVICE
if ((new_mtu > adapter->max_mtu) || (new_mtu < ENA_MIN_MTU)) {
netif_err(adapter, drv, dev,
"Invalid MTU setting. new_mtu: %d max mtu: %d min mtu: %d\n",
new_mtu, adapter->max_mtu, ENA_MIN_MTU);
return -EINVAL;
}
#endif
ret = ena_com_set_dev_mtu(adapter->ena_dev, new_mtu);
if (!ret) {
netif_dbg(adapter, drv, dev, "Set MTU to %d\n", new_mtu);
update_rx_ring_mtu(adapter, new_mtu);
dev->mtu = new_mtu;
} else {
netif_err(adapter, drv, dev, "Failed to set MTU to %d\n",
new_mtu);
}
return ret;
}
int ena_xmit_common(struct ena_adapter *adapter,
struct ena_ring *ring,
struct ena_tx_buffer *tx_info,
struct ena_com_tx_ctx *ena_tx_ctx,
u16 next_to_use,
u32 bytes)
{
int rc, nb_hw_desc;
if (unlikely(ena_com_is_doorbell_needed(ring->ena_com_io_sq,
ena_tx_ctx))) {
netif_dbg(adapter, tx_queued, adapter->netdev,
"llq tx max burst size of queue %d achieved, writing doorbell to send burst\n",
ring->qid);
ena_ring_tx_doorbell(ring);
}
/* prepare the packet's descriptors to dma engine */
rc = ena_com_prepare_tx(ring->ena_com_io_sq, ena_tx_ctx,
&nb_hw_desc);
/* In case there isn't enough space in the queue for the packet,
* we simply drop it. All other failure reasons of
* ena_com_prepare_tx() are fatal and therefore require a device reset.
*/
if (unlikely(rc)) {
netif_err(adapter, tx_queued, adapter->netdev,
"Failed to prepare tx bufs\n");
ena_increase_stat(&ring->tx_stats.prepare_ctx_err, 1,
&ring->syncp);
if (rc != -ENOMEM)
ena_reset_device(adapter,
ENA_REGS_RESET_DRIVER_INVALID_STATE);
return rc;
}
u64_stats_update_begin(&ring->syncp);
ring->tx_stats.cnt++;
ring->tx_stats.bytes += bytes;
u64_stats_update_end(&ring->syncp);
tx_info->tx_descs = nb_hw_desc;
tx_info->total_tx_size = bytes;
tx_info->last_jiffies = jiffies;
tx_info->print_once = 0;
ring->next_to_use = ENA_TX_RING_IDX_NEXT(next_to_use,
ring->ring_size);
return 0;
}
static int ena_init_rx_cpu_rmap(struct ena_adapter *adapter)
{
#ifdef CONFIG_RFS_ACCEL
u32 i;
int rc;
adapter->netdev->rx_cpu_rmap = alloc_irq_cpu_rmap(adapter->num_io_queues);
if (!adapter->netdev->rx_cpu_rmap)
return -ENOMEM;
for (i = 0; i < adapter->num_io_queues; i++) {
int irq_idx = ENA_IO_IRQ_IDX(i);
rc = irq_cpu_rmap_add(adapter->netdev->rx_cpu_rmap,
#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 8, 0)
adapter->msix_entries[irq_idx].vector);
#else
pci_irq_vector(adapter->pdev, irq_idx));
#endif
if (rc) {
free_irq_cpu_rmap(adapter->netdev->rx_cpu_rmap);
adapter->netdev->rx_cpu_rmap = NULL;
return rc;
}
}
#endif /* CONFIG_RFS_ACCEL */
return 0;
}
static void ena_init_io_rings_common(struct ena_adapter *adapter,
struct ena_ring *ring, u16 qid)
{
ring->qid = qid;
ring->pdev = adapter->pdev;
ring->dev = &adapter->pdev->dev;
ring->netdev = adapter->netdev;
ring->napi = &adapter->ena_napi[qid].napi;
ring->adapter = adapter;
ring->ena_dev = adapter->ena_dev;
ring->per_napi_packets = 0;
ring->cpu = 0;
ring->numa_node = 0;
ring->no_interrupt_event_cnt = 0;
u64_stats_init(&ring->syncp);
}
void ena_init_io_rings(struct ena_adapter *adapter,
int first_index, int count)
{
struct ena_com_dev *ena_dev;
struct ena_ring *txr, *rxr;
int i;
ena_dev = adapter->ena_dev;
for (i = first_index; i < first_index + count; i++) {
txr = &adapter->tx_ring[i];
rxr = &adapter->rx_ring[i];
/* TX common ring state */
ena_init_io_rings_common(adapter, txr, i);
/* TX specific ring state */
txr->ring_size = adapter->requested_tx_ring_size;
txr->tx_max_header_size = ena_dev->tx_max_header_size;
txr->tx_mem_queue_type = ena_dev->tx_mem_queue_type;
txr->sgl_size = adapter->max_tx_sgl_size;
txr->enable_bql = enable_bql;
txr->smoothed_interval =
ena_com_get_nonadaptive_moderation_interval_tx(ena_dev);
txr->disable_meta_caching = adapter->disable_meta_caching;
#ifdef ENA_XDP_SUPPORT
spin_lock_init(&txr->xdp_tx_lock);
#endif
/* Don't init RX queues for xdp queues */
if (!ENA_IS_XDP_INDEX(adapter, i)) {
/* RX common ring state */
ena_init_io_rings_common(adapter, rxr, i);
/* RX specific ring state */
rxr->ring_size = adapter->requested_rx_ring_size;
rxr->rx_copybreak = adapter->rx_copybreak;
rxr->sgl_size = adapter->max_rx_sgl_size;
rxr->smoothed_interval =
ena_com_get_nonadaptive_moderation_interval_rx(ena_dev);
rxr->empty_rx_queue = 0;
rxr->rx_headroom = NET_SKB_PAD;
adapter->ena_napi[i].dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
#ifdef ENA_XDP_SUPPORT
rxr->xdp_ring = &adapter->tx_ring[i + adapter->num_io_queues];
#endif
}
}
}
/* ena_setup_tx_resources - allocate I/O Tx resources (Descriptors)
* @adapter: network interface device structure
* @qid: queue index
*
* Return 0 on success, negative on failure
*/
static int ena_setup_tx_resources(struct ena_adapter *adapter, int qid)
{
struct ena_ring *tx_ring = &adapter->tx_ring[qid];
struct ena_irq *ena_irq = &adapter->irq_tbl[ENA_IO_IRQ_IDX(qid)];
int size, i, node;
if (tx_ring->tx_buffer_info) {
netif_err(adapter, ifup,
adapter->netdev, "tx_buffer_info info is not NULL");
return -EEXIST;
}
size = sizeof(struct ena_tx_buffer) * tx_ring->ring_size;
node = cpu_to_node(ena_irq->cpu);
tx_ring->tx_buffer_info = vzalloc_node(size, node);
if (!tx_ring->tx_buffer_info) {
tx_ring->tx_buffer_info = vzalloc(size);
if (!tx_ring->tx_buffer_info)
goto err_tx_buffer_info;
}
size = sizeof(u16) * tx_ring->ring_size;
tx_ring->free_ids = vzalloc_node(size, node);
if (!tx_ring->free_ids) {
tx_ring->free_ids = vzalloc(size);
if (!tx_ring->free_ids)
goto err_tx_free_ids;
}
size = tx_ring->tx_max_header_size;
tx_ring->push_buf_intermediate_buf = vzalloc_node(size, node);
if (!tx_ring->push_buf_intermediate_buf) {
tx_ring->push_buf_intermediate_buf = vzalloc(size);
if (!tx_ring->push_buf_intermediate_buf)
goto err_push_buf_intermediate_buf;
}
/* Req id ring for TX out of order completions */
for (i = 0; i < tx_ring->ring_size; i++)
tx_ring->free_ids[i] = i;
/* Reset tx statistics */
memset(&tx_ring->tx_stats, 0x0, sizeof(tx_ring->tx_stats));
tx_ring->next_to_use = 0;
tx_ring->next_to_clean = 0;
tx_ring->cpu = ena_irq->cpu;
tx_ring->numa_node = node;
return 0;
err_push_buf_intermediate_buf:
vfree(tx_ring->free_ids);
tx_ring->free_ids = NULL;
err_tx_free_ids:
vfree(tx_ring->tx_buffer_info);
tx_ring->tx_buffer_info = NULL;
err_tx_buffer_info:
return -ENOMEM;
}
/* ena_free_tx_resources - Free I/O Tx Resources per Queue
* @adapter: network interface device structure
* @qid: queue index
*
* Free all transmit software resources
*/
static void ena_free_tx_resources(struct ena_adapter *adapter, int qid)
{
struct ena_ring *tx_ring = &adapter->tx_ring[qid];
vfree(tx_ring->tx_buffer_info);
tx_ring->tx_buffer_info = NULL;
vfree(tx_ring->free_ids);
tx_ring->free_ids = NULL;
vfree(tx_ring->push_buf_intermediate_buf);
tx_ring->push_buf_intermediate_buf = NULL;
}
int ena_setup_tx_resources_in_range(struct ena_adapter *adapter,
int first_index, int count)
{
int i, rc = 0;
for (i = first_index; i < first_index + count; i++) {
rc = ena_setup_tx_resources(adapter, i);
if (rc)
goto err_setup_tx;
}
return 0;
err_setup_tx:
netif_err(adapter, ifup, adapter->netdev,
"Tx queue %d: allocation failed\n", i);
/* rewind the index freeing the rings as we go */
while (first_index < i--)
ena_free_tx_resources(adapter, i);
return rc;
}
void ena_free_all_io_tx_resources_in_range(struct ena_adapter *adapter,
int first_index, int count)
{
int i;
for (i = first_index; i < first_index + count; i++)
ena_free_tx_resources(adapter, i);
}
/* ena_free_all_io_tx_resources - Free I/O Tx Resources for All Queues
* @adapter: board private structure
*
* Free all transmit software resources
*/
void ena_free_all_io_tx_resources(struct ena_adapter *adapter)
{
ena_free_all_io_tx_resources_in_range(adapter,
0,
adapter->xdp_num_queues +
adapter->num_io_queues);
}
/* ena_setup_rx_resources - allocate I/O Rx resources (Descriptors)
* @adapter: network interface device structure
* @qid: queue index
*
* Returns 0 on success, negative on failure
*/
static int ena_setup_rx_resources(struct ena_adapter *adapter,
u32 qid)
{
struct ena_ring *rx_ring = &adapter->rx_ring[qid];
struct ena_irq *ena_irq = &adapter->irq_tbl[ENA_IO_IRQ_IDX(qid)];
int size, node, i;
if (rx_ring->rx_buffer_info) {
netif_err(adapter, ifup, adapter->netdev,
"rx_buffer_info is not NULL");
return -EEXIST;
}
/* alloc extra element so in rx path
* we can always prefetch rx_info + 1
*/
size = sizeof(struct ena_rx_buffer) * (rx_ring->ring_size + 1);
node = cpu_to_node(ena_irq->cpu);
rx_ring->rx_buffer_info = vzalloc_node(size, node);
if (!rx_ring->rx_buffer_info) {
rx_ring->rx_buffer_info = vzalloc(size);
if (!rx_ring->rx_buffer_info)
return -ENOMEM;
}
size = sizeof(u16) * rx_ring->ring_size;
rx_ring->free_ids = vzalloc_node(size, node);
if (!rx_ring->free_ids) {
rx_ring->free_ids = vzalloc(size);
if (!rx_ring->free_ids) {
vfree(rx_ring->rx_buffer_info);
rx_ring->rx_buffer_info = NULL;
return -ENOMEM;
}
}
/* Req id ring for receiving RX pkts out of order */
for (i = 0; i < rx_ring->ring_size; i++)
rx_ring->free_ids[i] = i;
/* Reset rx statistics */
memset(&rx_ring->rx_stats, 0x0, sizeof(rx_ring->rx_stats));
#ifdef ENA_BUSY_POLL_SUPPORT
ena_bp_init_lock(rx_ring);
#endif
rx_ring->next_to_clean = 0;
rx_ring->next_to_use = 0;
rx_ring->cpu = ena_irq->cpu;
rx_ring->numa_node = node;
return 0;
}
/* ena_free_rx_resources - Free I/O Rx Resources
* @adapter: network interface device structure
* @qid: queue index
*
* Free all receive software resources
*/
static void ena_free_rx_resources(struct ena_adapter *adapter,
u32 qid)
{
struct ena_ring *rx_ring = &adapter->rx_ring[qid];
vfree(rx_ring->rx_buffer_info);
rx_ring->rx_buffer_info = NULL;
vfree(rx_ring->free_ids);
rx_ring->free_ids = NULL;
}
/* ena_setup_all_rx_resources - allocate I/O Rx queues resources for all queues
* @adapter: board private structure
*
* Return 0 on success, negative on failure
*/
static int ena_setup_all_rx_resources(struct ena_adapter *adapter)
{
int i, rc = 0;
for (i = 0; i < adapter->num_io_queues; i++) {
rc = ena_setup_rx_resources(adapter, i);
if (rc)
goto err_setup_rx;
}
return 0;
err_setup_rx:
netif_err(adapter, ifup, adapter->netdev,
"Rx queue %d: allocation failed\n", i);
/* rewind the index freeing the rings as we go */
while (i--)
ena_free_rx_resources(adapter, i);
return rc;
}
/* ena_free_all_io_rx_resources - Free I/O Rx Resources for All Queues
* @adapter: board private structure
*
* Free all receive software resources
*/
static void ena_free_all_io_rx_resources(struct ena_adapter *adapter)
{
int i;
for (i = 0; i < adapter->num_io_queues; i++)
ena_free_rx_resources(adapter, i);
}
struct page *ena_alloc_map_page(struct ena_ring *rx_ring, dma_addr_t *dma)
{
struct page *page;
/* This would allocate the page on the same NUMA node the executing code
* is running on.
