net/funeth: Fix fun_xdp_tx() and XDP packet reclaim

The current implementation of fun_xdp_tx(), used for XPD_TX, is
incorrect in that it takes an address/length pair and later releases it
with page_frag_free(). It is OK for XDP_TX but the same code is used by
ndo_xdp_xmit. In that case it loses the XDP memory type and releases the
packet incorrectly for some of the types. Assorted breakage follows.

Change fun_xdp_tx() to take xdp_frame and rely on xdp_return_frame() in
reclaim.

Fixes: db37bc177dae ("net/funeth: add the data path")
Signed-off-by: Dimitris Michailidis <dmichail@fungible.com>
Link: https://lore.kernel.org/r/20220726215923.7887-1-dmichail@fungible.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

diff --git a/drivers/net/ethernet/fungible/funeth/funeth_rx.c b/drivers/net/ethernet/fungible/funeth/funeth_rx.c
index 0f6a549..29a6c2e 100644 (file)
--- a/drivers/net/ethernet/fungible/funeth/funeth_rx.c
+++ b/drivers/net/ethernet/fungible/funeth/funeth_rx.c
@@ -142,6 +142,7 @@ static void *fun_run_xdp(struct funeth_rxq *q, skb_frag_t *frags, void *buf_va,
                         int ref_ok, struct funeth_txq *xdp_q)
 {
        struct bpf_prog *xdp_prog;
+       struct xdp_frame *xdpf;
        struct xdp_buff xdp;
        u32 act;
 
@@ -163,7 +164,9 @@ static void *fun_run_xdp(struct funeth_rxq *q, skb_frag_t *frags, void *buf_va,
        case XDP_TX:
                if (unlikely(!ref_ok))
                        goto pass;
-               if (!fun_xdp_tx(xdp_q, xdp.data, xdp.data_end - xdp.data))
+
+               xdpf = xdp_convert_buff_to_frame(&xdp);
+               if (!xdpf || !fun_xdp_tx(xdp_q, xdpf))
                        goto xdp_error;
                FUN_QSTAT_INC(q, xdp_tx);
                q->xdp_flush |= FUN_XDP_FLUSH_TX;

diff --git a/drivers/net/ethernet/fungible/funeth/funeth_tx.c b/drivers/net/ethernet/fungible/funeth/funeth_tx.c
index ff6e292..2f6698b 100644 (file)
--- a/drivers/net/ethernet/fungible/funeth/funeth_tx.c
+++ b/drivers/net/ethernet/fungible/funeth/funeth_tx.c
@@ -466,7 +466,7 @@ static unsigned int fun_xdpq_clean(struct funeth_txq *q, unsigned int budget)
 
                do {
                        fun_xdp_unmap(q, reclaim_idx);
-                       page_frag_free(q->info[reclaim_idx].vaddr);
+                       xdp_return_frame(q->info[reclaim_idx].xdpf);
 
                        trace_funeth_tx_free(q, reclaim_idx, 1, head);
 
@@ -479,11 +479,11 @@ static unsigned int fun_xdpq_clean(struct funeth_txq *q, unsigned int budget)
        return npkts;
 }
 
-bool fun_xdp_tx(struct funeth_txq *q, void *data, unsigned int len)
+bool fun_xdp_tx(struct funeth_txq *q, struct xdp_frame *xdpf)
 {
        struct fun_eth_tx_req *req;
        struct fun_dataop_gl *gle;
-       unsigned int idx;
+       unsigned int idx, len;
        dma_addr_t dma;
 
        if (fun_txq_avail(q) < FUN_XDP_CLEAN_THRES)
@@ -494,7 +494,8 @@ bool fun_xdp_tx(struct funeth_txq *q, void *data, unsigned int len)
                return false;
        }
 
-       dma = dma_map_single(q->dma_dev, data, len, DMA_TO_DEVICE);
+       len = xdpf->len;
+       dma = dma_map_single(q->dma_dev, xdpf->data, len, DMA_TO_DEVICE);
        if (unlikely(dma_mapping_error(q->dma_dev, dma))) {
                FUN_QSTAT_INC(q, tx_map_err);
                return false;
@@ -514,7 +515,7 @@ bool fun_xdp_tx(struct funeth_txq *q, void *data, unsigned int len)
        gle = (struct fun_dataop_gl *)req->dataop.imm;
        fun_dataop_gl_init(gle, 0, 0, len, dma);
 
-       q->info[idx].vaddr = data;
+       q->info[idx].xdpf = xdpf;
 
        u64_stats_update_begin(&q->syncp);
        q->stats.tx_bytes += len;
@@ -545,12 +546,9 @@ int fun_xdp_xmit_frames(struct net_device *dev, int n,
        if (unlikely(q_idx >= fp->num_xdpqs))
                return -ENXIO;
 
-       for (q = xdpqs[q_idx], i = 0; i < n; i++) {
-               const struct xdp_frame *xdpf = frames[i];
-
-               if (!fun_xdp_tx(q, xdpf->data, xdpf->len))
+       for (q = xdpqs[q_idx], i = 0; i < n; i++)
+               if (!fun_xdp_tx(q, frames[i]))
                        break;
-       }
 
        if (unlikely(flags & XDP_XMIT_FLUSH))
                fun_txq_wr_db(q);
@@ -577,7 +575,7 @@ static void fun_xdpq_purge(struct funeth_txq *q)
                unsigned int idx = q->cons_cnt & q->mask;
 
                fun_xdp_unmap(q, idx);
-               page_frag_free(q->info[idx].vaddr);
+               xdp_return_frame(q->info[idx].xdpf);
                q->cons_cnt++;
        }
 }

diff --git a/drivers/net/ethernet/fungible/funeth/funeth_txrx.h b/drivers/net/ethernet/fungible/funeth/funeth_txrx.h
index 04c9f91..8708e28 100644 (file)
--- a/drivers/net/ethernet/fungible/funeth/funeth_txrx.h
+++ b/drivers/net/ethernet/fungible/funeth/funeth_txrx.h
@@ -95,8 +95,8 @@ struct funeth_txq_stats {  /* per Tx queue SW counters */
 
 struct funeth_tx_info {      /* per Tx descriptor state */
        union {
-               struct sk_buff *skb; /* associated packet */
-               void *vaddr;         /* start address for XDP */
+               struct sk_buff *skb;    /* associated packet (sk_buff path) */
+               struct xdp_frame *xdpf; /* associated XDP frame (XDP path) */
        };
 };
 
@@ -245,7 +245,7 @@ static inline int fun_irq_node(const struct fun_irq *p)
 int fun_rxq_napi_poll(struct napi_struct *napi, int budget);
 int fun_txq_napi_poll(struct napi_struct *napi, int budget);
 netdev_tx_t fun_start_xmit(struct sk_buff *skb, struct net_device *netdev);
-bool fun_xdp_tx(struct funeth_txq *q, void *data, unsigned int len);
+bool fun_xdp_tx(struct funeth_txq *q, struct xdp_frame *xdpf);
 int fun_xdp_xmit_frames(struct net_device *dev, int n,
                        struct xdp_frame **frames, u32 flags);