| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20200722155349.747-15-alobakin@marvell.com>
Date: Wed, 22 Jul 2020 18:53:48 +0300
From: Alexander Lobakin <alobakin@...vell.com>
To: "David S. Miller" <davem@...emloft.net>,
Jakub Kicinski <kuba@...nel.org>
CC: Alexander Lobakin <alobakin@...vell.com>,
Igor Russkikh <irusskikh@...vell.com>,
Michal Kalderon <michal.kalderon@...vell.com>,
"Ariel Elior" <aelior@...vell.com>,
Denis Bolotin <denis.bolotin@...vell.com>,
"Doug Ledford" <dledford@...hat.com>,
Jason Gunthorpe <jgg@...pe.ca>,
"Alexei Starovoitov" <ast@...nel.org>,
Daniel Borkmann <daniel@...earbox.net>,
"Jesper Dangaard Brouer" <hawk@...nel.org>,
John Fastabend <john.fastabend@...il.com>,
Martin KaFai Lau <kafai@...com>,
Song Liu <songliubraving@...com>, "Yonghong Song" <yhs@...com>,
Andrii Nakryiko <andriin@...com>,
KP Singh <kpsingh@...omium.org>,
<GR-everest-linux-l2@...vell.com>, <netdev@...r.kernel.org>,
<bpf@...r.kernel.org>, <linux-rdma@...r.kernel.org>,
<linux-kernel@...r.kernel.org>
Subject: [PATCH net-next 14/15] qede: refactor XDP Tx processing
Current XDP Tx logic is suboptimal and can't be reused for XDP_REDIRECT
path.
Make qede_xdp_{tx_int,xmit}() more universal and effective in general to
allow future expanding.
Misc: use unlikely() hints where appropriate and replace "fallthrough"
comments with pseudo-keywords.
Signed-off-by: Alexander Lobakin <alobakin@...vell.com>
Signed-off-by: Igor Russkikh <irusskikh@...vell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@...vell.com>
---
drivers/net/ethernet/qlogic/qede/qede.h | 1 +
drivers/net/ethernet/qlogic/qede/qede_fp.c | 89 +++++++++++-----------
2 files changed, 45 insertions(+), 45 deletions(-)
diff --git a/drivers/net/ethernet/qlogic/qede/qede.h b/drivers/net/ethernet/qlogic/qede/qede.h
index e8ed0bb94ee0..308c66a5f98f 100644
--- a/drivers/net/ethernet/qlogic/qede/qede.h
+++ b/drivers/net/ethernet/qlogic/qede/qede.h
@@ -455,6 +455,7 @@ struct qede_fastpath {
u8 id;
u8 xdp_xmit;
+#define QEDE_XDP_TX BIT(0)
struct napi_struct napi;
struct qed_sb_info *sb_info;
diff --git a/drivers/net/ethernet/qlogic/qede/qede_fp.c b/drivers/net/ethernet/qlogic/qede/qede_fp.c
index 1c4ece0713f8..c80bf6d37b89 100644
--- a/drivers/net/ethernet/qlogic/qede/qede_fp.c
+++ b/drivers/net/ethernet/qlogic/qede/qede_fp.c
@@ -302,48 +302,37 @@ static inline void qede_update_tx_producer(struct qede_tx_queue *txq)
wmb();
}
-static int qede_xdp_xmit(struct qede_dev *edev, struct qede_fastpath *fp,
- struct sw_rx_data *metadata, u16 padding, u16 length)
+static int qede_xdp_xmit(struct qede_tx_queue *txq, dma_addr_t dma, u16 pad,
+ u16 len, struct page *page)
{
- struct qede_tx_queue *txq = fp->xdp_tx;
- struct eth_tx_1st_bd *first_bd;
- u16 idx = txq->sw_tx_prod;
+ struct eth_tx_1st_bd *bd;
+ struct sw_tx_xdp *xdp;
u16 val;
- if (!qed_chain_get_elem_left(&txq->tx_pbl)) {
+ if (unlikely(qed_chain_get_elem_used(&txq->tx_pbl) >=
+ txq->num_tx_buffers)) {
txq->stopped_cnt++;
return -ENOMEM;
}
- first_bd = (struct eth_tx_1st_bd *)qed_chain_produce(&txq->tx_pbl);
+ bd = qed_chain_produce(&txq->tx_pbl);
+ bd->data.nbds = 1;
+ bd->data.bd_flags.bitfields = BIT(ETH_TX_1ST_BD_FLAGS_START_BD_SHIFT);
- memset(first_bd, 0, sizeof(*first_bd));
- first_bd->data.bd_flags.bitfields =
- BIT(ETH_TX_1ST_BD_FLAGS_START_BD_SHIFT);
-
- val = (length & ETH_TX_DATA_1ST_BD_PKT_LEN_MASK) <<
+ val = (len & ETH_TX_DATA_1ST_BD_PKT_LEN_MASK) <<
ETH_TX_DATA_1ST_BD_PKT_LEN_SHIFT;
- first_bd->data.bitfields |= cpu_to_le16(val);
- first_bd->data.nbds = 1;
+ bd->data.bitfields = cpu_to_le16(val);
/* We can safely ignore the offset, as it's 0 for XDP */
- BD_SET_UNMAP_ADDR_LEN(first_bd, metadata->mapping + padding, length);
+ BD_SET_UNMAP_ADDR_LEN(bd, dma + pad, len);
- /* Synchronize the buffer back to device, as program [probably]
- * has changed it.