*/
page = dev_alloc_page();
if (!page) {
ena_increase_stat(&rx_ring->rx_stats.page_alloc_fail, 1,
&rx_ring->syncp);
return ERR_PTR(-ENOSPC);
}
/* To enable NIC-side port-mirroring, AKA SPAN port,
* we make the buffer readable from the nic as well
*/
*dma = dma_map_page(rx_ring->dev, page, 0, ENA_PAGE_SIZE,
DMA_BIDIRECTIONAL);
if (unlikely(dma_mapping_error(rx_ring->dev, *dma))) {
ena_increase_stat(&rx_ring->rx_stats.dma_mapping_err, 1,
&rx_ring->syncp);
__free_page(page);
return ERR_PTR(-EIO);
}
return page;
}
static int ena_alloc_rx_buffer(struct ena_ring *rx_ring,
struct ena_rx_buffer *rx_info)
{
int headroom = rx_ring->rx_headroom;
struct ena_com_buf *ena_buf;
struct page *page;
dma_addr_t dma;
int tailroom;
/* restore page offset value in case it has been changed by device */
rx_info->buf_offset = headroom;
/* if previous allocated page is not used */
if (unlikely(rx_info->page))
return 0;
tailroom = SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
ena_buf = &rx_info->ena_buf;
#ifdef ENA_AF_XDP_SUPPORT
if (unlikely(ENA_IS_XSK_RING(rx_ring))) {
struct xdp_buff *xdp;
xdp = xsk_buff_alloc(rx_ring->xsk_pool);
if (!xdp)
return -ENOMEM;
ena_buf->paddr = xsk_buff_xdp_get_dma(xdp);
ena_buf->len = xsk_pool_get_rx_frame_size(rx_ring->xsk_pool);
rx_info->xdp = xdp;
return 0;
}
#endif /* ENA_AF_XDP_SUPPORT */
/* We handle DMA here */
page = ena_lpc_get_page(rx_ring, &dma, &rx_info->is_lpc_page);
if (unlikely(IS_ERR(page)))
return PTR_ERR(page);
netif_dbg(rx_ring->adapter, rx_status, rx_ring->netdev,
"Allocate page %p, rx_info %p\n", page, rx_info);
rx_info->page = page;
rx_info->dma_addr = dma;
rx_info->page_offset = 0;
ena_buf->paddr = dma + headroom;
ena_buf->len = ENA_PAGE_SIZE - headroom - tailroom;
return 0;
}
static void ena_unmap_rx_buff(struct ena_ring *rx_ring,
struct ena_rx_buffer *rx_info)
{
/* LPC pages are unmapped at cache destruction */
if (rx_info->is_lpc_page)
return;
dma_unmap_page(rx_ring->dev, rx_info->dma_addr,
ENA_PAGE_SIZE,
DMA_BIDIRECTIONAL);
}
static void ena_free_rx_page(struct ena_ring *rx_ring,
struct ena_rx_buffer *rx_info)
{
struct page *page = rx_info->page;
if (unlikely(!page)) {
netif_warn(rx_ring->adapter, rx_err, rx_ring->netdev,
"Trying to free unallocated buffer\n");
return;
}
ena_unmap_rx_buff(rx_ring, rx_info);
__free_page(page);
rx_info->page = NULL;
}
int ena_refill_rx_bufs(struct ena_ring *rx_ring, u32 num)
{
u16 next_to_use, req_id;
u32 i;
int rc;
next_to_use = rx_ring->next_to_use;
for (i = 0; i < num; i++) {
struct ena_rx_buffer *rx_info;
req_id = rx_ring->free_ids[next_to_use];
rx_info = &rx_ring->rx_buffer_info[req_id];
rc = ena_alloc_rx_buffer(rx_ring, rx_info);
if (unlikely(rc < 0)) {
if (!ENA_IS_XSK_RING(rx_ring))
netif_warn(rx_ring->adapter, rx_err, rx_ring->netdev,
"Failed to allocate buffer for rx queue %d\n",
rx_ring->qid);
break;
}
rc = ena_com_add_single_rx_desc(rx_ring->ena_com_io_sq,
&rx_info->ena_buf,
req_id);
if (unlikely(rc)) {
netif_warn(rx_ring->adapter, rx_status, rx_ring->netdev,
"Failed to add buffer for rx queue %d\n",
rx_ring->qid);
break;
}
next_to_use = ENA_RX_RING_IDX_NEXT(next_to_use,
rx_ring->ring_size);
}
if (unlikely(i < num)) {
ena_increase_stat(&rx_ring->rx_stats.refil_partial, 1,
&rx_ring->syncp);
if (!ENA_IS_XSK_RING(rx_ring))
netif_warn(rx_ring->adapter, rx_err, rx_ring->netdev,
"Refilled rx qid %d with only %d buffers (from %d)\n",
rx_ring->qid, i, num);
}
/* ena_com_write_sq_doorbell issues a wmb() */
if (likely(i))
ena_com_write_sq_doorbell(rx_ring->ena_com_io_sq);
rx_ring->next_to_use = next_to_use;
return i;
}
static void ena_free_rx_bufs(struct ena_adapter *adapter,
u32 qid)
{
struct ena_ring *rx_ring = &adapter->rx_ring[qid];
u32 i;
if (ENA_IS_XSK_RING(rx_ring)) {
ena_xdp_free_rx_bufs_zc(adapter, qid);
return;
}
for (i = 0; i < rx_ring->ring_size; i++) {
struct ena_rx_buffer *rx_info = &rx_ring->rx_buffer_info[i];
if (rx_info->page)
ena_free_rx_page(rx_ring, rx_info);
}
}
/* ena_refill_all_rx_bufs - allocate all queues Rx buffers
* @adapter: board private structure
*/
static void ena_refill_all_rx_bufs(struct ena_adapter *adapter)
{
struct ena_ring *rx_ring;
int i, rc, bufs_num;
for (i = 0; i < adapter->num_io_queues; i++) {
rx_ring = &adapter->rx_ring[i];
bufs_num = rx_ring->ring_size - 1;
rc = ena_refill_rx_bufs(rx_ring, bufs_num);
if (unlikely(rc != bufs_num))
netif_warn(rx_ring->adapter, rx_status, rx_ring->netdev,
"Refilling Queue %d failed. allocated %d buffers from: %d\n",
i, rc, bufs_num);
}
}
static void ena_free_all_rx_bufs(struct ena_adapter *adapter)
{
int i;
for (i = 0; i < adapter->num_io_queues; i++)
ena_free_rx_bufs(adapter, i);
}
void ena_unmap_tx_buff(struct ena_ring *tx_ring,
struct ena_tx_buffer *tx_info)
{
struct ena_com_buf *ena_buf;
u32 cnt;
int i;
ena_buf = tx_info->bufs;
cnt = tx_info->num_of_bufs;
if (unlikely(!cnt))
return;
if (tx_info->map_linear_data) {
dma_unmap_single(tx_ring->dev,
dma_unmap_addr(ena_buf, paddr),
dma_unmap_len(ena_buf, len),
DMA_TO_DEVICE);
ena_buf++;
cnt--;
}
/* unmap remaining mapped pages */
for (i = 0; i < cnt; i++) {
dma_unmap_page(tx_ring->dev, dma_unmap_addr(ena_buf, paddr),
dma_unmap_len(ena_buf, len), DMA_TO_DEVICE);
ena_buf++;
}
}
/* ena_free_tx_bufs - Free Tx Buffers per Queue
* @tx_ring: TX ring for which buffers be freed
*/
static void ena_free_tx_bufs(struct ena_ring *tx_ring)
{
bool print_once = true;
u32 i;
for (i = 0; i < tx_ring->ring_size; i++) {
struct ena_tx_buffer *tx_info = &tx_ring->tx_buffer_info[i];
if (!tx_info->skb)
continue;
if (print_once) {
netif_notice(tx_ring->adapter, ifdown, tx_ring->netdev,
"Free uncompleted tx skb qid %d idx 0x%x\n",
tx_ring->qid, i);
print_once = false;
} else {
netif_dbg(tx_ring->adapter, ifdown, tx_ring->netdev,
"Free uncompleted tx skb qid %d idx 0x%x\n",
tx_ring->qid, i);
}
ena_unmap_tx_buff(tx_ring, tx_info);
dev_kfree_skb_any(tx_info->skb);
}
netdev_tx_reset_queue(netdev_get_tx_queue(tx_ring->netdev,
tx_ring->qid));
}
static void ena_free_all_tx_bufs(struct ena_adapter *adapter)
{
struct ena_ring *tx_ring;
int i;
for (i = 0; i < adapter->num_io_queues + adapter->xdp_num_queues; i++) {
tx_ring = &adapter->tx_ring[i];
if (ENA_IS_XSK_RING(tx_ring)) {
ena_xdp_free_tx_bufs_zc(tx_ring);
continue;
}
ena_free_tx_bufs(tx_ring);
}
}
static void ena_destroy_all_tx_queues(struct ena_adapter *adapter)
{
u16 ena_qid;
int i;
for (i = 0; i < adapter->num_io_queues + adapter->xdp_num_queues; i++) {
ena_qid = ENA_IO_TXQ_IDX(i);
ena_com_destroy_io_queue(adapter->ena_dev, ena_qid);
}
}
static void ena_destroy_all_rx_queues(struct ena_adapter *adapter)
{
u16 ena_qid;
int i;
for (i = 0; i < adapter->num_io_queues; i++) {
ena_qid = ENA_IO_RXQ_IDX(i);
cancel_work_sync(&adapter->ena_napi[i].dim.work);
ena_xdp_unregister_rxq_info(&adapter->rx_ring[i]);
ena_com_destroy_io_queue(adapter->ena_dev, ena_qid);
}
}
static void ena_destroy_all_io_queues(struct ena_adapter *adapter)
{
ena_destroy_all_tx_queues(adapter);
ena_destroy_all_rx_queues(adapter);
}
int handle_invalid_req_id(struct ena_ring *ring, u16 req_id,
struct ena_tx_buffer *tx_info, bool is_xdp)
{
if (tx_info)
netif_err(ring->adapter,
tx_done,
ring->netdev,
"tx_info doesn't have valid %s. qid %u req_id %u",
is_xdp ? "xdp frame" : "skb", ring->qid, req_id);
else
netif_err(ring->adapter,
tx_done,
ring->netdev,
"Invalid req_id %u in qid %u\n",
req_id, ring->qid);
ena_increase_stat(&ring->tx_stats.bad_req_id, 1, &ring->syncp);
ena_reset_device(ring->adapter, ENA_REGS_RESET_INV_TX_REQ_ID);
return -EFAULT;
}
static int validate_tx_req_id(struct ena_ring *tx_ring, u16 req_id)
{
struct ena_tx_buffer *tx_info;
tx_info = &tx_ring->tx_buffer_info[req_id];
if (likely(tx_info->skb))
return 0;
return handle_invalid_req_id(tx_ring, req_id, tx_info, false);
}
static int ena_clean_tx_irq(struct ena_ring *tx_ring, u32 budget)
{
struct netdev_queue *txq;
bool above_thresh;
u32 tx_bytes = 0;
u32 total_done = 0;
u16 next_to_clean;
u16 req_id;
int tx_pkts = 0;
int rc;
next_to_clean = tx_ring->next_to_clean;
txq = netdev_get_tx_queue(tx_ring->netdev, tx_ring->qid);
while (tx_pkts < budget) {
struct ena_tx_buffer *tx_info;
struct sk_buff *skb;
rc = ena_com_tx_comp_req_id_get(tx_ring->ena_com_io_cq,
&req_id);
if (rc) {
if (unlikely(rc == -EINVAL))
handle_invalid_req_id(tx_ring, req_id, NULL,
false);
break;
}
/* validate that the request id points to a valid skb */
rc = validate_tx_req_id(tx_ring, req_id);
if (rc)
break;
tx_info = &tx_ring->tx_buffer_info[req_id];
skb = tx_info->skb;
/* prefetch skb_end_pointer() to speedup skb_shinfo(skb) */
prefetch(&skb->end);
tx_info->skb = NULL;
tx_info->last_jiffies = 0;
ena_unmap_tx_buff(tx_ring, tx_info);
netif_dbg(tx_ring->adapter, tx_done, tx_ring->netdev,
"tx_poll: q %d skb %p completed\n", tx_ring->qid,
skb);
tx_bytes += tx_info->total_tx_size;
dev_kfree_skb(skb);
tx_pkts++;
total_done += tx_info->tx_descs;
tx_ring->free_ids[next_to_clean] = req_id;
next_to_clean = ENA_TX_RING_IDX_NEXT(next_to_clean,
tx_ring->ring_size);
}
tx_ring->next_to_clean = next_to_clean;
ena_com_comp_ack(tx_ring->ena_com_io_sq, total_done);
ena_com_update_dev_comp_head(tx_ring->ena_com_io_cq);
if (tx_ring->enable_bql)
netdev_tx_completed_queue(txq, tx_pkts, tx_bytes);
netif_dbg(tx_ring->adapter, tx_done, tx_ring->netdev,
"tx_poll: q %d done. total pkts: %d\n",
tx_ring->qid, tx_pkts);
/* need to make the rings circular update visible to
* ena_start_xmit() before checking for netif_queue_stopped().
