[PATCH bpf-next v4 00/15] Introduce AF_XDP buffer allocation API

[PATCH bpf-next v4 00/15] Introduce AF_XDP buffer allocation API

[Björn Töpel]

Overview
========

Driver adoption for AF_XDP has been slow. The amount of code required
to proper support AF_XDP is substantial and the driver/core APIs are
vague or even non-existing. Drivers have to manually adjust data
offsets, updating AF_XDP handles differently for different modes
(aligned/unaligned).

This series attempts to improve the situation by introducing an AF_XDP
buffer allocation API. The implementation is based on a single core
(single producer/consumer) buffer pool for the AF_XDP UMEM.
    
A buffer is allocated using the xsk_buff_alloc() function, and
returned using xsk_buff_free(). If a buffer is disassociated with the
pool, e.g. when a buffer is passed to an AF_XDP socket, a buffer is
said to be released. Currently, the release function is only used by
the AF_XDP internals and not visible to the driver.
    
Drivers using this API should register the XDP memory model with the
new MEM_TYPE_XSK_BUFF_POOL type, which will supersede the
MEM_TYPE_ZERO_COPY type.

The buffer type is struct xdp_buff, and follows the lifetime of
regular xdp_buffs, i.e.  the lifetime of an xdp_buff is restricted to
a NAPI context. In other words, the API is not replacing xdp_frames.

DMA mapping/synching is folded into the buffer handling as well.

@JeffK The Intel drivers changes should go through the bpf-next tree,
       and not your regular Intel tree, since multiple (non-Intel)
       drivers are affected.

The outline of the series is as following:

Patch 1 is a fix for xsk_umem_xdp_frame_sz().

Patch 2 to 4 are restructures/clean ups. The XSKMAP implementation is
moved to net/xdp/. Functions/defines/enums that are only used by the
AF_XDP internals are moved from the global include/net/xdp_sock.h to
net/xdp/xsk.h. We are also introducing a new "driver include file",
include/net/xdp_sock_drv.h, which is the only file NIC driver
developers adding AF_XDP zero-copy support should care about.

Patch 5 adds the new API, and migrates the "copy-mode"/skb-mode AF_XDP
path to the new API.

Patch 6 to 11 migrates the existing zero-copy drivers to the new API.

Patch 12 removes the MEM_TYPE_ZERO_COPY memory type, and the "handle"
member of struct xdp_buff.

Patch 13 simplifies the xdp_return_{frame,frame_rx_napi,buff}
functions.

Patch 14 is a performance patch, where some functions are inlined.

Finally, patch 15 updates the MAINTAINERS file to correctly mirror the
new file layout.

Note that this series removes the "handle" member from struct
xdp_buff, which reduces the xdp_buff size.

After this series, the diff stat of drivers/net/ is:

 
This series is a first step of simplifying the driver side of
AF_XDP. I think more of the AF_XDP logic can be moved from the drivers
to the AF_XDP core, e.g. the "need wakeup" set/clear functionality.

Statistics when allocation fails can now be added to the socket
statistics via the XDP_STATISTICS getsockopt(). This will be added in
a follow up series.


Performance
===========

As a nice side effect, performance is up a bit as well.

  * i40e: 3% higher pps for rxdrop, zero-copy, aligned and unaligned
    (40 GbE, 64B packets).
  * mlx5: RX +0.8 Mpps, TX +0.4 Mpps


