// SPDX-License-Identifier: GPL-2.0-only /**************************************************************************** * Driver for Solarflare network controllers and boards * Copyright 2018 Solarflare Communications Inc. * * This program is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License version 2 as published * by the Free Software Foundation, incorporated herein by reference.
*/
/* This is the percentage fill level below which new RX descriptors * will be added to the RX descriptor ring.
*/ staticunsignedint rx_refill_threshold;
module_param(rx_refill_threshold, uint, 0444);
MODULE_PARM_DESC(rx_refill_threshold, "RX descriptor ring refill threshold (%)");
/* RX maximum head room required. * * This must be at least 1 to prevent overflow, plus one packet-worth * to allow pipelined receives.
*/ #define EFX_RXD_HEAD_ROOM (1 + EFX_RX_MAX_FRAGS)
/* Check the RX page recycle ring for a page that can be reused. */ staticstruct page *efx_reuse_page(struct efx_rx_queue *rx_queue)
{ struct efx_nic *efx = rx_queue->efx; struct efx_rx_page_state *state; unsignedint index; struct page *page;
if (unlikely(!rx_queue->page_ring)) return NULL;
index = rx_queue->page_remove & rx_queue->page_ptr_mask;
page = rx_queue->page_ring[index]; if (page == NULL) return NULL;
/* If page_count is 1 then we hold the only reference to this page. */ if (page_count(page) == 1) {
++rx_queue->page_recycle_count; return page;
} else {
state = page_address(page);
dma_unmap_page(&efx->pci_dev->dev, state->dma_addr,
PAGE_SIZE << efx->rx_buffer_order,
DMA_FROM_DEVICE);
put_page(page);
++rx_queue->page_recycle_failed;
}
return NULL;
}
/* Attempt to recycle the page if there is an RX recycle ring; the page can * only be added if this is the final RX buffer, to prevent pages being used in * the descriptor ring and appearing in the recycle ring simultaneously.
*/ staticvoid efx_recycle_rx_page(struct efx_channel *channel, struct efx_rx_buffer *rx_buf)
{ struct efx_rx_queue *rx_queue = efx_channel_get_rx_queue(channel); struct efx_nic *efx = rx_queue->efx; struct page *page = rx_buf->page; unsignedint index;
/* Only recycle the page after processing the final buffer. */ if (!(rx_buf->flags & EFX_RX_BUF_LAST_IN_PAGE)) return;
index = rx_queue->page_add & rx_queue->page_ptr_mask; if (rx_queue->page_ring[index] == NULL) { unsignedint read_index = rx_queue->page_remove &
rx_queue->page_ptr_mask;
/* The next slot in the recycle ring is available, but * increment page_remove if the read pointer currently * points here.
*/ if (read_index == index)
++rx_queue->page_remove;
rx_queue->page_ring[index] = page;
++rx_queue->page_add; return;
}
++rx_queue->page_recycle_full;
efx_unmap_rx_buffer(efx, rx_buf);
put_page(rx_buf->page);
}
/* Recycle the pages that are used by buffers that have just been received. */ void efx_recycle_rx_pages(struct efx_channel *channel, struct efx_rx_buffer *rx_buf, unsignedint n_frags)
{ struct efx_rx_queue *rx_queue = efx_channel_get_rx_queue(channel);
if (unlikely(!rx_queue->page_ring)) return;
do {
efx_recycle_rx_page(channel, rx_buf);
rx_buf = efx_rx_buf_next(rx_queue, rx_buf);
} while (--n_frags);
}
/* Unmap and release the pages in the recycle ring. Remove the ring. */ for (i = 0; i <= rx_queue->page_ptr_mask; i++) { struct page *page = rx_queue->page_ring[i]; struct efx_rx_page_state *state;
staticvoid efx_fini_rx_buffer(struct efx_rx_queue *rx_queue, struct efx_rx_buffer *rx_buf)
{ /* Release the page reference we hold for the buffer. */ if (rx_buf->page)
put_page(rx_buf->page);
/* If this is the last buffer in a page, unmap and free it. */ if (rx_buf->flags & EFX_RX_BUF_LAST_IN_PAGE) {
efx_unmap_rx_buffer(rx_queue->efx, rx_buf);
efx_free_rx_buffers(rx_queue, rx_buf, 1);
}
rx_buf->page = NULL;
}
int efx_probe_rx_queue(struct efx_rx_queue *rx_queue)
{ struct efx_nic *efx = rx_queue->efx; unsignedint entries; int rc;
/* Create the smallest power-of-two aligned ring */
entries = max(roundup_pow_of_two(efx->rxq_entries), EFX_MIN_DMAQ_SIZE);
EFX_WARN_ON_PARANOID(entries > EFX_MAX_DMAQ_SIZE);
rx_queue->ptr_mask = entries - 1;
if (rc) {
netif_err(efx, rx_err, efx->net_dev, "Failure to initialise XDP queue information rc=%d\n",
rc);
efx->xdp_rxq_info_failed = true;
}
/* Set up RX descriptor ring */
efx_nic_init_rx(rx_queue);
}
void efx_fini_rx_queue(struct efx_rx_queue *rx_queue)
{ struct efx_rx_buffer *rx_buf; int i;
netif_dbg(rx_queue->efx, drv, rx_queue->efx->net_dev, "shutting down RX queue %d\n", efx_rx_queue_index(rx_queue));
timer_delete_sync(&rx_queue->slow_fill); if (rx_queue->grant_credits)
flush_work(&rx_queue->grant_work);
/* Release RX buffers from the current read ptr to the write ptr */ if (rx_queue->buffer) { for (i = rx_queue->removed_count; i < rx_queue->added_count;
i++) { unsignedint index = i & rx_queue->ptr_mask;
/* Unmap a DMA-mapped page. This function is only called for the final RX * buffer in a page.
