/* VSI and Tx ring */ if (!vsi) return -ENOENT;
tx_ring = vsi->tx_rings[0]; if (!tx_ring || !tx_ring->desc) return -ENOENT;
dev = tx_ring->dev;
/* we are using two descriptors to add/del a filter and we can wait */ for (i = ICE_FDIR_CLEAN_DELAY; ICE_DESC_UNUSED(tx_ring) < 2; i--) { if (!i) return -EAGAIN;
msleep_interruptible(1);
}
/* grab the next descriptor */
i = tx_ring->next_to_use;
first = &tx_ring->tx_buf[i];
f_desc = ICE_TX_FDIRDESC(tx_ring, i);
memcpy(f_desc, fdir_desc, sizeof(*f_desc));
i++;
i = (i < tx_ring->count) ? i : 0;
tx_desc = ICE_TX_DESC(tx_ring, i);
tx_buf = &tx_ring->tx_buf[i];
i++;
tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
switch (tx_buf->type) { case ICE_TX_BUF_DUMMY:
devm_kfree(ring->dev, tx_buf->raw_buf); break; case ICE_TX_BUF_SKB:
dev_kfree_skb_any(tx_buf->skb); break; case ICE_TX_BUF_XDP_TX:
page_frag_free(tx_buf->raw_buf); break; case ICE_TX_BUF_XDP_XMIT:
xdp_return_frame(tx_buf->xdpf); break;
}
tx_buf->next_to_watch = NULL;
tx_buf->type = ICE_TX_BUF_EMPTY;
dma_unmap_len_set(tx_buf, len, 0); /* tx_buf must be completely set up in the transmit path */
}
/* get the bql data ready */
netdev_txq_bql_complete_prefetchw(txring_txq(tx_ring));
tx_buf = &tx_ring->tx_buf[i];
tx_desc = ICE_TX_DESC(tx_ring, i);
i -= tx_ring->count;
prefetch(&vsi->state);
do { struct ice_tx_desc *eop_desc = tx_buf->next_to_watch;
/* if next_to_watch is not set then there is no work pending */ if (!eop_desc) break;
/* follow the guidelines of other drivers */
prefetchw(&tx_buf->skb->users);
smp_rmb(); /* prevent any other reads prior to eop_desc */
ice_trace(clean_tx_irq, tx_ring, tx_desc, tx_buf); /* if the descriptor isn't done, no work yet to do */ if (!(eop_desc->cmd_type_offset_bsz &
cpu_to_le64(ICE_TX_DESC_DTYPE_DESC_DONE))) break;
/* unmap any remaining paged data */ if (dma_unmap_len(tx_buf, len)) {
dma_unmap_page(tx_ring->dev,
dma_unmap_addr(tx_buf, dma),
dma_unmap_len(tx_buf, len),
DMA_TO_DEVICE);
dma_unmap_len_set(tx_buf, len, 0);
}
}
ice_trace(clean_tx_irq_unmap_eop, tx_ring, tx_desc, tx_buf);
/* move us one more past the eop_desc for start of next pkt */
tx_buf++;
tx_desc++;
i++; if (unlikely(!i)) {
i -= tx_ring->count;
tx_buf = tx_ring->tx_buf;
tx_desc = ICE_TX_DESC(tx_ring, 0);
}
prefetch(tx_desc);
/* update budget accounting */
budget--;
} while (likely(budget));
/* warn if we are about to overwrite the pointer */
WARN_ON(tx_ring->tx_buf);
tx_ring->tx_buf =
devm_kcalloc(dev, sizeof(*tx_ring->tx_buf), tx_ring->count,
GFP_KERNEL); if (!tx_ring->tx_buf) return -ENOMEM;
/* round up to nearest page */
size = ALIGN(tx_ring->count * sizeof(struct ice_tx_desc),
PAGE_SIZE);
tx_ring->desc = dmam_alloc_coherent(dev, size, &tx_ring->dma,
GFP_KERNEL); if (!tx_ring->desc) {
dev_err(dev, "Unable to allocate memory for the Tx descriptor ring, size=%d\n",
size); goto err;
}
/* ring already cleared, nothing to do */ if (!rx_ring->rx_buf) return;
if (rx_ring->xsk_pool) {
ice_xsk_clean_rx_ring(rx_ring); goto rx_skip_free;
}
if (xdp->data) {
xdp_return_buff(xdp);
xdp->data = NULL;
}
/* Free all the Rx ring sk_buffs */ for (i = 0; i < rx_ring->count; i++) { struct ice_rx_buf *rx_buf = &rx_ring->rx_buf[i];
if (!rx_buf->page) continue;
/* Invalidate cache lines that may have been written to by *devicesothatweavoidcorruptingmemory.