*/
smp_mb();
above_thresh = ena_com_sq_have_enough_space(tx_ring->ena_com_io_sq,
ENA_TX_WAKEUP_THRESH);
if (unlikely(netif_tx_queue_stopped(txq) && above_thresh)) {
__netif_tx_lock(txq, smp_processor_id());
above_thresh =
ena_com_sq_have_enough_space(tx_ring->ena_com_io_sq,
ENA_TX_WAKEUP_THRESH);
if (netif_tx_queue_stopped(txq) && above_thresh &&
test_bit(ENA_FLAG_DEV_UP, &tx_ring->adapter->flags)) {
netif_tx_wake_queue(txq);
ena_increase_stat(&tx_ring->tx_stats.queue_wakeup, 1,
&tx_ring->syncp);
}
__netif_tx_unlock(txq);
}
return tx_pkts;
}
static struct sk_buff *ena_alloc_skb(struct ena_ring *rx_ring, void *first_frag, u16 len)
{
struct sk_buff *skb;
#ifdef ENA_LINEAR_FRAG_SUPPORTED
if (!first_frag)
skb = netdev_alloc_skb_ip_align(rx_ring->netdev, len);
else
skb = build_skb(first_frag, len);
#else
if (!first_frag)
skb = netdev_alloc_skb_ip_align(rx_ring->netdev, len);
else
skb = netdev_alloc_skb_ip_align(rx_ring->netdev,
ENA_SKB_PULL_MIN_LEN);
#endif
if (unlikely(!skb)) {
ena_increase_stat(&rx_ring->rx_stats.skb_alloc_fail, 1,
&rx_ring->syncp);
netif_dbg(rx_ring->adapter, rx_err, rx_ring->netdev,
"Failed to allocate skb. first_frag %s\n",
first_frag ? "provided" : "not provided");
}
return skb;
}
static bool ena_try_rx_buf_page_reuse(struct ena_rx_buffer *rx_info,
u16 buf_len, u16 len)
{
struct ena_com_buf *ena_buf = &rx_info->ena_buf;
/* More than ENA_MIN_RX_BUF_SIZE left in the reused buffer
* for data + headroom + tailroom
*/
if (SKB_DATA_ALIGN(len) + ENA_MIN_RX_BUF_SIZE <= ena_buf->len) {
page_ref_inc(rx_info->page);
rx_info->page_offset += buf_len;
ena_buf->paddr += buf_len;
ena_buf->len -= buf_len;
return true;
}
return false;
}
static struct sk_buff *ena_rx_skb(struct ena_ring *rx_ring,
struct ena_com_rx_buf_info *ena_bufs,
u32 descs,
u16 *next_to_clean)
{
int tailroom = SKB_DATA_ALIGN(sizeof(struct skb_shared_info));
bool is_xdp_loaded = ena_xdp_present_ring(rx_ring);
struct ena_rx_buffer *rx_info;
struct ena_adapter *adapter;
int page_offset, pkt_offset;
u16 len, req_id, buf = 0;
bool reuse_rx_buf_page;
struct sk_buff *skb;
void *buf_addr;
int buf_offset;
u16 buf_len;
#ifndef ENA_LINEAR_FRAG_SUPPORTED
void *data_addr;
u16 hlen;
#endif
len = ena_bufs[buf].len;
req_id = ena_bufs[buf].req_id;
rx_info = &rx_ring->rx_buffer_info[req_id];
if (unlikely(!rx_info->page)) {
adapter = rx_ring->adapter;
netif_err(adapter, rx_err, rx_ring->netdev,
"Page is NULL. qid %u req_id %u\n", rx_ring->qid, req_id);
ena_increase_stat(&rx_ring->rx_stats.bad_req_id, 1, &rx_ring->syncp);
ena_reset_device(adapter, ENA_REGS_RESET_INV_RX_REQ_ID);
return NULL;
}
netif_dbg(rx_ring->adapter, rx_status, rx_ring->netdev,
"rx_info %p page %p\n",
rx_info, rx_info->page);
buf_offset = rx_info->buf_offset;
page_offset = rx_info->page_offset;
buf_addr = page_address(rx_info->page) + page_offset;
if (len <= rx_ring->rx_copybreak) {
skb = ena_alloc_skb(rx_ring, NULL, len);
if (unlikely(!skb))
return NULL;
pkt_offset = buf_offset - rx_ring->rx_headroom;
/* sync this buffer for CPU use */
dma_sync_single_for_cpu(rx_ring->dev,
dma_unmap_addr(&rx_info->ena_buf, paddr) + pkt_offset,
len,
DMA_FROM_DEVICE);
skb_copy_to_linear_data(skb, buf_addr + buf_offset, len);
dma_sync_single_for_device(rx_ring->dev,
dma_unmap_addr(&rx_info->ena_buf, paddr) + pkt_offset,
len,
DMA_FROM_DEVICE);
skb_put(skb, len);
netif_dbg(rx_ring->adapter, rx_status, rx_ring->netdev,
"RX allocated small packet. len %d.\n", skb->len);
#ifdef ENA_BUSY_POLL_SUPPORT
skb_mark_napi_id(skb, rx_ring->napi);
#endif
skb->protocol = eth_type_trans(skb, rx_ring->netdev);
rx_ring->free_ids[*next_to_clean] = req_id;
*next_to_clean = ENA_RX_RING_IDX_ADD(*next_to_clean, descs,
rx_ring->ring_size);
return skb;
}
buf_len = SKB_DATA_ALIGN(len + buf_offset + tailroom);
/* If XDP isn't loaded try to reuse part of the RX buffer */
reuse_rx_buf_page = !is_xdp_loaded &&
ena_try_rx_buf_page_reuse(rx_info, buf_len, len);
if (!reuse_rx_buf_page)
ena_unmap_rx_buff(rx_ring, rx_info);
skb = ena_alloc_skb(rx_ring, buf_addr, buf_len);
if (unlikely(!skb))
return NULL;
#ifdef ENA_LINEAR_FRAG_SUPPORTED
/* Populate skb's linear part */
skb_reserve(skb, buf_offset);
skb_put(skb, len);
#else
data_addr = buf_addr + buf_offset;
/* GRO expects us to have the ethernet header in the linear part.
* Copy the first ENA_SKB_PULL_MIN_LEN bytes because it is more
* efficient.
*/
hlen = min_t(u16, len, ENA_SKB_PULL_MIN_LEN);
memcpy(__skb_put(skb, hlen), data_addr, hlen);
if (hlen < len)
skb_add_rx_frag(skb, 0, rx_info->page,
page_offset + buf_offset + hlen,
len - hlen, buf_len);
#endif
skb->protocol = eth_type_trans(skb, rx_ring->netdev);
do {
netif_dbg(rx_ring->adapter, rx_status, rx_ring->netdev,
"RX skb updated. len %d. data_len %d\n",
skb->len, skb->data_len);
if (!reuse_rx_buf_page)
rx_info->page = NULL;
rx_ring->free_ids[*next_to_clean] = req_id;
*next_to_clean =
ENA_RX_RING_IDX_NEXT(*next_to_clean,
rx_ring->ring_size);
if (likely(--descs == 0))
break;
buf++;
len = ena_bufs[buf].len;
req_id = ena_bufs[buf].req_id;
rx_info = &rx_ring->rx_buffer_info[req_id];
/* rx_info->buf_offset includes rx_ring->rx_headroom */
buf_offset = rx_info->buf_offset;
buf_len = SKB_DATA_ALIGN(len + buf_offset + tailroom);
page_offset = rx_info->page_offset;
reuse_rx_buf_page = !is_xdp_loaded &&
ena_try_rx_buf_page_reuse(rx_info, buf_len, len);
if (!reuse_rx_buf_page)
ena_unmap_rx_buff(rx_ring, rx_info);
skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_info->page,
page_offset + buf_offset, len, buf_len);
} while (1);
#ifdef ENA_BUSY_POLL_SUPPORT
skb_mark_napi_id(skb, rx_ring->napi);
#endif
return skb;
}
/* ena_rx_checksum - indicate in skb if hw indicated a good cksum
* @adapter: structure containing adapter specific data
* @ena_rx_ctx: received packet context/metadata
* @skb: skb currently being received and modified
*/
void ena_rx_checksum(struct ena_ring *rx_ring,
struct ena_com_rx_ctx *ena_rx_ctx,
struct sk_buff *skb)
{
/* Rx csum disabled */
if (unlikely(!(rx_ring->netdev->features & NETIF_F_RXCSUM))) {
skb->ip_summed = CHECKSUM_NONE;
return;
}
/* For fragmented packets the checksum isn't valid */
if (ena_rx_ctx->frag) {
skb->ip_summed = CHECKSUM_NONE;
return;
}
/* if IP and error */
if (unlikely((ena_rx_ctx->l3_proto == ENA_ETH_IO_L3_PROTO_IPV4) &&
(ena_rx_ctx->l3_csum_err))) {
/* ipv4 checksum error */
skb->ip_summed = CHECKSUM_NONE;
ena_increase_stat(&rx_ring->rx_stats.csum_bad, 1,
&rx_ring->syncp);
netif_dbg(rx_ring->adapter, rx_err, rx_ring->netdev,
"RX IPv4 header checksum error\n");
return;
}
/* if TCP/UDP */
if (likely((ena_rx_ctx->l4_proto == ENA_ETH_IO_L4_PROTO_TCP) ||
(ena_rx_ctx->l4_proto == ENA_ETH_IO_L4_PROTO_UDP))) {
if (unlikely(ena_rx_ctx->l4_csum_err)) {
/* TCP/UDP checksum error */
ena_increase_stat(&rx_ring->rx_stats.csum_bad, 1,
&rx_ring->syncp);
netif_dbg(rx_ring->adapter, rx_err, rx_ring->netdev,
"RX L4 checksum error\n");
skb->ip_summed = CHECKSUM_NONE;
return;
}
if (likely(ena_rx_ctx->l4_csum_checked)) {
skb->ip_summed = CHECKSUM_UNNECESSARY;
ena_increase_stat(&rx_ring->rx_stats.csum_good, 1,
&rx_ring->syncp);
} else {
ena_increase_stat(&rx_ring->rx_stats.csum_unchecked, 1,
&rx_ring->syncp);
skb->ip_summed = CHECKSUM_NONE;
}
} else {
skb->ip_summed = CHECKSUM_NONE;
return;
}
}
void ena_set_rx_hash(struct ena_ring *rx_ring,
struct ena_com_rx_ctx *ena_rx_ctx,
struct sk_buff *skb)
{
#ifdef NETIF_F_RXHASH
enum pkt_hash_types hash_type;
if (likely(rx_ring->netdev->features & NETIF_F_RXHASH)) {
if (likely((ena_rx_ctx->l4_proto == ENA_ETH_IO_L4_PROTO_TCP) ||
(ena_rx_ctx->l4_proto == ENA_ETH_IO_L4_PROTO_UDP)))
hash_type = PKT_HASH_TYPE_L4;
else
hash_type = PKT_HASH_TYPE_NONE;
/* Override hash type if the packet is fragmented */
if (ena_rx_ctx->frag)
hash_type = PKT_HASH_TYPE_NONE;
skb_set_hash(skb, ena_rx_ctx->hash, hash_type);
}
#endif /* NETIF_F_RXHASH */
}
#ifdef ENA_XDP_SUPPORT
static int ena_xdp_handle_buff(struct ena_ring *rx_ring, struct xdp_buff *xdp)
{
struct ena_rx_buffer *rx_info;
int ret;
rx_info = &rx_ring->rx_buffer_info[rx_ring->ena_bufs[0].req_id];
xdp_prepare_buff(xdp, page_address(rx_info->page),
rx_info->buf_offset,
rx_ring->ena_bufs[0].len, false);
/* If for some reason we received a bigger packet than
* we expect, then we simply drop it
*/
if (unlikely(rx_ring->ena_bufs[0].len > ENA_XDP_MAX_MTU))
return ENA_XDP_DROP;
ret = ena_xdp_execute(rx_ring, xdp);
/* The xdp program might expand the headers */
if (ret == ENA_XDP_PASS) {
rx_info->buf_offset = xdp->data - xdp->data_hard_start;
rx_ring->ena_bufs[0].len = xdp->data_end - xdp->data;
}
return ret;
}
#endif /* ENA_XDP_SUPPORT */
/* ena_clean_rx_irq - Cleanup RX irq
* @rx_ring: RX ring to clean
* @napi: napi handler
* @budget: how many packets driver is allowed to clean
*
* Returns the number of cleaned buffers.