*/ void efx_unmap_rx_buffer(struct efx_nic *efx, struct efx_rx_buffer *rx_buf)
{ struct page *page = rx_buf->page;
if (page) { struct efx_rx_page_state *state = page_address(page);
/* efx_init_rx_buffers - create EFX_RX_BATCH page-based RX buffers * * @rx_queue: Efx RX queue * * This allocates a batch of pages, maps them for DMA, and populates * struct efx_rx_buffers for each one. Return a negative error code or * 0 on success. If a single page can be used for multiple buffers, * then the page will either be inserted fully, or not at all.
*/ staticint efx_init_rx_buffers(struct efx_rx_queue *rx_queue, bool atomic)
{ unsignedint page_offset, index, count; struct efx_nic *efx = rx_queue->efx; struct efx_rx_page_state *state; struct efx_rx_buffer *rx_buf;
dma_addr_t dma_addr; struct page *page;
/* efx_fast_push_rx_descriptors - push new RX descriptors quickly * @rx_queue: RX descriptor queue * * This will aim to fill the RX descriptor queue up to * @rx_queue->@max_fill. If there is insufficient atomic * memory to do so, a slow fill will be scheduled. * * The caller must provide serialisation (none is used here). In practise, * this means this function must run from the NAPI handler, or be called * when NAPI is disabled.
*/ void efx_fast_push_rx_descriptors(struct efx_rx_queue *rx_queue, bool atomic)
{ struct efx_nic *efx = rx_queue->efx; unsignedint fill_level, batch_size; int space, rc = 0;
if (!rx_queue->refill_enabled) return;
/* Calculate current fill level, and exit if we don't need to fill */
fill_level = (rx_queue->added_count - rx_queue->removed_count);
EFX_WARN_ON_ONCE_PARANOID(fill_level > rx_queue->efx->rxq_entries); if (fill_level >= rx_queue->fast_fill_trigger) goto out;
/* Record minimum fill level */ if (unlikely(fill_level < rx_queue->min_fill)) { if (fill_level)
rx_queue->min_fill = fill_level;
}
out: if (rx_queue->notified_count != rx_queue->added_count)
efx_nic_notify_rx_desc(rx_queue);
}
/* Pass a received packet up through GRO. GRO can handle pages * regardless of checksum state and skbs with a good checksum.
*/ void
efx_rx_packet_gro(struct efx_channel *channel, struct efx_rx_buffer *rx_buf, unsignedint n_frags, u8 *eh, __wsum csum)
{ struct napi_struct *napi = &channel->napi_str; struct efx_nic *efx = channel->efx; struct sk_buff *skb;
skb = napi_get_frags(napi); if (unlikely(!skb)) { struct efx_rx_queue *rx_queue;
for (i = 0; i < ARRAY_SIZE(efx->rss_context.rx_indir_table); i++)
indir[i] = ethtool_rxfh_indir_default(i, efx->rss_spread);
}
/** * efx_filter_is_mc_recipient - test whether spec is a multicast recipient * @spec: Specification to test * * Return: %true if the specification is a non-drop RX filter that * matches a local MAC address I/G bit value of 1 or matches a local * IPv4 or IPv6 address value in the respective multicast address * range. Otherwise %false.