Changelog
=========

v3->v4:
    * i40e: Made i40e_fd_handle_status() static. (kbuild test robot)
    * mlx5: Remove unused variable num_xsk_frames. (Jakub)

v2->v3: 
  * Added xsk_umem_xdp_frame_sz() fix to the series. (Björn)
  * Initialize struct xdp_buff member frame_sz. (Björn)
  * Add API to query the DMA address of a frame. (Maxim)
  * Do DMA sync for CPU till the end of the frame to handle possible
    growth (frame_sz). (Maxim)
  * mlx5: Handle frame_sz, use xsk_buff_xdp_get_frame_dma, use
    xsk_buff API for DMA sync on TX, add performance numbers. (Maxim)

v1->v2: 
  * mlx5: Fix DMA address handling, set XDP metadata to invalid. (Maxim)
  * ixgbe: Fixed xdp_buff data_end update. (Björn)
  * Swapped SoBs in patch 4. (Maxim)

rfc->v1:
  * Fixed build errors/warnings for m68k and riscv. (kbuild test
    robot)
  * Added headroom/chunk size getter. (Maxim/Björn)
  * mlx5: Put back the sanity check for XSK params, use XSK API to get
    the total headroom size. (Maxim)
  * Fixed spelling in commit message. (Björn)
  * Make sure xp_validate_desc() is inlined for Tx perf. (Maxim)
  * Sorted file entries. (Joe)
  * Added xdp_return_{frame,frame_rx_napi,buff} simplification (Björn)

Thanks for all the comments/input/help!


Cheers,
Björn

Björn Töpel (14):
  xsk: fix xsk_umem_xdp_frame_sz()
  xsk: move xskmap.c to net/xdp/
  xsk: move defines only used by AF_XDP internals to xsk.h
  xsk: introduce AF_XDP buffer allocation API
  i40e: refactor rx_bi accesses
  i40e: separate kernel allocated rx_bi rings from AF_XDP rings
  i40e, xsk: migrate to new MEM_TYPE_XSK_BUFF_POOL
  ice, xsk: migrate to new MEM_TYPE_XSK_BUFF_POOL
  ixgbe, xsk: migrate to new MEM_TYPE_XSK_BUFF_POOL
  mlx5, xsk: migrate to new MEM_TYPE_XSK_BUFF_POOL
  xsk: remove MEM_TYPE_ZERO_COPY and corresponding code
  xdp: simplify xdp_return_{frame,frame_rx_napi,buff}
  xsk: explicitly inline functions and move definitions
  MAINTAINERS, xsk: update AF_XDP section after moves/adds