*/
static int ena_clean_rx_irq(struct ena_ring *rx_ring, struct napi_struct *napi,
u32 budget)
{
u16 next_to_clean = rx_ring->next_to_clean;
struct ena_com_rx_ctx ena_rx_ctx;
struct ena_rx_buffer *rx_info;
struct ena_adapter *adapter;
u32 res_budget, work_done;
int rx_copybreak_pkt = 0;
int refill_threshold;
struct sk_buff *skb;
int refill_required;
#ifdef ENA_XDP_SUPPORT
struct xdp_buff xdp;
int xdp_flags = 0;
#endif /* ENA_XDP_SUPPORT */
int total_len = 0;
#ifdef ENA_XDP_SUPPORT
int xdp_verdict;
#endif /* ENA_XDP_SUPPORT */
int rc = 0;
int i;
netif_dbg(rx_ring->adapter, rx_status, rx_ring->netdev,
"%s qid %d\n", __func__, rx_ring->qid);
res_budget = budget;
#ifdef ENA_XDP_SUPPORT
xdp_init_buff(&xdp, ENA_PAGE_SIZE, &rx_ring->xdp_rxq);
#endif /* ENA_XDP_SUPPORT */
do {
#ifdef ENA_XDP_SUPPORT
xdp_verdict = ENA_XDP_PASS;
skb = NULL;
#endif /* ENA_XDP_SUPPORT */
ena_rx_ctx.ena_bufs = rx_ring->ena_bufs;
ena_rx_ctx.max_bufs = rx_ring->sgl_size;
ena_rx_ctx.descs = 0;
ena_rx_ctx.pkt_offset = 0;
rc = ena_com_rx_pkt(rx_ring->ena_com_io_cq,
rx_ring->ena_com_io_sq,
&ena_rx_ctx);
if (unlikely(rc))
goto error;
if (unlikely(ena_rx_ctx.descs == 0))
break;
/* First descriptor might have an offset set by the device */
rx_info = &rx_ring->rx_buffer_info[rx_ring->ena_bufs[0].req_id];
rx_info->buf_offset += ena_rx_ctx.pkt_offset;
netif_dbg(rx_ring->adapter, rx_status, rx_ring->netdev,
"rx_poll: q %d got packet from ena. descs #: %d l3 proto %d l4 proto %d hash: %x\n",
rx_ring->qid, ena_rx_ctx.descs, ena_rx_ctx.l3_proto,
ena_rx_ctx.l4_proto, ena_rx_ctx.hash);
#ifdef ENA_XDP_SUPPORT
if (ena_xdp_present_ring(rx_ring))
xdp_verdict = ena_xdp_handle_buff(rx_ring, &xdp);
/* allocate skb and fill it */
if (xdp_verdict == ENA_XDP_PASS)
skb = ena_rx_skb(rx_ring,
rx_ring->ena_bufs,
ena_rx_ctx.descs,
&next_to_clean);
#else
skb = ena_rx_skb(rx_ring, rx_ring->ena_bufs, ena_rx_ctx.descs,
&next_to_clean);
#endif /* ENA_XDP_SUPPORT */
if (unlikely(!skb)) {
for (i = 0; i < ena_rx_ctx.descs; i++) {
int req_id = rx_ring->ena_bufs[i].req_id;
rx_ring->free_ids[next_to_clean] = req_id;
next_to_clean =
ENA_RX_RING_IDX_NEXT(next_to_clean,
rx_ring->ring_size);
#ifdef ENA_XDP_SUPPORT
/* Packets was passed for transmission, unmap it
* from RX side.
*/
if (xdp_verdict & ENA_XDP_FORWARDED) {
ena_unmap_rx_buff(rx_ring,
&rx_ring->rx_buffer_info[req_id]);
rx_ring->rx_buffer_info[req_id].page = NULL;
}
#endif /* ENA_XDP_SUPPORT */
}
#ifdef ENA_XDP_SUPPORT
if (xdp_verdict != ENA_XDP_PASS) {
xdp_flags |= xdp_verdict;
total_len += ena_rx_ctx.ena_bufs[0].len;
res_budget--;
continue;
}
#endif /* ENA_XDP_SUPPORT */
break;
}
ena_rx_checksum(rx_ring, &ena_rx_ctx, skb);
ena_set_rx_hash(rx_ring, &ena_rx_ctx, skb);
skb_record_rx_queue(skb, rx_ring->qid);
if (rx_ring->ena_bufs[0].len <= rx_ring->rx_copybreak)
rx_copybreak_pkt++;
total_len += skb->len;
#ifdef ENA_BUSY_POLL_SUPPORT
if (ena_bp_busy_polling(rx_ring))
netif_receive_skb(skb);
else
napi_gro_receive(napi, skb);
#else
napi_gro_receive(napi, skb);
#endif /* ENA_BUSY_POLL_SUPPORT */
res_budget--;
} while (likely(res_budget));
work_done = budget - res_budget;
rx_ring->per_napi_packets += work_done;
u64_stats_update_begin(&rx_ring->syncp);
rx_ring->rx_stats.bytes += total_len;
rx_ring->rx_stats.cnt += work_done;
rx_ring->rx_stats.rx_copybreak_pkt += rx_copybreak_pkt;
u64_stats_update_end(&rx_ring->syncp);
rx_ring->next_to_clean = next_to_clean;
refill_required = ena_com_free_q_entries(rx_ring->ena_com_io_sq);
refill_threshold =
min_t(int, rx_ring->ring_size / ENA_RX_REFILL_THRESH_DIVIDER,
ENA_RX_REFILL_THRESH_PACKET);
/* Optimization, try to batch new rx buffers */
if (refill_required > refill_threshold) {
ena_com_update_dev_comp_head(rx_ring->ena_com_io_cq);
ena_refill_rx_bufs(rx_ring, refill_required);
}
#ifdef ENA_XDP_SUPPORT
if (xdp_flags & ENA_XDP_REDIRECT)
xdp_do_flush_map();
#endif
return work_done;
error:
adapter = netdev_priv(rx_ring->netdev);
if (rc == -ENOSPC) {
ena_increase_stat(&rx_ring->rx_stats.bad_desc_num, 1,
&rx_ring->syncp);
ena_reset_device(adapter, ENA_REGS_RESET_TOO_MANY_RX_DESCS);
} else {
ena_increase_stat(&rx_ring->rx_stats.bad_req_id, 1,
&rx_ring->syncp);
ena_reset_device(adapter, ENA_REGS_RESET_INV_RX_REQ_ID);
}
return 0;
}
static void ena_dim_work(struct work_struct *w)
{
struct dim *dim = container_of(w, struct dim, work);
struct dim_cq_moder cur_moder =
net_dim_get_rx_moderation(dim->mode, dim->profile_ix);
struct ena_napi *ena_napi = container_of(dim, struct ena_napi, dim);
ena_napi->rx_ring->smoothed_interval = cur_moder.usec;
dim->state = DIM_START_MEASURE;
}
static void ena_adjust_adaptive_rx_intr_moderation(struct ena_napi *ena_napi)
{
struct dim_sample dim_sample;
struct ena_ring *rx_ring = ena_napi->rx_ring;
if (!rx_ring->per_napi_packets)
return;
rx_ring->non_empty_napi_events++;
dim_update_sample(rx_ring->non_empty_napi_events,
rx_ring->rx_stats.cnt,
rx_ring->rx_stats.bytes,
&dim_sample);
net_dim(&ena_napi->dim, dim_sample);
rx_ring->per_napi_packets = 0;
}
void ena_unmask_interrupt(struct ena_ring *tx_ring,
struct ena_ring *rx_ring)
{
struct ena_eth_io_intr_reg intr_reg;
u32 rx_interval = 0;
/* Rx ring can be NULL when for XDP tx queues which don't have an
* accompanying rx_ring pair.
*/
if (rx_ring)
rx_interval = ena_com_get_adaptive_moderation_enabled(rx_ring->ena_dev) ?
rx_ring->smoothed_interval :
ena_com_get_nonadaptive_moderation_interval_rx(rx_ring->ena_dev);
/* Update intr register: rx intr delay,
* tx intr delay and interrupt unmask
*/
ena_com_update_intr_reg(&intr_reg,
rx_interval,
tx_ring->smoothed_interval,
true);
ena_increase_stat(&tx_ring->tx_stats.unmask_interrupt, 1,
&tx_ring->syncp);
/* It is a shared MSI-X.
* Tx and Rx CQ have pointer to it.
* So we use one of them to reach the intr reg
* The Tx ring is used because the rx_ring is NULL for XDP queues
*/
ena_com_unmask_intr(tx_ring->ena_com_io_cq, &intr_reg);
}
void ena_update_ring_numa_node(struct ena_ring *tx_ring,
struct ena_ring *rx_ring)
{
int cpu = get_cpu();
int numa_node;
/* Check only one ring since the 2 rings are running on the same cpu */
if (likely(tx_ring->cpu == cpu))
goto out;
tx_ring->cpu = cpu;
if (rx_ring)
rx_ring->cpu = cpu;
numa_node = cpu_to_node(cpu);
if (likely(tx_ring->numa_node == numa_node))
goto out;
put_cpu();
if (numa_node != NUMA_NO_NODE) {
ena_com_update_numa_node(tx_ring->ena_com_io_cq, numa_node);
tx_ring->numa_node = numa_node;
if (rx_ring) {
rx_ring->numa_node = numa_node;
ena_com_update_numa_node(rx_ring->ena_com_io_cq,
numa_node);
}
}
return;
out:
put_cpu();
}
static int ena_io_poll(struct napi_struct *napi, int budget)
{
struct ena_napi *ena_napi = container_of(napi, struct ena_napi, napi);
struct ena_ring *tx_ring, *rx_ring;
int tx_work_done;
int rx_work_done = 0;
int tx_budget;
int napi_comp_call = 0;
int ret;
tx_ring = ena_napi->tx_ring;
rx_ring = ena_napi->rx_ring;
tx_budget = tx_ring->ring_size / ENA_TX_POLL_BUDGET_DIVIDER;
if (!test_bit(ENA_FLAG_DEV_UP, &tx_ring->adapter->flags) ||
test_bit(ENA_FLAG_TRIGGER_RESET, &tx_ring->adapter->flags)) {
napi_complete_done(napi, 0);
return 0;
}
#ifdef ENA_BUSY_POLL_SUPPORT
if (!ena_bp_lock_napi(rx_ring))
return budget;
#endif
tx_work_done = ena_clean_tx_irq(tx_ring, tx_budget);
/* On netpoll the budget is zero and the handler should only clean the
* tx completions.
*/
if (likely(budget))
rx_work_done = ena_clean_rx_irq(rx_ring, napi, budget);
/* If the device is about to reset or down, avoid unmask
* the interrupt and return 0 so NAPI won't reschedule
*/
if (unlikely(!test_bit(ENA_FLAG_DEV_UP, &tx_ring->adapter->flags) ||
test_bit(ENA_FLAG_TRIGGER_RESET, &tx_ring->adapter->flags))) {
napi_complete_done(napi, 0);
ret = 0;
} else if ((budget > rx_work_done) && (tx_budget > tx_work_done)) {
napi_comp_call = 1;
/* Update numa and unmask the interrupt only when schedule
* from the interrupt context (vs from sk_busy_loop)
*/
#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 10, 0)
if (napi_complete_done(napi, rx_work_done) &&
READ_ONCE(ena_napi->interrupts_masked)) {
#else
napi_complete_done(napi, rx_work_done);
if (READ_ONCE(ena_napi->interrupts_masked)) {
#endif
smp_rmb(); /* make sure interrupts_masked is read */
WRITE_ONCE(ena_napi->interrupts_masked, false);
/* We apply adaptive moderation on Rx path only.