*/ bool efx_filter_is_mc_recipient(conststruct efx_filter_spec *spec)
{ if (!(spec->flags & EFX_FILTER_FLAG_RX) ||
spec->dmaq_id == EFX_FILTER_RX_DMAQ_ID_DROP) returnfalse;
if (spec->match_flags &
(EFX_FILTER_MATCH_LOC_MAC | EFX_FILTER_MATCH_LOC_MAC_IG) &&
is_multicast_ether_addr(spec->loc_mac)) returntrue;
if ((spec->match_flags &
(EFX_FILTER_MATCH_ETHER_TYPE | EFX_FILTER_MATCH_LOC_HOST)) ==
(EFX_FILTER_MATCH_ETHER_TYPE | EFX_FILTER_MATCH_LOC_HOST)) { if (spec->ether_type == htons(ETH_P_IP) &&
ipv4_is_multicast(spec->loc_host[0])) returntrue; if (spec->ether_type == htons(ETH_P_IPV6) &&
((const u8 *)spec->loc_host)[0] == 0xff) returntrue;
}
#ifdef CONFIG_RFS_ACCEL bool efx_rps_check_rule(struct efx_arfs_rule *rule, unsignedint filter_idx, bool *force)
{ if (rule->filter_id == EFX_ARFS_FILTER_ID_PENDING) { /* ARFS is currently updating this entry, leave it */ returnfalse;
} if (rule->filter_id == EFX_ARFS_FILTER_ID_ERROR) { /* ARFS tried and failed to update this, so it's probably out * of date. Remove the filter and the ARFS rule entry.
*/
rule->filter_id = EFX_ARFS_FILTER_ID_REMOVING;
*force = true; returntrue;
} elseif (WARN_ON(rule->filter_id != filter_idx)) { /* can't happen */ /* ARFS has moved on, so old filter is not needed. Since we did * not mark the rule with EFX_ARFS_FILTER_ID_REMOVING, it will * not be removed by efx_rps_hash_del() subsequently.
*/
*force = true; returntrue;
} /* Remove it iff ARFS wants to. */ returntrue;
}
head = efx_rps_hash_bucket(efx, spec); if (WARN_ON(!head)) return;
hlist_for_each(node, head) {
rule = container_of(node, struct efx_arfs_rule, node); if (efx_filter_spec_equal(spec, &rule->spec)) { /* Someone already reused the entry. We know that if * this check doesn't fire (i.e. filter_id == REMOVING) * then the REMOVING mark was put there by our caller, * because caller is holding a lock on filter table and * only holders of that lock set REMOVING.
*/ if (rule->filter_id != EFX_ARFS_FILTER_ID_REMOVING) return;
hlist_del(node);
kfree(rule); return;
}
} /* We didn't find it. */
WARN_ON(1);
} #endif
int efx_probe_filters(struct efx_nic *efx)
{ int rc;
mutex_lock(&efx->mac_lock);
rc = efx->type->filter_table_probe(efx); if (rc) goto out_unlock;
#ifdef CONFIG_RFS_ACCEL if (efx->type->offload_features & NETIF_F_NTUPLE) { struct efx_channel *channel; int i, success = 1;
efx_for_each_channel(channel, efx) {
channel->rps_flow_id =
kcalloc(efx->type->max_rx_ip_filters, sizeof(*channel->rps_flow_id),
GFP_KERNEL); if (!channel->rps_flow_id)
success = 0; else for (i = 0;
i < efx->type->max_rx_ip_filters;
++i)
channel->rps_flow_id[i] =
RPS_FLOW_ID_INVALID;
channel->rfs_expire_index = 0;
channel->rfs_filter_count = 0;
}
rc = efx->type->filter_insert(efx, &req->spec, true); if (rc >= 0) /* Discard 'priority' part of EF10+ filter ID (mcdi_filters) */
rc %= efx->type->max_rx_ip_filters; if (efx->rps_hash_table) {
spin_lock_bh(&efx->rps_hash_lock);
rule = efx_rps_hash_find(efx, &req->spec); /* The rule might have already gone, if someone else's request * for the same spec was already worked and then expired before * we got around to our work. In that case we have nothing * tying us to an arfs_id, meaning that as soon as the filter * is considered for expiry it will be removed.
*/ if (rule) { if (rc < 0)
rule->filter_id = EFX_ARFS_FILTER_ID_ERROR; else
rule->filter_id = rc;
arfs_id = rule->arfs_id;
}
spin_unlock_bh(&efx->rps_hash_lock);
} if (rc >= 0) { /* Remember this so we can check whether to expire the filter * later.