Magnus Karlsson (1):
  xsk: move driver interface to xdp_sock_drv.h

 MAINTAINERS                                   |   6 +-
 drivers/net/ethernet/intel/i40e/i40e_main.c   |  28 +-
 drivers/net/ethernet/intel/i40e/i40e_txrx.c   | 134 +++----
 drivers/net/ethernet/intel/i40e/i40e_txrx.h   |  17 +-
 .../ethernet/intel/i40e/i40e_txrx_common.h    |  40 +-
 drivers/net/ethernet/intel/i40e/i40e_type.h   |   5 +-
 drivers/net/ethernet/intel/i40e/i40e_xsk.c    | 378 +++---------------
 drivers/net/ethernet/intel/i40e/i40e_xsk.h    |   3 +-
 drivers/net/ethernet/intel/ice/ice_base.c     |  16 +-
 drivers/net/ethernet/intel/ice/ice_txrx.h     |   8 +-
 drivers/net/ethernet/intel/ice/ice_xsk.c      | 376 ++---------------
 drivers/net/ethernet/intel/ice/ice_xsk.h      |  13 +-
 drivers/net/ethernet/intel/ixgbe/ixgbe.h      |   9 +-
 drivers/net/ethernet/intel/ixgbe/ixgbe_main.c |  15 +-
 .../ethernet/intel/ixgbe/ixgbe_txrx_common.h  |   2 +-
 drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c  | 309 +++-----------
 drivers/net/ethernet/mellanox/mlx5/core/en.h  |   7 +-
 .../ethernet/mellanox/mlx5/core/en/params.c   |  13 +-
 .../net/ethernet/mellanox/mlx5/core/en/xdp.c  |  33 +-
 .../net/ethernet/mellanox/mlx5/core/en/xdp.h  |   2 +-
 .../ethernet/mellanox/mlx5/core/en/xsk/rx.c   | 113 +-----
 .../ethernet/mellanox/mlx5/core/en/xsk/rx.h   |  25 +-
 .../ethernet/mellanox/mlx5/core/en/xsk/tx.c   |   9 +-
 .../ethernet/mellanox/mlx5/core/en/xsk/tx.h   |   2 +-
 .../ethernet/mellanox/mlx5/core/en/xsk/umem.c |  51 +--
 .../net/ethernet/mellanox/mlx5/core/en_main.c |  25 +-
 .../net/ethernet/mellanox/mlx5/core/en_rx.c   |  34 +-
 drivers/net/hyperv/netvsc_bpf.c               |   1 -
 include/net/xdp.h                             |   9 +-
 include/net/xdp_sock.h                        | 287 +------------
 include/net/xdp_sock_drv.h                    | 232 +++++++++++
 include/net/xsk_buff_pool.h                   | 140 +++++++
 include/trace/events/xdp.h                    |   2 +-
 kernel/bpf/Makefile                           |   3 -
 net/core/xdp.c                                |  51 +--
 net/ethtool/channels.c                        |   2 +-
 net/ethtool/ioctl.c                           |   2 +-
 net/xdp/Makefile                              |   3 +-
 net/xdp/xdp_umem.c                            |  55 +--
 net/xdp/xdp_umem.h                            |   2 +-
 net/xdp/xsk.c                                 | 204 ++++------
 net/xdp/xsk.h                                 |  30 ++
 net/xdp/xsk_buff_pool.c                       | 336 ++++++++++++++++
 net/xdp/xsk_diag.c                            |   2 +-
 net/xdp/xsk_queue.c                           |  62 ---
 net/xdp/xsk_queue.h                           | 117 ++----
 {kernel/bpf => net/xdp}/xskmap.c              |   2 +
 47 files changed, 1278 insertions(+), 1937 deletions(-)
 create mode 100644 include/net/xdp_sock_drv.h
 create mode 100644 include/net/xsk_buff_pool.h
 create mode 100644 net/xdp/xsk_buff_pool.c
 rename {kernel/bpf => net/xdp}/xskmap.c (99%)


base-commit: dda18a5c0b75461d1ed228f80b59c67434b8d601
-- 
2.25.1

[PATCH bpf-next v4 01/15] xsk: fix xsk_umem_xdp_frame_sz()

[Björn Töpel]

From: Björn Töpel <bjorn.topel@intel.com>

Calculating the "data_hard_end" for an XDP buffer coming from AF_XDP
zero-copy mode, the return value of xsk_umem_xdp_frame_sz() is added
to "data_hard_start".

Currently, the chunk size of the UMEM is returned by
xsk_umem_xdp_frame_sz(). This is not correct, if the fixed UMEM
headroom is non-zero. Fix this by returning the chunk_size without the
UMEM headroom.

Fixes: 2a637c5b1aaf ("xdp: For Intel AF_XDP drivers add XDP frame_sz")
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
 include/net/xdp_sock.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/net/xdp_sock.h b/include/net/xdp_sock.h
index abd72de25fa4..6b1137ce1692 100644
--- a/include/net/xdp_sock.h
+++ b/include/net/xdp_sock.h
@@ -239,7 +239,7 @@ static inline u64 xsk_umem_adjust_offset(struct xdp_umem *umem, u64 address,
 
 static inline u32 xsk_umem_xdp_frame_sz(struct xdp_umem *umem)
 {
-	return umem->chunk_size_nohr + umem->headroom;
+	return umem->chunk_size_nohr;
 }
 
 #else
-- 
2.25.1

[PATCH bpf-next v4 02/15] xsk: move xskmap.c to net/xdp/

[Björn Töpel]

From: Björn Töpel <bjorn.topel@intel.com>

The XSKMAP is partly implemented by net/xdp/xsk.c. Move xskmap.c from
kernel/bpf/ to net/xdp/, which is the logical place for AF_XDP related
code. Also, move AF_XDP struct definitions, and function declarations
only used by AF_XDP internals into net/xdp/xsk.h.

Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
 include/net/xdp_sock.h           | 20 --------------------
 kernel/bpf/Makefile              |  3 ---
 net/xdp/Makefile                 |  2 +-
 net/xdp/xsk.h                    | 16 ++++++++++++++++
 {kernel/bpf => net/xdp}/xskmap.c |  2 ++
 5 files changed, 19 insertions(+), 24 deletions(-)
 rename {kernel/bpf => net/xdp}/xskmap.c (99%)

diff --git a/include/net/xdp_sock.h b/include/net/xdp_sock.h
index 6b1137ce1692..8f3f6f5b0dfe 100644
--- a/include/net/xdp_sock.h
+++ b/include/net/xdp_sock.h
@@ -65,22 +65,12 @@ struct xdp_umem {
 	struct list_head xsk_tx_list;
 };
 
-/* Nodes are linked in the struct xdp_sock map_list field, and used to
- * track which maps a certain socket reside in.
- */
-
 struct xsk_map {
 	struct bpf_map map;
 	spinlock_t lock; /* Synchronize map updates */
 	struct xdp_sock *xsk_map[];
 };
 
-struct xsk_map_node {
-	struct list_head node;
-	struct xsk_map *map;
-	struct xdp_sock **map_entry;
-};
-
 struct xdp_sock {
 	/* struct sock must be the first member of struct xdp_sock */
 	struct sock sk;
@@ -114,7 +104,6 @@ struct xdp_sock {
 struct xdp_buff;
 #ifdef CONFIG_XDP_SOCKETS
 int xsk_generic_rcv(struct xdp_sock *xs, struct xdp_buff *xdp);
-bool xsk_is_setup_for_bpf_map(struct xdp_sock *xs);
 /* Used from netdev driver */
 bool xsk_umem_has_addrs(struct xdp_umem *umem, u32 cnt);
 bool xsk_umem_peek_addr(struct xdp_umem *umem, u64 *addr);
@@ -133,10 +122,6 @@ void xsk_clear_rx_need_wakeup(struct xdp_umem *umem);
 void xsk_clear_tx_need_wakeup(struct xdp_umem *umem);
 bool xsk_umem_uses_need_wakeup(struct xdp_umem *umem);
 
-void xsk_map_try_sock_delete(struct xsk_map *map, struct xdp_sock *xs,
-			     struct xdp_sock **map_entry);
-int xsk_map_inc(struct xsk_map *map);
-void xsk_map_put(struct xsk_map *map);
 int __xsk_map_redirect(struct xdp_sock *xs, struct xdp_buff *xdp);
 void __xsk_map_flush(void);
 
@@ -248,11 +233,6 @@ static inline int xsk_generic_rcv(struct xdp_sock *xs, struct xdp_buff *xdp)
 	return -ENOTSUPP;
 }
 
-static inline bool xsk_is_setup_for_bpf_map(struct xdp_sock *xs)
-{
-	return false;
-}
-
 static inline bool xsk_umem_has_addrs(struct xdp_umem *umem, u32 cnt)
 {
 	return false;
diff --git a/kernel/bpf/Makefile b/kernel/bpf/Makefile
index 37b2d8620153..375b933010dd 100644
--- a/kernel/bpf/Makefile
+++ b/kernel/bpf/Makefile
@@ -12,9 +12,6 @@ obj-$(CONFIG_BPF_JIT) += dispatcher.o
 ifeq ($(CONFIG_NET),y)
 obj-$(CONFIG_BPF_SYSCALL) += devmap.o
 obj-$(CONFIG_BPF_SYSCALL) += cpumap.o
-ifeq ($(CONFIG_XDP_SOCKETS),y)
-obj-$(CONFIG_BPF_SYSCALL) += xskmap.o
-endif
 obj-$(CONFIG_BPF_SYSCALL) += offload.o
 endif
 ifeq ($(CONFIG_PERF_EVENTS),y)
diff --git a/net/xdp/Makefile b/net/xdp/Makefile
index 71e2bdafb2ce..90b5460d6166 100644
--- a/net/xdp/Makefile
+++ b/net/xdp/Makefile
@@ -1,3 +1,3 @@
 # SPDX-License-Identifier: GPL-2.0-only
-obj-$(CONFIG_XDP_SOCKETS) += xsk.o xdp_umem.o xsk_queue.o
+obj-$(CONFIG_XDP_SOCKETS) += xsk.o xdp_umem.o xsk_queue.o xskmap.o
 obj-$(CONFIG_XDP_SOCKETS_DIAG) += xsk_diag.o
diff --git a/net/xdp/xsk.h b/net/xdp/xsk.h
index 4cfd106bdb53..d6a0979050e6 100644
--- a/net/xdp/xsk.h
+++ b/net/xdp/xsk.h
@@ -17,9 +17,25 @@ struct xdp_mmap_offsets_v1 {
 	struct xdp_ring_offset_v1 cr;
 };
 