- */
- dma_sync_single_for_device(&edev->pdev->dev,
- metadata->mapping + padding,
- length, PCI_DMA_TODEVICE);
+ xdp = txq->sw_tx_ring.xdp + txq->sw_tx_prod;
+ xdp->mapping = dma;
+ xdp->page = page;
- txq->sw_tx_ring.xdp[idx].page = metadata->data;
- txq->sw_tx_ring.xdp[idx].mapping = metadata->mapping;
txq->sw_tx_prod = (txq->sw_tx_prod + 1) % txq->num_tx_buffers;
- /* Mark the fastpath for future XDP doorbell */
- fp->xdp_xmit = 1;
-
return 0;
}
@@ -362,20 +351,21 @@ int qede_txq_has_work(struct qede_tx_queue *txq)
static void qede_xdp_tx_int(struct qede_dev *edev, struct qede_tx_queue *txq)
{
- u16 hw_bd_cons, idx;
+ struct sw_tx_xdp *xdp_info, *xdp_arr = txq->sw_tx_ring.xdp;
+ struct device *dev = &edev->pdev->dev;
+ u16 hw_bd_cons;
hw_bd_cons = le16_to_cpu(*txq->hw_cons_ptr);
barrier();
while (hw_bd_cons != qed_chain_get_cons_idx(&txq->tx_pbl)) {
- qed_chain_consume(&txq->tx_pbl);
- idx = txq->sw_tx_cons;
+ xdp_info = xdp_arr + txq->sw_tx_cons;
- dma_unmap_page(&edev->pdev->dev,
- txq->sw_tx_ring.xdp[idx].mapping,
- PAGE_SIZE, DMA_BIDIRECTIONAL);
- __free_page(txq->sw_tx_ring.xdp[idx].page);
+ dma_unmap_page(dev, xdp_info->mapping, PAGE_SIZE,
+ DMA_BIDIRECTIONAL);
+ __free_page(xdp_info->page);
+ qed_chain_consume(&txq->tx_pbl);
txq->sw_tx_cons = (txq->sw_tx_cons + 1) % txq->num_tx_buffers;
txq->xmit_pkts++;
}
@@ -1064,32 +1054,39 @@ static bool qede_rx_xdp(struct qede_dev *edev,
switch (act) {
case XDP_TX:
/* We need the replacement buffer before transmit. */
- if (qede_alloc_rx_buffer(rxq, true)) {
+ if (unlikely(qede_alloc_rx_buffer(rxq, true))) {
qede_recycle_rx_bd_ring(rxq, 1);
+
trace_xdp_exception(edev->ndev, prog, act);
- return false;
+ break;
}
/* Now if there's a transmission problem, we'd still have to
* throw current buffer, as replacement was already allocated.
*/
- if (qede_xdp_xmit(edev, fp, bd, *data_offset, *len)) {
- dma_unmap_page(rxq->dev, bd->mapping,
- PAGE_SIZE, DMA_BIDIRECTIONAL);
+ if (unlikely(qede_xdp_xmit(fp->xdp_tx, bd->mapping,
+ *data_offset, *len, bd->data))) {
+ dma_unmap_page(rxq->dev, bd->mapping, PAGE_SIZE,
+ rxq->data_direction);
__free_page(bd->data);
+
trace_xdp_exception(edev->ndev, prog, act);
+ } else {
+ dma_sync_single_for_device(rxq->dev,
+ bd->mapping + *data_offset,
+ *len, rxq->data_direction);
+ fp->xdp_xmit |= QEDE_XDP_TX;
}
/* Regardless, we've consumed an Rx BD */
qede_rx_bd_ring_consume(rxq);
- return false;
-
+ break;
default:
bpf_warn_invalid_xdp_action(act);
- /* Fall through */
+ fallthrough;
case XDP_ABORTED:
trace_xdp_exception(edev->ndev, prog, act);
- /* Fall through */
+ fallthrough;
case XDP_DROP:
qede_recycle_rx_bd_ring(rxq, cqe->bd_num);
}
@@ -1353,6 +1350,9 @@ int qede_poll(struct napi_struct *napi, int budget)
napi);
struct qede_dev *edev = fp->edev;
int rx_work_done = 0;
+ u16 xdp_prod;
+
+ fp->xdp_xmit = 0;
if (likely(fp->type & QEDE_FASTPATH_TX)) {
int cos;
@@ -1380,10 +1380,9 @@ int qede_poll(struct napi_struct *napi, int budget)
}
}
- if (fp->xdp_xmit) {
- u16 xdp_prod = qed_chain_get_prod_idx(&fp->xdp_tx->tx_pbl);
+ if (fp->xdp_xmit & QEDE_XDP_TX) {
+ xdp_prod = qed_chain_get_prod_idx(&fp->xdp_tx->tx_pbl);
- fp->xdp_xmit = 0;
fp->xdp_tx->tx_db.data.bd_prod = cpu_to_le16(xdp_prod);
qede_update_tx_producer(fp->xdp_tx);
}
--
2.25.1
Powered by blists - more mailing lists