* Tx uses static interrupt moderation.
*/
if (ena_com_get_adaptive_moderation_enabled(rx_ring->ena_dev))
ena_adjust_adaptive_rx_intr_moderation(ena_napi);
ena_update_ring_numa_node(tx_ring, rx_ring);
ena_unmask_interrupt(tx_ring, rx_ring);
}
ret = rx_work_done;
} else {
ret = budget;
}
u64_stats_update_begin(&tx_ring->syncp);
tx_ring->tx_stats.napi_comp += napi_comp_call;
tx_ring->tx_stats.tx_poll++;
u64_stats_update_end(&tx_ring->syncp);
#ifdef ENA_BUSY_POLL_SUPPORT
ena_bp_unlock_napi(rx_ring);
#endif
tx_ring->tx_stats.last_napi_jiffies = jiffies;
return ret;
}
static irqreturn_t ena_intr_msix_mgmnt(int irq, void *data)
{
struct ena_adapter *adapter = (struct ena_adapter *)data;
ena_com_admin_q_comp_intr_handler(adapter->ena_dev);
/* Don't call the aenq handler before probe is done */
if (likely(test_bit(ENA_FLAG_DEVICE_RUNNING, &adapter->flags)))
ena_com_aenq_intr_handler(adapter->ena_dev, data);
return IRQ_HANDLED;
}
/* ena_intr_msix_io - MSI-X Interrupt Handler for Tx/Rx
* @irq: interrupt number
* @data: pointer to a network interface private napi device structure
*/
static irqreturn_t ena_intr_msix_io(int irq, void *data)
{
struct ena_napi *ena_napi = data;
/* Used to check HW health */
WRITE_ONCE(ena_napi->first_interrupt, true);
WRITE_ONCE(ena_napi->interrupts_masked, true);
smp_wmb(); /* write interrupts_masked before calling napi */
napi_schedule_irqoff(&ena_napi->napi);
return IRQ_HANDLED;
}
/* Reserve a single MSI-X vector for management (admin + aenq).
* plus reserve one vector for each potential io queue.
* the number of potential io queues is the minimum of what the device
* supports and the number of vCPUs.
*/
static int ena_enable_msix(struct ena_adapter *adapter)
{
int msix_vecs, irq_cnt;
#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 8, 0)
int i;
#endif
if (test_bit(ENA_FLAG_MSIX_ENABLED, &adapter->flags)) {
netif_err(adapter, probe, adapter->netdev,
"Error, MSI-X is already enabled\n");
return -EPERM;
}
/* Reserved the max msix vectors we might need */
msix_vecs = ENA_MAX_MSIX_VEC(adapter->max_num_io_queues);
netif_dbg(adapter, probe, adapter->netdev,
"Trying to enable MSI-X, vectors %d\n", msix_vecs);
#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 8, 0)
adapter->msix_entries = vzalloc(msix_vecs * sizeof(struct msix_entry));
if (!adapter->msix_entries)
return -ENOMEM;
for (i = 0; i < msix_vecs; i++)
adapter->msix_entries[i].entry = i;
irq_cnt = pci_enable_msix_range(adapter->pdev, adapter->msix_entries,
ENA_MIN_MSIX_VEC, msix_vecs);
#else
irq_cnt = pci_alloc_irq_vectors(adapter->pdev, ENA_MIN_MSIX_VEC,
msix_vecs, PCI_IRQ_MSIX);
#endif
if (irq_cnt < 0) {
netif_err(adapter, probe, adapter->netdev,
"Failed to enable MSI-X. irq_cnt %d\n", irq_cnt);
#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 8, 0)
vfree(adapter->msix_entries);
adapter->msix_entries = NULL;
#endif
return -ENOSPC;
}
if (irq_cnt != msix_vecs) {
netif_notice(adapter, probe, adapter->netdev,
"Enable only %d MSI-X (out of %d), reduce the number of queues\n",
irq_cnt, msix_vecs);
adapter->num_io_queues = irq_cnt - ENA_ADMIN_MSIX_VEC;
}
if (ena_init_rx_cpu_rmap(adapter))
netif_warn(adapter, probe, adapter->netdev,
"Failed to map IRQs to CPUs\n");
adapter->msix_vecs = irq_cnt;
set_bit(ENA_FLAG_MSIX_ENABLED, &adapter->flags);
return 0;
}
static void ena_setup_mgmnt_intr(struct ena_adapter *adapter)
{
u32 cpu;
snprintf(adapter->irq_tbl[ENA_MGMNT_IRQ_IDX].name,
ENA_IRQNAME_SIZE, "ena-mgmnt@pci:%s",
pci_name(adapter->pdev));
adapter->irq_tbl[ENA_MGMNT_IRQ_IDX].handler =
ena_intr_msix_mgmnt;
adapter->irq_tbl[ENA_MGMNT_IRQ_IDX].data = adapter;
adapter->irq_tbl[ENA_MGMNT_IRQ_IDX].vector =
#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 8, 0)
adapter->msix_entries[ENA_MGMNT_IRQ_IDX].vector;
#else
pci_irq_vector(adapter->pdev, ENA_MGMNT_IRQ_IDX);
#endif
cpu = cpumask_first(cpu_online_mask);
adapter->irq_tbl[ENA_MGMNT_IRQ_IDX].cpu = cpu;
cpumask_set_cpu(cpu,
&adapter->irq_tbl[ENA_MGMNT_IRQ_IDX].affinity_hint_mask);
}
static void ena_setup_io_intr(struct ena_adapter *adapter)
{
struct net_device *netdev;
int irq_idx, i, cpu;
int io_queue_count;
netdev = adapter->netdev;
io_queue_count = adapter->num_io_queues + adapter->xdp_num_queues;
for (i = 0; i < io_queue_count; i++) {
irq_idx = ENA_IO_IRQ_IDX(i);
cpu = i % num_online_cpus();
snprintf(adapter->irq_tbl[irq_idx].name, ENA_IRQNAME_SIZE,
"%s-Tx-Rx-%d", netdev->name, i);
adapter->irq_tbl[irq_idx].handler = ena_intr_msix_io;
adapter->irq_tbl[irq_idx].data = &adapter->ena_napi[i];
adapter->irq_tbl[irq_idx].vector =
#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 8, 0)
adapter->msix_entries[irq_idx].vector;
#else
pci_irq_vector(adapter->pdev, irq_idx);
#endif
adapter->irq_tbl[irq_idx].cpu = cpu;
cpumask_set_cpu(cpu,
&adapter->irq_tbl[irq_idx].affinity_hint_mask);
}
}
static int ena_request_mgmnt_irq(struct ena_adapter *adapter)
{
unsigned long flags = 0;
struct ena_irq *irq;
int rc;
irq = &adapter->irq_tbl[ENA_MGMNT_IRQ_IDX];
rc = request_irq(irq->vector, irq->handler, flags, irq->name,
irq->data);
if (rc) {
netif_err(adapter, probe, adapter->netdev,
"Failed to request admin irq\n");
return rc;
}
netif_dbg(adapter, probe, adapter->netdev,
"Set affinity hint of mgmnt irq.to 0x%lx (irq vector: %d)\n",
irq->affinity_hint_mask.bits[0], irq->vector);
return rc;
}
static int ena_request_io_irq(struct ena_adapter *adapter)
{
u32 io_queue_count = adapter->num_io_queues + adapter->xdp_num_queues;
unsigned long flags = 0;
struct ena_irq *irq;
int rc = 0, i, k;
if (!test_bit(ENA_FLAG_MSIX_ENABLED, &adapter->flags)) {
netif_err(adapter, ifup, adapter->netdev,
"Failed to request I/O IRQ: MSI-X is not enabled\n");
return -EINVAL;
}
for (i = ENA_IO_IRQ_FIRST_IDX; i < ENA_MAX_MSIX_VEC(io_queue_count); i++) {
irq = &adapter->irq_tbl[i];
rc = request_irq(irq->vector, irq->handler, flags, irq->name,
irq->data);
if (rc) {
netif_err(adapter, ifup, adapter->netdev,
"Failed to request I/O IRQ. index %d rc %d\n",
i, rc);
goto err;
}
netif_dbg(adapter, ifup, adapter->netdev,
"Set affinity hint of irq. index %d to 0x%lx (irq vector: %d)\n",
i, irq->affinity_hint_mask.bits[0], irq->vector);
}
return rc;
err:
for (k = ENA_IO_IRQ_FIRST_IDX; k < i; k++) {
irq = &adapter->irq_tbl[k];
free_irq(irq->vector, irq->data);
}
return rc;
}
static void ena_free_mgmnt_irq(struct ena_adapter *adapter)
{
struct ena_irq *irq;
irq = &adapter->irq_tbl[ENA_MGMNT_IRQ_IDX];
synchronize_irq(irq->vector);
irq_set_affinity_hint(irq->vector, NULL);
free_irq(irq->vector, irq->data);
}
static void ena_free_io_irq(struct ena_adapter *adapter)
{
u32 io_queue_count = adapter->num_io_queues + adapter->xdp_num_queues;
struct ena_irq *irq;
int i;
#ifdef CONFIG_RFS_ACCEL
if (adapter->msix_vecs >= 1) {
free_irq_cpu_rmap(adapter->netdev->rx_cpu_rmap);
adapter->netdev->rx_cpu_rmap = NULL;
}
#endif /* CONFIG_RFS_ACCEL */
for (i = ENA_IO_IRQ_FIRST_IDX; i < ENA_MAX_MSIX_VEC(io_queue_count); i++) {
irq = &adapter->irq_tbl[i];
irq_set_affinity_hint(irq->vector, NULL);
free_irq(irq->vector, irq->data);
}
}
static void ena_disable_msix(struct ena_adapter *adapter)
{
#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 8, 0)
if (test_and_clear_bit(ENA_FLAG_MSIX_ENABLED, &adapter->flags))
pci_disable_msix(adapter->pdev);
if (adapter->msix_entries)
vfree(adapter->msix_entries);
adapter->msix_entries = NULL;
#else
if (test_and_clear_bit(ENA_FLAG_MSIX_ENABLED, &adapter->flags))
pci_free_irq_vectors(adapter->pdev);
#endif
}
static void ena_disable_io_intr_sync(struct ena_adapter *adapter)
{
u32 io_queue_count = adapter->num_io_queues + adapter->xdp_num_queues;
int i;
if (!netif_running(adapter->netdev))
return;
for (i = ENA_IO_IRQ_FIRST_IDX; i < ENA_MAX_MSIX_VEC(io_queue_count); i++)
synchronize_irq(adapter->irq_tbl[i].vector);
}
static void ena_del_napi_in_range(struct ena_adapter *adapter,
int first_index,
int count)
{
int i;
for (i = first_index; i < first_index + count; i++) {
#ifdef ENA_BUSY_POLL_SUPPORT
napi_hash_del(&adapter->ena_napi[i].napi);
#endif /* ENA_BUSY_POLL_SUPPORT */
netif_napi_del(&adapter->ena_napi[i].napi);
#ifdef ENA_XDP_SUPPORT
WARN_ON(ENA_IS_XDP_INDEX(adapter, i) &&
adapter->ena_napi[i].rx_ring);
#endif /* ENA_XDP_SUPPORT */
}
#ifdef ENA_BUSY_POLL_SUPPORT
/* Wait until all uses of napi struct complete */
synchronize_net();
#endif /* ENA_BUSY_POLL_SUPPORT */
}
static void ena_init_napi_in_range(struct ena_adapter *adapter,
int first_index, int count)
{
int i;
int (*napi_handler)(struct napi_struct *napi, int budget);
for (i = first_index; i < first_index + count; i++) {
struct ena_napi *napi = &adapter->ena_napi[i];
struct ena_ring *rx_ring, *tx_ring;
memset(napi, 0, sizeof(*napi));
rx_ring = &adapter->rx_ring[i];
tx_ring = &adapter->tx_ring[i];
napi_handler = ena_io_poll;
#ifdef ENA_XDP_SUPPORT
if (ENA_IS_XDP_INDEX(adapter, i) || ENA_IS_XSK_RING(rx_ring))
napi_handler = ena_xdp_io_poll;
#endif /* ENA_XDP_SUPPORT */
netif_napi_add(adapter->netdev,
&napi->napi,
napi_handler,
ENA_NAPI_BUDGET);
#ifdef ENA_BUSY_POLL_SUPPORT
napi_hash_add(&adapter->ena_napi[i].napi);
#endif /* ENA_BUSY_POLL_SUPPORT */
if (!ENA_IS_XDP_INDEX(adapter, i))
napi->rx_ring = rx_ring;
napi->tx_ring = tx_ring;
napi->qid = i;
}
}
#ifdef ENA_BUSY_POLL_SUPPORT
static void ena_napi_disable_in_range(struct ena_adapter *adapter,
int first_index,
int count)
{
struct ena_ring *rx_ring;
int i, timeout;
for (i = first_index; i < first_index + count; i++) {
napi_disable(&adapter->ena_napi[i].napi);
rx_ring = &adapter->rx_ring[i];
timeout = 1000;
while (!ena_bp_disable(rx_ring)) {
netif_info(adapter, ifdown, adapter->netdev,
"Rx queue %d locked\n", i);
usleep_range(1000, 2000);
timeout--;
if (!timeout) {
WARN(!ena_bp_disable(rx_ring),
"Unable to disable busy poll at ring %d\n", i);
break;
}
}
}
}
#else
static void ena_napi_disable_in_range(struct ena_adapter *adapter,
int first_index,
int count)
{
int i;
for (i = first_index; i < first_index + count; i++)
napi_disable(&adapter->ena_napi[i].napi);
}
#endif
static void ena_napi_enable_in_range(struct ena_adapter *adapter,
int first_index,
int count)
{
int i;
for (i = first_index; i < first_index + count; i++)
napi_enable(&adapter->ena_napi[i].napi);
}
/* Configure the Rx forwarding */
static int ena_rss_configure(struct ena_adapter *adapter)
{
struct ena_com_dev *ena_dev = adapter->ena_dev;
int rc;
/* In case the RSS table wasn't initialized by probe */
if (!ena_dev->rss.tbl_log_size) {
rc = ena_rss_init_default(adapter);
if (rc && (rc != -EOPNOTSUPP)) {
netif_err(adapter, ifup, adapter->netdev,
"Failed to init RSS rc: %d\n", rc);
return rc;
}
}
/* Set indirect table */
rc = ena_com_indirect_table_set(ena_dev);
