Merge branch 'cpsw-Add-XDP-support'

Ivan Khoronzhuk says:

====================
net: ethernet: ti: cpsw: Add XDP support

This patchset adds XDP support for TI cpsw driver and base it on
page_pool allocator. It was verified on af_xdp socket drop,
af_xdp l2f, ebpf XDP_DROP, XDP_REDIRECT, XDP_PASS, XDP_TX.

It was verified with following configs enabled:
CONFIG_JIT=y
CONFIG_BPFILTER=y
CONFIG_BPF_SYSCALL=y
CONFIG_XDP_SOCKETS=y
CONFIG_BPF_EVENTS=y
CONFIG_HAVE_EBPF_JIT=y
CONFIG_BPF_JIT=y
CONFIG_CGROUP_BPF=y

Link on previous v7:
https://lkml.org/lkml/2019/7/4/715

Also regular tests with iperf2 were done in order to verify impact on
regular netstack performance, compared with base commit:
https://pastebin.com/JSMT0iZ4

v8..v9:
- fix warnings on arm64 caused by typos in type casting

v7..v8:
- corrected dma calculation based on headroom instead of hard start
- minor comment changes

v6..v7:
- rolled back to v4 solution but with small modification
- picked up patch:
  https://www.spinics.net/lists/netdev/msg583145.html
- added changes related to netsec fix and cpsw

v5..v6:
- do changes that is rx_dev while redirect/flush cycle is kept the same
- dropped net: ethernet: ti: davinci_cpdma: return handler status
- other changes desc in patches

v4..v5:
- added two plreliminary patches:
  net: ethernet: ti: davinci_cpdma: allow desc split while down
  net: ethernet: ti: cpsw_ethtool: allow res split while down
- added xdp alocator refcnt on xdp level, avoiding page pool refcnt
- moved flush status as separate argument for cpdma_chan_process
- reworked cpsw code according to last changes to allocator
- added missed statistic counter

v3..v4:
- added page pool user counter
- use same pool for ndevs in dual mac
- restructured page pool create/destroy according to the last changes in API

v2..v3:
- each rxq and ndev has its own page pool

v1..v2:
- combined xdp_xmit functions
- used page allocation w/o refcnt juggle
- unmapped page for skb netstack
- moved rxq/page pool allocation to open/close pair
- added several preliminary patches:
  net: page_pool: add helper function to retrieve dma addresses
  net: page_pool: add helper function to unmap dma addresses
  net: ethernet: ti: cpsw: use cpsw as drv data
  net: ethernet: ti: cpsw_ethtool: simplify slave loops
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/mellanox/mlx5/core/en_main.c

@@ -577,8 +577,6 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 		}
 		err = xdp_rxq_info_reg_mem_model(&rq->xdp_rxq,
 						 MEM_TYPE_PAGE_POOL, rq->page_pool);
-		if (err)
-			page_pool_free(rq->page_pool);
 	}
 	if (err)
 		goto err_free;
@@ -646,6 +644,7 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 	if (rq->xdp_prog)
 		bpf_prog_put(rq->xdp_prog);
 	xdp_rxq_info_unreg(&rq->xdp_rxq);
+	page_pool_destroy(rq->page_pool);
 	mlx5_wq_destroy(&rq->wq_ctrl);
 
 	return err;
@@ -680,6 +679,7 @@ static void mlx5e_free_rq(struct mlx5e_rq *rq)
 	}
 
 	xdp_rxq_info_unreg(&rq->xdp_rxq);
+	page_pool_destroy(rq->page_pool);
 	mlx5_wq_destroy(&rq->wq_ctrl);
 }
 

drivers/net/ethernet/socionext/netsec.c

@@ -1212,15 +1212,11 @@ static void netsec_uninit_pkt_dring(struct netsec_priv *priv, int id)
 		}
 	}
 
-	/* Rx is currently using page_pool
-	 * since the pool is created during netsec_setup_rx_dring(), we need to
-	 * free the pool manually if the registration failed
-	 */
+	/* Rx is currently using page_pool */
 	if (id == NETSEC_RING_RX) {
 		if (xdp_rxq_info_is_reg(&dring->xdp_rxq))
 			xdp_rxq_info_unreg(&dring->xdp_rxq);
-		else
-			page_pool_free(dring->page_pool);
+		page_pool_destroy(dring->page_pool);
 	}
 
 	memset(dring->desc, 0, sizeof(struct netsec_desc) * DESC_NUM);

drivers/net/ethernet/ti/Kconfig

@@ -50,6 +50,7 @@ config TI_CPSW
 	depends on ARCH_DAVINCI || ARCH_OMAP2PLUS || COMPILE_TEST
 	select TI_DAVINCI_MDIO
 	select MFD_SYSCON
+	select PAGE_POOL
 	select REGMAP
 	---help---
 	  This driver supports TI's CPSW Ethernet Switch.