*/
dma_sync_single_range_for_cpu(dev, rx_buf->dma,
rx_buf->page_offset,
rx_ring->rx_buf_len,
DMA_FROM_DEVICE);
/* warn if we are about to overwrite the pointer */
WARN_ON(rx_ring->rx_buf);
rx_ring->rx_buf =
kcalloc(rx_ring->count, sizeof(*rx_ring->rx_buf), GFP_KERNEL); if (!rx_ring->rx_buf) return -ENOMEM;
/* round up to nearest page */
size = ALIGN(rx_ring->count * sizeof(union ice_32byte_rx_desc),
PAGE_SIZE);
rx_ring->desc = dmam_alloc_coherent(dev, size, &rx_ring->dma,
GFP_KERNEL); if (!rx_ring->desc) {
dev_err(dev, "Unable to allocate memory for the Rx descriptor ring, size=%d\n",
size); goto err;
}
/* since we are recycling buffers we should seldom need to alloc */ if (likely(page)) returntrue;
/* alloc new page for storage */
page = dev_alloc_pages(ice_rx_pg_order(rx_ring)); if (unlikely(!page)) {
rx_ring->ring_stats->rx_stats.alloc_page_failed++; returnfalse;
}
/* map page for use */
dma = dma_map_page_attrs(rx_ring->dev, page, 0, ice_rx_pg_size(rx_ring),
DMA_FROM_DEVICE, ICE_RX_DMA_ATTR);
/* if mapping failed free memory back to system since *thereisn'tmuchpointinholdingmemorywecan'tuse
*/ if (dma_mapping_error(rx_ring->dev, dma)) {
__free_pages(page, ice_rx_pg_order(rx_ring));
rx_ring->ring_stats->rx_stats.alloc_page_failed++; returnfalse;
}
/* do nothing if no valid netdev defined */ if (!rx_ring->netdev || !cleaned_count) returnfalse;
/* get the Rx descriptor and buffer based on next_to_use */
rx_desc = ICE_RX_DESC(rx_ring, ntu);
bi = &rx_ring->rx_buf[ntu];
do { /* if we fail here, we have work remaining */ if (!ice_alloc_mapped_page(rx_ring, bi)) break;
/* sync the buffer for use by the device */
dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
bi->page_offset,
rx_ring->rx_buf_len,
DMA_FROM_DEVICE);
/* Refresh the desc even if buffer_addrs didn't change *becauseeachwrite-backerasesthisinfo.