*/
mutex_lock(&efx->rps_mutex); if (channel->rps_flow_id[rc] == RPS_FLOW_ID_INVALID)
channel->rfs_filter_count++;
channel->rps_flow_id[rc] = req->flow_id;
mutex_unlock(&efx->rps_mutex);
if (req->spec.ether_type == htons(ETH_P_IP))
netif_info(efx, rx_status, efx->net_dev, "steering %s %pI4:%u:%pI4:%u to queue %u [flow %u filter %d id %u]\n",
(req->spec.ip_proto == IPPROTO_TCP) ? "TCP" : "UDP",
req->spec.rem_host, ntohs(req->spec.rem_port),
req->spec.loc_host, ntohs(req->spec.loc_port),
req->rxq_index, req->flow_id, rc, arfs_id); else
netif_info(efx, rx_status, efx->net_dev, "steering %s [%pI6]:%u:[%pI6]:%u to queue %u [flow %u filter %d id %u]\n",
(req->spec.ip_proto == IPPROTO_TCP) ? "TCP" : "UDP",
req->spec.rem_host, ntohs(req->spec.rem_port),
req->spec.loc_host, ntohs(req->spec.loc_port),
req->rxq_index, req->flow_id, rc, arfs_id);
channel->n_rfs_succeeded++;
} else { if (req->spec.ether_type == htons(ETH_P_IP))
netif_dbg(efx, rx_status, efx->net_dev, "failed to steer %s %pI4:%u:%pI4:%u to queue %u [flow %u rc %d id %u]\n",
(req->spec.ip_proto == IPPROTO_TCP) ? "TCP" : "UDP",
req->spec.rem_host, ntohs(req->spec.rem_port),
req->spec.loc_host, ntohs(req->spec.loc_port),
req->rxq_index, req->flow_id, rc, arfs_id); else
netif_dbg(efx, rx_status, efx->net_dev, "failed to steer %s [%pI6]:%u:[%pI6]:%u to queue %u [flow %u rc %d id %u]\n",
(req->spec.ip_proto == IPPROTO_TCP) ? "TCP" : "UDP",
req->spec.rem_host, ntohs(req->spec.rem_port),
req->spec.loc_host, ntohs(req->spec.loc_port),
req->rxq_index, req->flow_id, rc, arfs_id);
channel->n_rfs_failed++; /* We're overloading the NIC's filter tables, so let's do a * chunk of extra expiry work.
*/
__efx_filter_rfs_expire(channel, min(channel->rfs_filter_count,
100u));
}
if (efx->rps_hash_table) { /* Add it to ARFS hash table */
spin_lock(&efx->rps_hash_lock);
rule = efx_rps_hash_add(efx, &req->spec, &new); if (!rule) {
rc = -ENOMEM; goto out_unlock;
} if (new)
rule->arfs_id = efx->rps_next_id++ % RPS_NO_FILTER;
rc = rule->arfs_id; /* Skip if existing or pending filter already does the right thing */ if (!new && rule->rxq_index == rxq_index &&
rule->filter_id >= EFX_ARFS_FILTER_ID_PENDING) goto out_unlock;
rule->rxq_index = rxq_index;
rule->filter_id = EFX_ARFS_FILTER_ID_PENDING;
spin_unlock(&efx->rps_hash_lock);
} else { /* Without an ARFS hash table, we just use arfs_id 0 for all * filters. This means if multiple flows hash to the same * flow_id, all but the most recently touched will be eligible * for expiry.
*/
rc = 0;
}
if (!mutex_trylock(&efx->rps_mutex)) returnfalse;
expire_one = efx->type->filter_rfs_expire_one;
index = channel->rfs_expire_index;
start = index;
size = efx->type->max_rx_ip_filters; while (quota) {
flow_id = channel->rps_flow_id[index];
if (flow_id != RPS_FLOW_ID_INVALID) {
quota--; if (expire_one(efx, flow_id, index)) {
netif_info(efx, rx_status, efx->net_dev, "expired filter %d [channel %u flow %u]\n",
index, channel->channel, flow_id);
channel->rps_flow_id[index] = RPS_FLOW_ID_INVALID;
channel->rfs_filter_count--;
}
} if (++index == size)
index = 0; /* If we were called with a quota that exceeds the total number * of filters in the table (which shouldn't happen, but could * if two callers race), ensure that we don't loop forever - * stop when we've examined every row of the table.
*/ if (index == start) break;
}
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung und die Messung sind noch experimentell.