if (unlikely(rc && rc != -EOPNOTSUPP))
return rc;
/* Configure hash function (if supported) */
rc = ena_com_set_hash_function(ena_dev);
if (unlikely(rc && (rc != -EOPNOTSUPP)))
return rc;
/* Configure hash inputs (if supported) */
rc = ena_com_set_hash_ctrl(ena_dev);
if (unlikely(rc && (rc != -EOPNOTSUPP)))
return rc;
return 0;
}
static int ena_up_complete(struct ena_adapter *adapter)
{
int rc;
rc = ena_rss_configure(adapter);
if (rc)
return rc;
ena_change_mtu(adapter->netdev, adapter->netdev->mtu);
ena_refill_all_rx_bufs(adapter);
/* enable transmits */
netif_tx_start_all_queues(adapter->netdev);
ena_napi_enable_in_range(adapter,
0,
adapter->xdp_num_queues + adapter->num_io_queues);
return 0;
}
static int ena_create_io_tx_queue(struct ena_adapter *adapter, int qid)
{
struct ena_com_create_io_ctx ctx;
struct ena_com_dev *ena_dev;
struct ena_ring *tx_ring;
u32 msix_vector;
u16 ena_qid;
int rc;
ena_dev = adapter->ena_dev;
tx_ring = &adapter->tx_ring[qid];
msix_vector = ENA_IO_IRQ_IDX(qid);
ena_qid = ENA_IO_TXQ_IDX(qid);
memset(&ctx, 0x0, sizeof(ctx));
ctx.direction = ENA_COM_IO_QUEUE_DIRECTION_TX;
ctx.qid = ena_qid;
ctx.mem_queue_type = ena_dev->tx_mem_queue_type;
ctx.msix_vector = msix_vector;
ctx.queue_size = tx_ring->ring_size;
ctx.numa_node = tx_ring->numa_node;
rc = ena_com_create_io_queue(ena_dev, &ctx);
if (rc) {
netif_err(adapter, ifup, adapter->netdev,
"Failed to create I/O TX queue num %d rc: %d\n",
qid, rc);
return rc;
}
rc = ena_com_get_io_handlers(ena_dev, ena_qid,
&tx_ring->ena_com_io_sq,
&tx_ring->ena_com_io_cq);
if (rc) {
netif_err(adapter, ifup, adapter->netdev,
"Failed to get TX queue handlers. TX queue num %d rc: %d\n",
qid, rc);
ena_com_destroy_io_queue(ena_dev, ena_qid);
return rc;
}
ena_com_update_numa_node(tx_ring->ena_com_io_cq, ctx.numa_node);
return rc;
}
int ena_create_io_tx_queues_in_range(struct ena_adapter *adapter,
int first_index, int count)
{
struct ena_com_dev *ena_dev = adapter->ena_dev;
int rc, i;
for (i = first_index; i < first_index + count; i++) {
rc = ena_create_io_tx_queue(adapter, i);
if (rc)
goto create_err;
}
return 0;
create_err:
while (i-- > first_index)
ena_com_destroy_io_queue(ena_dev, ENA_IO_TXQ_IDX(i));
return rc;
}
static int ena_create_io_rx_queue(struct ena_adapter *adapter, int qid)
{
struct ena_com_dev *ena_dev;
struct ena_com_create_io_ctx ctx;
struct ena_ring *rx_ring;
u32 msix_vector;
u16 ena_qid;
int rc;
ena_dev = adapter->ena_dev;
rx_ring = &adapter->rx_ring[qid];
msix_vector = ENA_IO_IRQ_IDX(qid);
ena_qid = ENA_IO_RXQ_IDX(qid);
memset(&ctx, 0x0, sizeof(ctx));
ctx.qid = ena_qid;
ctx.direction = ENA_COM_IO_QUEUE_DIRECTION_RX;
ctx.mem_queue_type = ENA_ADMIN_PLACEMENT_POLICY_HOST;
ctx.msix_vector = msix_vector;
ctx.queue_size = rx_ring->ring_size;
ctx.numa_node = rx_ring->numa_node;
rc = ena_com_create_io_queue(ena_dev, &ctx);
if (rc) {
netif_err(adapter, ifup, adapter->netdev,
"Failed to create I/O RX queue num %d rc: %d\n",
qid, rc);
return rc;
}
rc = ena_com_get_io_handlers(ena_dev, ena_qid,
&rx_ring->ena_com_io_sq,
&rx_ring->ena_com_io_cq);
if (rc) {
netif_err(adapter, ifup, adapter->netdev,
"Failed to get RX queue handlers. RX queue num %d rc: %d\n",
qid, rc);
goto err;
}
ena_com_update_numa_node(rx_ring->ena_com_io_cq, ctx.numa_node);
return rc;
err:
ena_com_destroy_io_queue(ena_dev, ena_qid);
return rc;
}
static int ena_create_all_io_rx_queues(struct ena_adapter *adapter)
{
struct ena_com_dev *ena_dev = adapter->ena_dev;
int rc, i;
for (i = 0; i < adapter->num_io_queues; i++) {
rc = ena_create_io_rx_queue(adapter, i);
if (rc)
goto create_err;
INIT_WORK(&adapter->ena_napi[i].dim.work, ena_dim_work);
ena_xdp_register_rxq_info(&adapter->rx_ring[i]);
}
return 0;
create_err:
while (i--) {
ena_xdp_unregister_rxq_info(&adapter->rx_ring[i]);
cancel_work_sync(&adapter->ena_napi[i].dim.work);
ena_com_destroy_io_queue(ena_dev, ENA_IO_RXQ_IDX(i));
}
return rc;
}
static void set_io_rings_size(struct ena_adapter *adapter,
int new_tx_size,
int new_rx_size)
{
int i;
for (i = 0; i < adapter->num_io_queues; i++) {
adapter->tx_ring[i].ring_size = new_tx_size;
adapter->rx_ring[i].ring_size = new_rx_size;
}
}
/* This function allows queue allocation to backoff when the system is
* low on memory. If there is not enough memory to allocate io queues
* the driver will try to allocate smaller queues.
*
* The backoff algorithm is as follows:
* 1. Try to allocate TX and RX and if successful.
* 1.1. return success
*
* 2. Divide by 2 the size of the larger of RX and TX queues (or both if their size is the same).
*
* 3. If TX or RX is smaller than 256
* 3.1. return failure.
* 4. else
* 4.1. go back to 1.
*/
static int create_queues_with_size_backoff(struct ena_adapter *adapter)
{
int rc, cur_rx_ring_size, cur_tx_ring_size;
int new_rx_ring_size, new_tx_ring_size;
/* current queue sizes might be set to smaller than the requested
* ones due to past queue allocation failures.
*/
set_io_rings_size(adapter, adapter->requested_tx_ring_size,
adapter->requested_rx_ring_size);
while (1) {
#ifdef ENA_XDP_SUPPORT
if (ena_xdp_present(adapter)) {
rc = ena_setup_and_create_all_xdp_queues(adapter);
if (rc)
goto err_setup_tx;
}
#endif /* ENA_XDP_SUPPORT */
rc = ena_setup_tx_resources_in_range(adapter,
0,
adapter->num_io_queues);
if (rc)
goto err_setup_tx;
rc = ena_create_io_tx_queues_in_range(adapter,
0,
adapter->num_io_queues);
if (rc)
goto err_create_tx_queues;
rc = ena_setup_all_rx_resources(adapter);
if (rc)
goto err_setup_rx;
rc = ena_create_all_io_rx_queues(adapter);
if (rc)
goto err_create_rx_queues;
rc = ena_create_page_caches(adapter);
if (rc) /* Cache memory is freed in case of failure */
goto err_create_rx_queues;
return 0;
err_create_rx_queues:
ena_free_all_io_rx_resources(adapter);
err_setup_rx:
ena_destroy_all_tx_queues(adapter);
err_create_tx_queues:
ena_free_all_io_tx_resources(adapter);
err_setup_tx:
if (rc != -ENOMEM) {
netif_err(adapter, ifup, adapter->netdev,
"Queue creation failed with error code %d\n",
rc);
return rc;
}
cur_tx_ring_size = adapter->tx_ring[0].ring_size;
cur_rx_ring_size = adapter->rx_ring[0].ring_size;
netif_err(adapter, ifup, adapter->netdev,
"Not enough memory to create queues with sizes TX=%d, RX=%d\n",
cur_tx_ring_size, cur_rx_ring_size);
new_tx_ring_size = cur_tx_ring_size;
new_rx_ring_size = cur_rx_ring_size;
/* Decrease the size of the larger queue, or
* decrease both if they are the same size.
*/
if (cur_rx_ring_size <= cur_tx_ring_size)
new_tx_ring_size = cur_tx_ring_size / 2;
if (cur_rx_ring_size >= cur_tx_ring_size)
new_rx_ring_size = cur_rx_ring_size / 2;
if (new_tx_ring_size < ENA_MIN_RING_SIZE ||
new_rx_ring_size < ENA_MIN_RING_SIZE) {
netif_err(adapter, ifup, adapter->netdev,
"Queue creation failed with the smallest possible queue size of %d for both queues. Not retrying with smaller queues\n",
ENA_MIN_RING_SIZE);
return rc;
}
netif_err(adapter, ifup, adapter->netdev,
"Retrying queue creation with sizes TX=%d, RX=%d\n",
new_tx_ring_size,
new_rx_ring_size);
set_io_rings_size(adapter, new_tx_ring_size,
new_rx_ring_size);
}
}
int ena_up(struct ena_adapter *adapter)
{
int io_queue_count, rc, i;
netif_dbg(adapter, ifup, adapter->netdev, "%s\n", __func__);
io_queue_count = adapter->num_io_queues + adapter->xdp_num_queues;
ena_setup_io_intr(adapter);
/* napi poll functions should be initialized before running
* request_irq(), to handle a rare condition where there is a pending
* interrupt, causing the ISR to fire immediately while the poll
* function wasn't set yet, causing a null dereference
*/
ena_init_napi_in_range(adapter, 0, io_queue_count);
#ifdef CONFIG_ARM64
/* enable DIM by default on ARM machines, also needs to happen
* before enabling IRQs since DIM is ran from napi routine
*/
if (ena_com_interrupt_moderation_supported(adapter->ena_dev))
ena_com_enable_adaptive_moderation(adapter->ena_dev);
#endif
rc = ena_request_io_irq(adapter);
if (rc)
goto err_req_irq;
rc = create_queues_with_size_backoff(adapter);
if (rc)
goto err_create_queues_with_backoff;
rc = ena_up_complete(adapter);
if (rc)
goto err_up;
if (test_bit(ENA_FLAG_LINK_UP, &adapter->flags))
netif_carrier_on(adapter->netdev);
ena_increase_stat(&adapter->dev_stats.interface_up, 1,
&adapter->syncp);
set_bit(ENA_FLAG_DEV_UP, &adapter->flags);
/* Enable completion queues interrupt */
for (i = 0; i < adapter->num_io_queues; i++)
ena_unmask_interrupt(&adapter->tx_ring[i],
&adapter->rx_ring[i]);
/* schedule napi in case we had pending packets
* from the last time we disable napi
*/
for (i = 0; i < io_queue_count; i++)
napi_schedule(&adapter->ena_napi[i].napi);
return rc;
err_up:
ena_free_page_caches(adapter);
ena_destroy_all_tx_queues(adapter);
ena_free_all_io_tx_resources(adapter);
ena_destroy_all_rx_queues(adapter);
ena_free_all_io_rx_resources(adapter);
err_create_queues_with_backoff:
ena_free_io_irq(adapter);
err_req_irq:
ena_del_napi_in_range(adapter, 0, io_queue_count);
return rc;
}
void ena_down(struct ena_adapter *adapter)
{
int io_queue_count = adapter->num_io_queues + adapter->xdp_num_queues;
netif_info(adapter, ifdown, adapter->netdev, "%s\n", __func__);
clear_bit(ENA_FLAG_DEV_UP, &adapter->flags);
ena_increase_stat(&adapter->dev_stats.interface_down, 1,
&adapter->syncp);
netif_carrier_off(adapter->netdev);
netif_tx_disable(adapter->netdev);
/* After this point the napi handler won't enable the tx queue */
ena_napi_disable_in_range(adapter, 0, io_queue_count);
/* After destroy the queue there won't be any new interrupts */
if (test_bit(ENA_FLAG_TRIGGER_RESET, &adapter->flags)) {
int rc;
rc = ena_com_dev_reset(adapter->ena_dev, adapter->reset_reason);
if (rc)
netif_err(adapter, ifdown, adapter->netdev,
"Device reset failed\n");
/* stop submitting admin commands on a device that was reset */
ena_com_set_admin_running_state(adapter->ena_dev, false);
}
ena_destroy_all_io_queues(adapter);
ena_disable_io_intr_sync(adapter);
ena_free_io_irq(adapter);
ena_del_napi_in_range(adapter, 0, io_queue_count);
ena_free_all_tx_bufs(adapter);
ena_free_all_rx_bufs(adapter);
ena_free_all_cache_pages(adapter);
ena_free_page_caches(adapter);
ena_free_all_io_tx_resources(adapter);
ena_free_all_io_rx_resources(adapter);
}
/* ena_open - Called when a network interface is made active
* @netdev: network interface device structure
*
* Returns 0 on success, negative value on failure
*
* The open entry point is called when a network interface is made
* active by the system (IFF_UP). At this point all resources needed
* for transmit and receive operations are allocated, the interrupt
* handler is registered with the OS, the watchdog timer is started,
* and the stack is notified that the interface is ready.