*/
rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
rx_desc++;
bi++;
ntu++; if (unlikely(ntu == rx_ring->count)) {
rx_desc = ICE_RX_DESC(rx_ring, 0);
bi = rx_ring->rx_buf;
ntu = 0;
}
/* clear the status bits for the next_to_use descriptor */
rx_desc->wb.status_error0 = 0;
cleaned_count--;
} while (cleaned_count);
if (rx_ring->next_to_use != ntu)
ice_release_rx_desc(rx_ring, ntu);
return !!cleaned_count;
}
/** *ice_rx_buf_adjust_pg_offset-PrepareRxbufferforreuse *@rx_buf:Rxbuffertoadjust *@size:Sizeofadjustment * *UpdatetheoffsetwithinpagesothatRxbufwillbereadytobereused. *ForsystemswithPAGE_SIZE<8192thisfunctionwillflipthepageoffset *sothesecondhalfofpageassignedtoRxbufferwillbeused,otherwise *theoffsetismovedby"size"bytes
*/ staticvoid
ice_rx_buf_adjust_pg_offset(struct ice_rx_buf *rx_buf, unsignedint size)
{ #if (PAGE_SIZE < 8192) /* flip page offset to other buffer */
rx_buf->page_offset ^= size; #else /* move offset up to the next cache line */
rx_buf->page_offset += size; #endif
}
/* avoid re-using remote and pfmemalloc pages */ if (!dev_page_is_reusable(page)) returnfalse;
/* if we are only owner of page we can reuse it */ if (unlikely(rx_buf->pgcnt - pagecnt_bias > 1)) returnfalse; #if (PAGE_SIZE >= 8192) #define ICE_LAST_OFFSET \
(SKB_WITH_OVERHEAD(PAGE_SIZE) - ICE_RXBUF_3072) if (rx_buf->page_offset > ICE_LAST_OFFSET) returnfalse; #endif/* PAGE_SIZE >= 8192) */
/* If we have drained the page fragment pool we need to update *thepagecnt_biasandpagecountsothatwefullyrestockthe *numberofreferencesthedriverholds.
*/ if (unlikely(pagecnt_bias == 1)) {
page_ref_add(page, USHRT_MAX - 1);
rx_buf->pagecnt_bias = USHRT_MAX;
}
/* update, and store next to alloc */
nta++;
rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
/* Transfer page from old buffer to new buffer. *Moveeachmemberindividuallytoavoidpossiblestore *forwardingstallsandunnecessarycopyofskb.
*/
new_buf->dma = old_buf->dma;
new_buf->page = old_buf->page;
new_buf->page_offset = old_buf->page_offset;
new_buf->pagecnt_bias = old_buf->pagecnt_bias;
}
if (!size) return rx_buf; /* we are reusing so sync this buffer for CPU use */
dma_sync_single_range_for_cpu(rx_ring->dev, rx_buf->dma,
rx_buf->page_offset, size,
DMA_FROM_DEVICE);
/* We have pulled a buffer for use, so decrement pagecnt_bias */
rx_buf->pagecnt_bias--;
if (unlikely(xdp_buff_has_frags(xdp))) {
sinfo = xdp_get_shared_info_from_buff(xdp);
nr_frags = sinfo->nr_frags;
}
/* Prefetch first cache line of first page. If xdp->data_meta *isunused,thispointsexactlyasxdp->data,otherwisewe *likelyhaveaconsumeraccessingfirstfewbytesofmeta *data,andthenactualdata.
*/
net_prefetch(xdp->data_meta); /* build an skb around the page buffer */
skb = napi_build_skb(xdp->data_hard_start, xdp->frame_sz); if (unlikely(!skb)) return NULL;
/* must to record Rx queue, otherwise OS features such as *symmetricqueuewon'twork
*/
skb_record_rx_queue(skb, rx_ring->q_index);
/* update pointers within the skb to store the data */
skb_reserve(skb, xdp->data - xdp->data_hard_start);
__skb_put(skb, xdp->data_end - xdp->data); if (metasize)
skb_metadata_set(skb, metasize);
if (unlikely(xdp_buff_has_frags(xdp)))
xdp_update_skb_shared_info(skb, nr_frags,
sinfo->xdp_frags_size,
nr_frags * xdp->frame_sz,
xdp_buff_is_frag_pfmemalloc(xdp));
/* prefetch first cache line of first page */
net_prefetch(xdp->data);
if (unlikely(xdp_buff_has_frags(xdp))) {
sinfo = xdp_get_shared_info_from_buff(xdp);
nr_frags = sinfo->nr_frags;
}
/* allocate a skb to store the frags */
skb = napi_alloc_skb(&rx_ring->q_vector->napi, ICE_RX_HDR_SIZE); if (unlikely(!skb)) return NULL;
rx_buf = &rx_ring->rx_buf[rx_ring->first_desc];
skb_record_rx_queue(skb, rx_ring->q_index); /* Determine available headroom for copy */
headlen = size; if (headlen > ICE_RX_HDR_SIZE)
headlen = eth_get_headlen(skb->dev, xdp->data, ICE_RX_HDR_SIZE);
/* align pull length to size of long to optimize memcpy performance */
memcpy(__skb_put(skb, headlen), xdp->data, ALIGN(headlen, sizeof(long)));
/* if we exhaust the linear part then add what is left as a frag */
size -= headlen; if (size) { /* besides adding here a partial frag, we are going to add *fragsfromxdp_buff,makesurethereisenoughspacefor *them
*/ if (unlikely(nr_frags >= MAX_SKB_FRAGS - 1)) {
dev_kfree_skb(skb); return NULL;
}
skb_add_rx_frag(skb, 0, rx_buf->page,
rx_buf->page_offset + headlen, size,
xdp->frame_sz);
} else { /* buffer is unused, restore biased page count in Rx buffer; *datawascopiedontoskb'slinearpartsothere'sno *needforadjustingpageoffsetandwecanreusethisbuffer *as-is
*/
rx_buf->pagecnt_bias++;
}
if (unlikely(xdp_buff_has_frags(xdp))) { struct skb_shared_info *skinfo = skb_shinfo(skb);
if (ice_can_reuse_rx_page(rx_buf)) { /* hand second half of page back to the ring */
ice_reuse_rx_page(rx_ring, rx_buf);
} else { /* we are not reusing the buffer so unmap it */
dma_unmap_page_attrs(rx_ring->dev, rx_buf->dma,
ice_rx_pg_size(rx_ring), DMA_FROM_DEVICE,
ICE_RX_DMA_ATTR);
__page_frag_cache_drain(rx_buf->page, rx_buf->pagecnt_bias);
}
/* clear contents of buffer_info */
rx_buf->page = NULL;
}
if (unlikely(xdp_buff_has_frags(xdp)))
xdp_frags = xdp_get_shared_info_from_buff(xdp)->nr_frags;
while (idx != ntc) {
buf = &rx_ring->rx_buf[idx]; if (++idx == cnt)
idx = 0;
/* An XDP program could release fragments from the end of the *buffer.Forthese,weneedtokeepthepagecnt_biasas-is. *Todothis,onlyadjustpagecnt_biasforfragmentsupto *thetotalremainingaftertheXDPprogramhasrun.
*/ if (verdict != ICE_XDP_CONSUMED)
ice_rx_buf_adjust_pg_offset(buf, xdp->frame_sz); elseif (i++ <= xdp_frags)
buf->pagecnt_bias++;
/* start the loop to process Rx packets bounded by 'budget' */ while (likely(total_rx_pkts < (unsignedint)budget)) { union ice_32b_rx_flex_desc *rx_desc; struct ice_rx_buf *rx_buf; struct sk_buff *skb; unsignedint size;
u16 stat_err_bits;
u16 vlan_tci;
/* get the Rx desc from Rx ring based on 'next_to_clean' */
rx_desc = ICE_RX_DESC(rx_ring, ntc);
/* status_error_len will always be zero for unused descriptors *becauseit'sclearedincleanup,andoverlapswithhdr_addr *whichisalwayszerobecausepacketsplitisn'tused,ifthe *hardwarewroteDDthenitwillbenon-zero
*/
stat_err_bits = BIT(ICE_RX_FLEX_DESC_STATUS0_DD_S); if (!ice_test_staterr(rx_desc->wb.status_error0, stat_err_bits)) break;
/* This memory barrier is needed to keep us from reading *anyotherfieldsoutoftherx_descuntilweknowthe *DDbitisset.
*/
dma_rmb();
/* if dim settings get stale, like when not updated for 1 *secondorlonger,forceittostartagain.Thisaddressesthe *frequentcaseofanidlequeuebeingswitchedtobythe *scheduler.The1,000heremeans1,000milliseconds.