*/
static int ena_open(struct net_device *netdev)
{
struct ena_adapter *adapter = netdev_priv(netdev);
int rc;
/* Notify the stack of the actual queue counts. */
rc = netif_set_real_num_tx_queues(netdev, adapter->num_io_queues);
if (rc) {
netif_err(adapter, ifup, netdev, "Can't set num tx queues\n");
return rc;
}
rc = netif_set_real_num_rx_queues(netdev, adapter->num_io_queues);
if (rc) {
netif_err(adapter, ifup, netdev, "Can't set num rx queues\n");
return rc;
}
rc = ena_up(adapter);
if (rc)
return rc;
return rc;
}
/* ena_close - Disables a network interface
* @netdev: network interface device structure
*
* Returns 0, this is not allowed to fail
*
* The close entry point is called when an interface is de-activated
* by the OS. The hardware is still under the drivers control, but
* needs to be disabled. A global MAC reset is issued to stop the
* hardware, and all transmit and receive resources are freed.
*/
static int ena_close(struct net_device *netdev)
{
struct ena_adapter *adapter = netdev_priv(netdev);
netif_dbg(adapter, ifdown, netdev, "%s\n", __func__);
if (!test_bit(ENA_FLAG_DEVICE_RUNNING, &adapter->flags))
return 0;
if (test_bit(ENA_FLAG_DEV_UP, &adapter->flags))
ena_down(adapter);
/* Check for device status and issue reset if needed*/
check_for_admin_com_state(adapter);
if (unlikely(test_bit(ENA_FLAG_TRIGGER_RESET, &adapter->flags))) {
netif_err(adapter, ifdown, adapter->netdev,
"Destroy failure, restarting device\n");
ena_dump_stats_to_dmesg(adapter);
/* rtnl lock already obtained in dev_ioctl() layer */
ena_destroy_device(adapter, false);
ena_restore_device(adapter);
}
return 0;
}
int ena_set_lpc_state(struct ena_adapter *adapter, bool enabled)
{
/* In XDP, lpc_size might be positive even with LPC disabled, use cache
* pointer instead.
*/
struct ena_page_cache *page_cache = adapter->rx_ring->page_cache;
/* Exit early if LPC state doesn't change */
if (enabled == !!page_cache)
return 0;
if (enabled && !ena_is_lpc_supported(adapter, adapter->rx_ring, true))
return -EOPNOTSUPP;
adapter->used_lpc_size = enabled ? adapter->configured_lpc_size : 0;
/* rtnl lock is already obtained in dev_ioctl() layer, so it's safe to
* re-initialize IO resources.
*/
if (test_bit(ENA_FLAG_DEV_UP, &adapter->flags)) {
ena_close(adapter->netdev);
ena_up(adapter);
}
return 0;
}
int ena_update_queue_sizes(struct ena_adapter *adapter,
u32 new_tx_size,
u32 new_rx_size)
{
bool dev_was_up;
dev_was_up = test_bit(ENA_FLAG_DEV_UP, &adapter->flags);
ena_close(adapter->netdev);
adapter->requested_tx_ring_size = new_tx_size;
adapter->requested_rx_ring_size = new_rx_size;
ena_init_io_rings(adapter,
0,
adapter->xdp_num_queues +
adapter->num_io_queues);
return dev_was_up ? ena_up(adapter) : 0;
}
int ena_update_queue_count(struct ena_adapter *adapter, u32 new_channel_count)
{
struct ena_com_dev *ena_dev = adapter->ena_dev;
#ifdef ENA_XDP_SUPPORT
int prev_channel_count;
#endif /* ENA_XDP_SUPPORT */
bool dev_was_up;
dev_was_up = test_bit(ENA_FLAG_DEV_UP, &adapter->flags);
ena_close(adapter->netdev);
#ifdef ENA_XDP_SUPPORT
prev_channel_count = adapter->num_io_queues;
#endif /* ENA_XDP_SUPPORT */
adapter->num_io_queues = new_channel_count;
#ifdef ENA_XDP_SUPPORT
if (ena_xdp_present(adapter) &&
ena_xdp_allowed(adapter) == ENA_XDP_ALLOWED) {
adapter->xdp_first_ring = new_channel_count;
adapter->xdp_num_queues = new_channel_count;
if (prev_channel_count > new_channel_count)
ena_xdp_exchange_program_rx_in_range(adapter,
NULL,
new_channel_count,
prev_channel_count);
else
ena_xdp_exchange_program_rx_in_range(adapter,
adapter->xdp_bpf_prog,
prev_channel_count,
new_channel_count);
}
#endif /* ENA_XDP_SUPPORT */
/* We need to destroy the rss table so that the indirection
* table will be reinitialized by ena_up()
*/
ena_com_rss_destroy(ena_dev);
ena_init_io_rings(adapter,
0,
adapter->xdp_num_queues +
adapter->num_io_queues);
return dev_was_up ? ena_open(adapter->netdev) : 0;
}
static void ena_tx_csum(struct ena_com_tx_ctx *ena_tx_ctx,
struct sk_buff *skb,
bool disable_meta_caching)
{
u32 mss = skb_shinfo(skb)->gso_size;
struct ena_com_tx_meta *ena_meta = &ena_tx_ctx->ena_meta;
u8 l4_protocol = 0;
if ((skb->ip_summed == CHECKSUM_PARTIAL) || mss) {
ena_tx_ctx->l4_csum_enable = 1;
if (mss) {
ena_tx_ctx->tso_enable = 1;
ena_meta->l4_hdr_len = tcp_hdr(skb)->doff;
ena_tx_ctx->l4_csum_partial = 0;
} else {
ena_tx_ctx->tso_enable = 0;
ena_meta->l4_hdr_len = 0;
ena_tx_ctx->l4_csum_partial = 1;
}
switch (ip_hdr(skb)->version) {
case IPVERSION:
ena_tx_ctx->l3_proto = ENA_ETH_IO_L3_PROTO_IPV4;
if (ip_hdr(skb)->frag_off & htons(IP_DF))
ena_tx_ctx->df = 1;
if (mss)
ena_tx_ctx->l3_csum_enable = 1;
l4_protocol = ip_hdr(skb)->protocol;
break;
case 6:
ena_tx_ctx->l3_proto = ENA_ETH_IO_L3_PROTO_IPV6;
l4_protocol = ipv6_hdr(skb)->nexthdr;
break;
default:
break;
}
if (l4_protocol == IPPROTO_TCP)
ena_tx_ctx->l4_proto = ENA_ETH_IO_L4_PROTO_TCP;
else
ena_tx_ctx->l4_proto = ENA_ETH_IO_L4_PROTO_UDP;
ena_meta->mss = mss;
ena_meta->l3_hdr_len = skb_network_header_len(skb);
ena_meta->l3_hdr_offset = skb_network_offset(skb);
ena_tx_ctx->meta_valid = 1;
} else if (disable_meta_caching) {
memset(ena_meta, 0, sizeof(*ena_meta));
ena_tx_ctx->meta_valid = 1;
} else {
ena_tx_ctx->meta_valid = 0;
}
}
static int ena_check_and_linearize_skb(struct ena_ring *tx_ring,
struct sk_buff *skb)
{
int num_frags, header_len, rc;
num_frags = skb_shinfo(skb)->nr_frags;
header_len = skb_headlen(skb);
if (num_frags < tx_ring->sgl_size)
return 0;
if ((num_frags == tx_ring->sgl_size) &&
(header_len < tx_ring->tx_max_header_size))
return 0;
ena_increase_stat(&tx_ring->tx_stats.linearize, 1, &tx_ring->syncp);
rc = skb_linearize(skb);
if (unlikely(rc)) {
ena_increase_stat(&tx_ring->tx_stats.linearize_failed, 1,
&tx_ring->syncp);
}
return rc;
}
static int ena_tx_map_skb(struct ena_ring *tx_ring,
struct ena_tx_buffer *tx_info,
struct sk_buff *skb,
void **push_hdr,
u16 *header_len)
{
struct ena_adapter *adapter = tx_ring->adapter;
struct ena_com_buf *ena_buf;
dma_addr_t dma;
u32 skb_head_len, frag_len, last_frag;
u16 push_len = 0;
u16 delta = 0;
int i = 0;
skb_head_len = skb_headlen(skb);
tx_info->skb = skb;
ena_buf = tx_info->bufs;
if (tx_ring->tx_mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_DEV) {
/* When the device is LLQ mode, the driver will copy
* the header into the device memory space.
* the ena_com layer assume the header is in a linear
* memory space.
* This assumption might be wrong since part of the header
* can be in the fragmented buffers.
* Use skb_header_pointer to make sure the header is in a
* linear memory space.