*/ if (ktime_ms_delta(sample->time, rc->dim.start_sample.time) >= 1000)
rc->dim.state = DIM_START_MEASURE;
}
/** *ice_buildreg_itr-buildvalueforwritingtotheGLINT_DYN_CTLregister *@itr_idx:interruptthrottlingindex *@itr:interruptthrottlingvalueinusecs
*/ static u32 ice_buildreg_itr(u16 itr_idx, u16 itr)
{ /* The ITR value is reported in microseconds, and the register value is *recordedin2microsecondunits.Forthisreasonweonlyneedto *shiftbytheGLINT_DYN_CTL_INTERVAL_S-ICE_ITR_GRAN_Stoapplythis *granularityasashiftinsteadofdivision.Themaskmakessurethe *ITRvalueisneveroddsowedon'taccidentallywriteintothefield *priortotheITRfield.
*/
itr &= ICE_ITR_MASK;
/* trigger an ITR delayed software interrupt when exiting busy poll, to *makesuretocatchanypendingcleanupsthatmighthavebeenmissed *duetointerruptstatetransition.Ifbusypollorpollisn't *enabled,thendon'tupdateITR,andjustenabletheinterrupt.
*/ if (!wb_en) {
itr_val = ice_buildreg_itr(ICE_ITR_NONE, 0);
} else {
q_vector->wb_on_itr = false;
/* do two things here with a single write. Set up the third ITR *indextobeusedforsoftwareinterruptmoderation,andthen *triggerasoftwareinterruptwitharatelimitof20Kon *softwareinterrupts,thiswillhelpavoidhighinterrupt *loadsduetofrequentlypollingandexitingpolling.
*/
itr_val = ice_buildreg_itr(ICE_IDX_ITR2, ICE_ITR_20K);
itr_val |= GLINT_DYN_CTL_SWINT_TRIG_M |
ICE_IDX_ITR2 << GLINT_DYN_CTL_SW_ITR_INDX_S |
GLINT_DYN_CTL_SW_ITR_INDX_ENA_M;
}
wr32(&vsi->back->hw, GLINT_DYN_CTL(q_vector->reg_idx), itr_val);
}
/* already in wb_on_itr mode no need to change it */ if (q_vector->wb_on_itr) return;
/* use previously set ITR values for all of the ITR indices by *specifyingICE_ITR_NONE,whichwillvaryinadaptive(AIM)modeand *bestaticinnon-adaptivemode(userconfigured)
*/
wr32(&vsi->back->hw, GLINT_DYN_CTL(q_vector->reg_idx),
FIELD_PREP(GLINT_DYN_CTL_ITR_INDX_M, ICE_ITR_NONE) |
FIELD_PREP(GLINT_DYN_CTL_INTENA_MSK_M, 1) |
FIELD_PREP(GLINT_DYN_CTL_WB_ON_ITR_M, 1));
q_vector->wb_on_itr = true;
}
/** *ice_napi_poll-NAPIpollingRx/Txcleanuproutine *@napi:napistructwithourdevicesinfoinit *@budget:amountofworkdriverisallowedtodothispass,inpackets * *Thisfunctionwillcleanallqueuesassociatedwithaq_vector. * *Returnstheamountofworkdone
*/ int ice_napi_poll(struct napi_struct *napi, int budget)
{ struct ice_q_vector *q_vector =
container_of(napi, struct ice_q_vector, napi); struct ice_tx_ring *tx_ring; struct ice_rx_ring *rx_ring; bool clean_complete = true; int budget_per_ring; int work_done = 0;
/* Since the actual Tx work is minimal, we can give the Tx a larger *budgetandbemoreaggressiveaboutcleaninguptheTxdescriptors.
*/
ice_for_each_tx_ring(tx_ring, q_vector->tx) { struct xsk_buff_pool *xsk_pool = READ_ONCE(tx_ring->xsk_pool); bool wd;
/* Handle case where we are called by netpoll with a budget of 0 */ if (unlikely(budget <= 0)) return budget;
/* normally we have 1 Rx ring per q_vector */ if (unlikely(q_vector->num_ring_rx > 1)) /* We attempt to distribute budget to each Rx queue fairly, but *don'tallowthebudgettogobelow1becausethatwouldexit *pollingearly.
*/
budget_per_ring = max_t(int, budget / q_vector->num_ring_rx, 1); else /* Max of 1 Rx ring in this q_vector so give it the budget */
budget_per_ring = budget;
ice_for_each_rx_ring(rx_ring, q_vector->rx) { struct xsk_buff_pool *xsk_pool = READ_ONCE(rx_ring->xsk_pool); int cleaned;
/* A dedicated path for zero-copy allows making a single *comparisonintheirqcontextinsteadofmanyinsidethe *ice_clean_rx_irqfunctionandmakesthecodebasecleaner.
*/
cleaned = rx_ring->xsk_pool ?
ice_clean_rx_irq_zc(rx_ring, xsk_pool, budget_per_ring) :
ice_clean_rx_irq(rx_ring, budget_per_ring);
work_done += cleaned; /* if we clean as many as budgeted, we must not be done */ if (cleaned >= budget_per_ring)
clean_complete = false;
}
/* If work not completed, return budget and polling will return */ if (!clean_complete) { /* Set the writeback on ITR so partial completions of *cache-lineswillstillcontinueevenifwe'repolling.
*/
ice_set_wb_on_itr(q_vector); return budget;
}
/* Exit the polling mode, but don't re-enable interrupts if stack might *pollusduetobusy-polling
*/ if (napi_complete_done(napi, work_done)) {
ice_net_dim(q_vector);
ice_enable_interrupt(q_vector);
} else {
ice_set_wb_on_itr(q_vector);
}
return min_t(int, work_done, budget - 1);
}
/** *__ice_maybe_stop_tx-2ndlevelcheckforTxstopconditions *@tx_ring:theringtobechecked *@size:thesizebufferwewanttoassureisavailable * *Returns-EBUSYifastopisneeded,else0
*/ staticint __ice_maybe_stop_tx(struct ice_tx_ring *tx_ring, unsignedint size)
{
netif_tx_stop_queue(txring_txq(tx_ring)); /* Memory barrier before checking head and tail */
smp_mb();
/* Check again in a case another CPU has just made room available. */ if (likely(ICE_DESC_UNUSED(tx_ring) < size)) return -EBUSY;
/* A reprieve! - use start_queue because it doesn't call schedule */
netif_tx_start_queue(txring_txq(tx_ring));
++tx_ring->ring_stats->tx_stats.restart_q; return0;
}
for (frag = &skb_shinfo(skb)->frags[0];; frag++) { unsignedint max_data = ICE_MAX_DATA_PER_TXD_ALIGNED;
if (dma_mapping_error(tx_ring->dev, dma)) goto dma_error;
/* record length, and DMA address */
dma_unmap_len_set(tx_buf, len, size);
dma_unmap_addr_set(tx_buf, dma, dma);
/* align size to end of page */
max_data += -dma & (ICE_MAX_READ_REQ_SIZE - 1);
tx_desc->buf_addr = cpu_to_le64(dma);
/* account for data chunks larger than the hardware *canhandle
*/ while (unlikely(size > ICE_MAX_DATA_PER_TXD)) {
tx_desc->cmd_type_offset_bsz =
ice_build_ctob(td_cmd, td_offset, max_data,
td_tag);
tx_desc++;
i++;
if (i == tx_ring->count) {
tx_desc = ICE_TX_DESC(tx_ring, 0);
i = 0;
}
/* record SW timestamp if HW timestamp is not available */
skb_tx_timestamp(first->skb);
i++; if (i == tx_ring->count)
i = 0;
/* write last descriptor with RS and EOP bits */
td_cmd |= (u64)ICE_TXD_LAST_DESC_CMD;
tx_desc->cmd_type_offset_bsz =
ice_build_ctob(td_cmd, td_offset, size, td_tag);
/* Force memory writes to complete before letting h/w know there *arenewdescriptorstofetch. * *Wealsousethismemorybarriertomakecertainallofthe *statusbitshavebeenupdatedbeforenext_to_watchiswritten.
*/
wmb();
/* set next_to_watch value indicating a packet is present */
first->next_to_watch = tx_desc;
tx_ring->next_to_use = i;
ice_maybe_stop_tx(tx_ring, DESC_NEEDED);
/* notify HW of packet */
kick = __netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount,
netdev_xmit_more()); if (kick) /* notify HW of packet */
writel(i, tx_ring->tail);
return;
dma_error: /* clear DMA mappings for failed tx_buf map */ for (;;) {
tx_buf = &tx_ring->tx_buf[i];
ice_unmap_and_free_tx_buf(tx_ring, tx_buf); if (tx_buf == first) break; if (i == 0)
i = tx_ring->count;
i--;
}
/* set the tx_flags to indicate the IP protocol type. this is *requiredsothatchecksumheadercomputationbelowisaccurate.