*/
push_len = min_t(u32, skb->len, tx_ring->tx_max_header_size);
*push_hdr = skb_header_pointer(skb, 0, push_len,
tx_ring->push_buf_intermediate_buf);
*header_len = push_len;
if (unlikely(skb->data != *push_hdr)) {
ena_increase_stat(&tx_ring->tx_stats.llq_buffer_copy, 1,
&tx_ring->syncp);
delta = push_len - skb_head_len;
}
} else {
*push_hdr = NULL;
*header_len = min_t(u32, skb_head_len,
tx_ring->tx_max_header_size);
}
netif_dbg(adapter, tx_queued, adapter->netdev,
"skb: %p header_buf->vaddr: %p push_len: %d\n", skb,
*push_hdr, push_len);
if (skb_head_len > push_len) {
dma = dma_map_single(tx_ring->dev, skb->data + push_len,
skb_head_len - push_len, DMA_TO_DEVICE);
if (unlikely(dma_mapping_error(tx_ring->dev, dma)))
goto error_report_dma_error;
ena_buf->paddr = dma;
ena_buf->len = skb_head_len - push_len;
ena_buf++;
tx_info->num_of_bufs++;
tx_info->map_linear_data = 1;
} else {
tx_info->map_linear_data = 0;
}
last_frag = skb_shinfo(skb)->nr_frags;
for (i = 0; i < last_frag; i++) {
const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
frag_len = skb_frag_size(frag);
if (unlikely(delta >= frag_len)) {
delta -= frag_len;
continue;
}
dma = skb_frag_dma_map(tx_ring->dev, frag, delta,
frag_len - delta, DMA_TO_DEVICE);
if (unlikely(dma_mapping_error(tx_ring->dev, dma)))
goto error_report_dma_error;
ena_buf->paddr = dma;
ena_buf->len = frag_len - delta;
ena_buf++;
tx_info->num_of_bufs++;
delta = 0;
}
return 0;
error_report_dma_error:
ena_increase_stat(&tx_ring->tx_stats.dma_mapping_err, 1,
&tx_ring->syncp);
netif_warn(adapter, tx_queued, adapter->netdev, "Failed to map skb\n");
tx_info->skb = NULL;
tx_info->num_of_bufs += i;
ena_unmap_tx_buff(tx_ring, tx_info);
return -EINVAL;
}
/* Called with netif_tx_lock. */
static netdev_tx_t ena_start_xmit(struct sk_buff *skb, struct net_device *dev)
{
struct ena_adapter *adapter = netdev_priv(dev);
struct ena_tx_buffer *tx_info;
struct ena_com_tx_ctx ena_tx_ctx;
struct ena_ring *tx_ring;
struct netdev_queue *txq;
void *push_hdr;
u16 next_to_use, req_id, header_len;
int qid, rc;
netif_dbg(adapter, tx_queued, dev, "%s skb %p\n", __func__, skb);
/* Determine which tx ring we will be placed on */
qid = skb_get_queue_mapping(skb);
tx_ring = &adapter->tx_ring[qid];
txq = netdev_get_tx_queue(dev, qid);
rc = ena_check_and_linearize_skb(tx_ring, skb);
if (unlikely(rc))
goto error_drop_packet;
skb_tx_timestamp(skb);
next_to_use = tx_ring->next_to_use;
req_id = tx_ring->free_ids[next_to_use];
tx_info = &tx_ring->tx_buffer_info[req_id];
tx_info->num_of_bufs = 0;
WARN(tx_info->skb, "SKB isn't NULL req_id %d\n", req_id);
rc = ena_tx_map_skb(tx_ring, tx_info, skb, &push_hdr, &header_len);
if (unlikely(rc))
goto error_drop_packet;
memset(&ena_tx_ctx, 0x0, sizeof(struct ena_com_tx_ctx));
ena_tx_ctx.ena_bufs = tx_info->bufs;
ena_tx_ctx.push_header = push_hdr;
ena_tx_ctx.num_bufs = tx_info->num_of_bufs;
ena_tx_ctx.req_id = req_id;
ena_tx_ctx.header_len = header_len;
/* set flags and meta data */
ena_tx_csum(&ena_tx_ctx, skb, tx_ring->disable_meta_caching);
rc = ena_xmit_common(adapter,
tx_ring,
tx_info,
&ena_tx_ctx,
next_to_use,
skb->len);
if (rc)
goto error_unmap_dma;
if (tx_ring->enable_bql)
netdev_tx_sent_queue(txq, skb->len);
/* stop the queue when no more space available, the packet can have up
* to sgl_size + 2. one for the meta descriptor and one for header
* (if the header is larger than tx_max_header_size).
*/
if (unlikely(!ena_com_sq_have_enough_space(tx_ring->ena_com_io_sq,
tx_ring->sgl_size + 2))) {
netif_dbg(adapter, tx_queued, dev, "%s stop queue %d\n",
__func__, qid);
netif_tx_stop_queue(txq);
ena_increase_stat(&tx_ring->tx_stats.queue_stop, 1,
&tx_ring->syncp);
/* There is a rare condition where this function decide to
* stop the queue but meanwhile clean_tx_irq updates
* next_to_completion and terminates.
* The queue will remain stopped forever.
* To solve this issue add a mb() to make sure that
* netif_tx_stop_queue() write is vissible before checking if
* there is additional space in the queue.
*/
smp_mb();
if (ena_com_sq_have_enough_space(tx_ring->ena_com_io_sq,
ENA_TX_WAKEUP_THRESH)) {
netif_tx_wake_queue(txq);
ena_increase_stat(&tx_ring->tx_stats.queue_wakeup, 1,
&tx_ring->syncp);
}
}
#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 18, 0)
#ifdef HAVE_NETDEV_XMIT_MORE
if (netif_xmit_stopped(txq) || !netdev_xmit_more())
#else
if (netif_xmit_stopped(txq) || !skb->xmit_more)
#endif /* HAVE_NETDEV_XMIT_MORE */
#endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(3, 18, 0) */
/* trigger the dma engine. ena_ring_tx_doorbell()
* calls a memory barrier inside it.
*/
ena_ring_tx_doorbell(tx_ring);
return NETDEV_TX_OK;
error_unmap_dma:
ena_unmap_tx_buff(tx_ring, tx_info);
tx_info->skb = NULL;
error_drop_packet:
dev_kfree_skb(skb);
return NETDEV_TX_OK;
}
#if defined HAVE_NDO_SELECT_QUEUE_ACCEL_FALLBACK_V3
static u16 ena_select_queue(struct net_device *dev, struct sk_buff *skb,
struct net_device *sb_dev)
#elif defined HAVE_NDO_SELECT_QUEUE_ACCEL_FALLBACK_V2
static u16 ena_select_queue(struct net_device *dev, struct sk_buff *skb,
struct net_device *sb_dev,
select_queue_fallback_t fallback)
#elif defined HAVE_NDO_SELECT_QUEUE_ACCEL_FALLBACK_V1
static u16 ena_select_queue(struct net_device *dev, struct sk_buff *skb,
void *accel_priv,
select_queue_fallback_t fallback)
#elif defined HAVE_NDO_SELECT_QUEUE_ACCEL
/* Return subqueue id on this core (one per core). */
static u16 ena_select_queue(struct net_device *dev, struct sk_buff *skb,
void *accel_priv)
#else
static u16 ena_select_queue(struct net_device *dev, struct sk_buff *skb)
#endif
{
u16 qid;
/* we suspect that this is good for in--kernel network services that
* want to loop incoming skb rx to tx in normal user generated traffic,
* most probably we will not get to this
*/
if (skb_rx_queue_recorded(skb))
qid = skb_get_rx_queue(skb);
else
#if (defined HAVE_NDO_SELECT_QUEUE_ACCEL_FALLBACK_V3)
qid = netdev_pick_tx(dev, skb, NULL);
#elif (defined HAVE_NDO_SELECT_QUEUE_ACCEL_FALLBACK_V2)
qid = fallback(dev, skb, NULL);
#elif (defined HAVE_NDO_SELECT_QUEUE_ACCEL_FALLBACK_V1)
qid = fallback(dev, skb);
#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3,9,0)
qid = __netdev_pick_tx(dev, skb);
#else
qid = skb_tx_hash(dev, skb);
#endif /* HAVE_NDO_SELECT_QUEUE_ACCEL_FALLBACK_V2 */
return qid;
}
#ifdef HAVE_SET_RX_MODE
/* Unicast, Multicast and Promiscuous mode set
* @netdev: network interface device structure
*
* The set_rx_mode entry point is called whenever the unicast or multicast
* address lists or the network interface flags are updated. This routine is
* responsible for configuring the hardware for proper unicast, multicast,
* promiscuous mode, and all-multi behavior.
*/
static void ena_set_rx_mode(struct net_device *netdev)
{
/* struct ena_adapter *adapter = netdev_priv(netdev); */
/* TODO set Rx mode */
if (netdev->flags & IFF_PROMISC) {
} else if (netdev->flags & IFF_ALLMULTI) {
} else if (netdev_mc_empty(netdev)) {
} else {
}
}
#endif /* HAVE_SET_RX_MODE */
static void ena_config_host_info(struct ena_com_dev *ena_dev, struct pci_dev *pdev)
{
struct device *dev = &pdev->dev;
struct ena_admin_host_info *host_info;
int rc;
/* Allocate only the host info */
rc = ena_com_allocate_host_info(ena_dev);
if (rc) {
dev_err(dev, "Cannot allocate host info\n");
return;
}
host_info = ena_dev->host_attr.host_info;
host_info->bdf = (pdev->bus->number << 8) | pdev->devfn;
host_info->os_type = ENA_ADMIN_OS_LINUX;
host_info->kernel_ver = LINUX_VERSION_CODE;
strlcpy(host_info->kernel_ver_str, utsname()->version,
sizeof(host_info->kernel_ver_str) - 1);
host_info->os_dist = 0;
strncpy(host_info->os_dist_str, utsname()->release,
sizeof(host_info->os_dist_str) - 1);
host_info->driver_version =
(DRV_MODULE_GEN_MAJOR) |
(DRV_MODULE_GEN_MINOR << ENA_ADMIN_HOST_INFO_MINOR_SHIFT) |
(DRV_MODULE_GEN_SUBMINOR << ENA_ADMIN_HOST_INFO_SUB_MINOR_SHIFT) |
("g"[0] << ENA_ADMIN_HOST_INFO_MODULE_TYPE_SHIFT);
host_info->num_cpus = num_online_cpus();
host_info->driver_supported_features =
ENA_ADMIN_HOST_INFO_RX_OFFSET_MASK |
ENA_ADMIN_HOST_INFO_INTERRUPT_MODERATION_MASK |
ENA_ADMIN_HOST_INFO_RX_BUF_MIRRORING_MASK |
ENA_ADMIN_HOST_INFO_RSS_CONFIGURABLE_FUNCTION_KEY_MASK |
ENA_ADMIN_HOST_INFO_RX_PAGE_REUSE_MASK;
rc = ena_com_set_host_attributes(ena_dev);
if (rc) {
if (rc == -EOPNOTSUPP)
dev_warn(dev, "Cannot set host attributes\n");
else
dev_err(dev, "Cannot set host attributes\n");
goto err;
}
return;
err:
ena_com_delete_host_info(ena_dev);
}
static void ena_config_debug_area(struct ena_adapter *adapter)
{
u32 debug_area_size;
int rc, ss_count;
ss_count = ena_get_sset_count(adapter->netdev, ETH_SS_STATS);
if (ss_count <= 0) {
netif_err(adapter, drv, adapter->netdev,
"SS count is negative\n");
return;
}
/* allocate 32 bytes for each string and 64bit for the value */
debug_area_size = ss_count * ETH_GSTRING_LEN + sizeof(u64) * ss_count;
rc = ena_com_allocate_debug_area(adapter->ena_dev, debug_area_size);
if (rc) {
netif_err(adapter, drv, adapter->netdev,
"Cannot allocate debug area\n");
return;
}
rc = ena_com_set_host_attributes(adapter->ena_dev);
if (rc) {
if (rc == -EOPNOTSUPP)
netif_warn(adapter, drv, adapter->netdev,
"Cannot set host attributes\n");
else
netif_err(adapter, drv, adapter->netdev,
"Cannot set host attributes\n");
goto err;
}
return;
err:
ena_com_delete_debug_area(adapter->ena_dev);
}
int ena_update_hw_stats(struct ena_adapter *adapter)
{
int rc;
rc = ena_com_get_eni_stats(adapter->ena_dev, &adapter->eni_stats);
if (rc) {
netdev_err(adapter->netdev, "Failed to get ENI stats\n");
return rc;
}
return 0;
}
#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36))
#ifdef NDO_GET_STATS_64_V2
static void ena_get_stats64(struct net_device *netdev,
struct rtnl_link_stats64 *stats)
#else
static struct rtnl_link_stats64 *ena_get_stats64(struct net_device *netdev,
struct rtnl_link_stats64 *stats)
#endif
{
struct ena_adapter *adapter = netdev_priv(netdev);
struct ena_ring *rx_ring, *tx_ring;
unsigned int start;
u64 rx_drops;
u64 tx_drops;
int i;
if (!test_bit(ENA_FLAG_DEV_UP, &adapter->flags))
#ifdef NDO_GET_STATS_64_V2
return;
#else
return NULL;
#endif
for (i = 0; i < adapter->num_io_queues; i++) {
u64 bytes, packets;
tx_ring = &adapter->tx_ring[i];
do {
start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
packets = tx_ring->tx_stats.cnt;
bytes = tx_ring->tx_stats.bytes;
} while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
stats->tx_packets += packets;
stats->tx_bytes += bytes;
rx_ring = &adapter->rx_ring[i];
do {
start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
packets = rx_ring->rx_stats.cnt;
bytes = rx_ring->rx_stats.bytes;
} while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
stats->rx_packets += packets;
stats->rx_bytes += bytes;
}
do {
start = u64_stats_fetch_begin_irq(&adapter->syncp);
rx_drops = adapter->dev_stats.rx_drops;
tx_drops = adapter->dev_stats.tx_drops;
} while (u64_stats_fetch_retry_irq(&adapter->syncp, start));
stats->rx_dropped = rx_drops;
stats->tx_dropped = tx_drops;
stats->multicast = 0;
stats->collisions = 0;
stats->rx_length_errors = 0;
stats->rx_crc_errors = 0;
stats->rx_frame_errors = 0;
stats->rx_fifo_errors = 0;
stats->rx_missed_errors = 0;
stats->tx_window_errors = 0;
stats->rx_errors = 0;
stats->tx_errors = 0;
#ifndef NDO_GET_STATS_64_V2
return stats;
#endif
}
#else /* kernel > 2.6.36 */