*/ if (ip.v4->version == 4)
first->tx_flags |= ICE_TX_FLAGS_IPV4; elseif (ip.v6->version == 6)
first->tx_flags |= ICE_TX_FLAGS_IPV6;
gso_ena = skb_shinfo(skb)->gso_type & SKB_GSO_PARTIAL; /* indicate if we need to offload outer UDP header */ if ((first->tx_flags & ICE_TX_FLAGS_TSO) && !gso_ena &&
(skb_shinfo(skb)->gso_type & SKB_GSO_UDP_TUNNEL_CSUM))
tunnel |= ICE_TXD_CTX_QW0_L4T_CS_M;
/* record tunnel offload values */
off->cd_tunnel_params |= tunnel;
/* set DTYP=1 to indicate that it's an Tx context descriptor *inIPsectunnelmodewithTxoffloadsinQuadword1
*/
off->cd_qw1 |= (u64)ICE_TX_DESC_DTYPE_CTX;
/* switch L4 header pointer from outer to inner */
l4.hdr = skb_inner_transport_header(skb);
l4_proto = 0;
/* reset type as we transition from outer to inner headers */
first->tx_flags &= ~(ICE_TX_FLAGS_IPV4 | ICE_TX_FLAGS_IPV6); if (ip.v4->version == 4)
first->tx_flags |= ICE_TX_FLAGS_IPV4; if (ip.v6->version == 6)
first->tx_flags |= ICE_TX_FLAGS_IPV6;
}
/* Enable IP checksum offloads */ if (first->tx_flags & ICE_TX_FLAGS_IPV4) {
l4_proto = ip.v4->protocol; /* the stack computes the IP header already, the only time we *needthehardwaretorecomputeitisinthecaseofTSO.
*/ if (first->tx_flags & ICE_TX_FLAGS_TSO)
cmd |= ICE_TX_DESC_CMD_IIPT_IPV4_CSUM; else
cmd |= ICE_TX_DESC_CMD_IIPT_IPV4;
/* nothing left to do, software offloaded VLAN */ if (!skb_vlan_tag_present(skb) && eth_type_vlan(skb->protocol)) return;
/* the VLAN ethertype/tpid is determined by VSI configuration and netdev *featureflags,whichthedriveronlyallowseither802.1Qor802.1ad *VLANoffloadsexclusivelysoweonlycareabouttheVLANIDhere
*/ if (skb_vlan_tag_present(skb)) {
first->vid = skb_vlan_tag_get(skb); if (tx_ring->flags & ICE_TX_FLAGS_RING_VLAN_L2TAG2)
first->tx_flags |= ICE_TX_FLAGS_HW_OUTER_SINGLE_VLAN; else
first->tx_flags |= ICE_TX_FLAGS_HW_VLAN;
}
/* no need to check if number of frags is less than 7 */
nr_frags = skb_shinfo(skb)->nr_frags; if (nr_frags < (ICE_MAX_BUF_TXD - 1)) returnfalse;
/* We need to walk through the list and validate that each group *of6fragmentstotalsatleastgso_size.
*/
nr_frags -= ICE_MAX_BUF_TXD - 2;
frag = &skb_shinfo(skb)->frags[0];
/* Initialize size to the negative value of gso_size minus 1. We *usethisastheworstcasescenarioinwhichthefragahead *ofusonlyprovidesonebytewhichiswhywearelimitedto6 *descriptorsforasingletransmitastheheaderandprevious *fragmentarealreadyconsuming2descriptors.
*/
sum = 1 - skb_shinfo(skb)->gso_size;
/* Add size of frags 0 through 4 to create our initial sum */
sum += skb_frag_size(frag++);
sum += skb_frag_size(frag++);
sum += skb_frag_size(frag++);
sum += skb_frag_size(frag++);
sum += skb_frag_size(frag++);
/* Walk through fragments adding latest fragment, testing it, and *thenremovingstalefragmentsfromthesum.
*/ for (stale = &skb_shinfo(skb)->frags[0];; stale++) { int stale_size = skb_frag_size(stale);
sum += skb_frag_size(frag++);
/* The stale fragment may present us with a smaller *descriptorthantheactualfragmentsize.Toaccount *forthatweneedtoremoveallthedataonthefrontand *figureoutwhattheremainderwouldbeinthelast *descriptorassociatedwiththefragment.
*/ if (stale_size > ICE_MAX_DATA_PER_TXD) { int align_pad = -(skb_frag_off(stale)) &
(ICE_MAX_READ_REQ_SIZE - 1);
sum -= align_pad;
stale_size -= align_pad;
do {
sum -= ICE_MAX_DATA_PER_TXD_ALIGNED;
stale_size -= ICE_MAX_DATA_PER_TXD_ALIGNED;
} while (stale_size > ICE_MAX_DATA_PER_TXD);
}
/* if sum is negative we failed to make sufficient progress */ if (sum < 0) returntrue;
if (!nr_frags--) break;
sum -= stale_size;
}
returnfalse;
}
/** *ice_chk_linearize-Checkiftherearemorethan8fragmentsperpacket *@skb:sendbuffer *@count:numberofbuffersused * *Note:OurHWcan'tscatter-gathermorethan8fragmentstobuild *apacketonthewireandsoweneedtofigureoutthecaseswherewe *needtolinearizetheskb.
*/ staticbool ice_chk_linearize(struct sk_buff *skb, unsignedint count)
{ /* Both TSO and single send will work if count is less than 8 */ if (likely(count < ICE_MAX_BUF_TXD)) returnfalse;
if (skb_is_gso(skb)) return __ice_chk_linearize(skb);
/* we can support up to 8 data buffers for a single send */ return count != ICE_MAX_BUF_TXD;
}
/* prefetch for bql data which is infrequently used */
netdev_txq_bql_enqueue_prefetchw(txring_txq(tx_ring));
offload.tx_ring = tx_ring;
/* record the location of the first descriptor for this packet */
first = &tx_ring->tx_buf[tx_ring->next_to_use];
first->skb = skb;
first->type = ICE_TX_BUF_SKB;
first->bytecount = max_t(unsignedint, skb->len, ETH_ZLEN);
first->gso_segs = 1;
first->tx_flags = 0;
/* prepare the VLAN tagging flags for Tx */
ice_tx_prepare_vlan_flags(tx_ring, first); if (first->tx_flags & ICE_TX_FLAGS_HW_OUTER_SINGLE_VLAN) {
offload.cd_qw1 |= (u64)(ICE_TX_DESC_DTYPE_CTX |
(ICE_TX_CTX_DESC_IL2TAG2 <<
ICE_TXD_CTX_QW1_CMD_S));
offload.cd_l2tag2 = first->vid;
}
/* set up TSO offload */
tso = ice_tso(first, &offload); if (tso < 0) goto out_drop;
/* always set up Tx checksum offload */
csum = ice_tx_csum(first, &offload); if (csum < 0) goto out_drop;
/* allow CONTROL frames egress from main VSI if FW LLDP disabled */
eth = (struct ethhdr *)skb_mac_header(skb);
/* move past eop_desc for start of next FD desc */
tx_buf++;
tx_desc++;
i++; if (unlikely(!i)) {
i -= tx_ring->count;
tx_buf = tx_ring->tx_buf;
tx_desc = ICE_TX_DESC(tx_ring, 0);
}
budget--;
} while (likely(budget));
i += tx_ring->count;
tx_ring->next_to_clean = i;
/* re-enable interrupt if needed */
ice_irq_dynamic_ena(&vsi->back->hw, vsi, vsi->q_vectors[0]);
}
Messung V0.5 in Prozent
¤ Dauer der Verarbeitung: 0.69 Sekunden
(vorverarbeitet am 2026-10-03)
¤
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.