+/* Nodes are linked in the struct xdp_sock map_list field, and used to
+ * track which maps a certain socket reside in.
+ */
+
+struct xsk_map_node {
+	struct list_head node;
+	struct xsk_map *map;
+	struct xdp_sock **map_entry;
+};
+
 static inline struct xdp_sock *xdp_sk(struct sock *sk)
 {
 	return (struct xdp_sock *)sk;
 }
 
+bool xsk_is_setup_for_bpf_map(struct xdp_sock *xs);
+void xsk_map_try_sock_delete(struct xsk_map *map, struct xdp_sock *xs,
+			     struct xdp_sock **map_entry);
+int xsk_map_inc(struct xsk_map *map);
+void xsk_map_put(struct xsk_map *map);
+
 #endif /* XSK_H_ */
diff --git a/kernel/bpf/xskmap.c b/net/xdp/xskmap.c
similarity index 99%
rename from kernel/bpf/xskmap.c
rename to net/xdp/xskmap.c
index 2cc5c8f4c800..1dc7208c71ba 100644
--- a/kernel/bpf/xskmap.c
+++ b/net/xdp/xskmap.c
@@ -9,6 +9,8 @@
 #include <linux/slab.h>
 #include <linux/sched.h>
 
+#include "xsk.h"
+
 int xsk_map_inc(struct xsk_map *map)
 {
 	bpf_map_inc(&map->map);
-- 
2.25.1

[PATCH bpf-next v4 03/15] xsk: move driver interface to xdp_sock_drv.h

[Björn Töpel]

From: Magnus Karlsson <magnus.karlsson@intel.com>

Move the AF_XDP zero-copy driver interface to its own include file
called xdp_sock_drv.h. This, hopefully, will make it more clear for
NIC driver implementors to know what functions to use for zero-copy
support.

Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_main.c   |   2 +-
 drivers/net/ethernet/intel/i40e/i40e_xsk.c    |   2 +-
 drivers/net/ethernet/intel/ice/ice_xsk.c      |   2 +-
 drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c  |   2 +-
 .../net/ethernet/mellanox/mlx5/core/en/xdp.c  |   2 +-
 .../ethernet/mellanox/mlx5/core/en/xsk/rx.h   |   2 +-
 .../ethernet/mellanox/mlx5/core/en/xsk/tx.h   |   2 +-
 .../ethernet/mellanox/mlx5/core/en/xsk/umem.c |   2 +-
 include/net/xdp_sock.h                        | 214 +----------------
 include/net/xdp_sock_drv.h                    | 217 ++++++++++++++++++
 net/ethtool/channels.c                        |   2 +-
 net/ethtool/ioctl.c                           |   2 +-
 net/xdp/xdp_umem.h                            |   2 +-
 net/xdp/xsk.c                                 |   2 +-
 14 files changed, 237 insertions(+), 218 deletions(-)
 create mode 100644 include/net/xdp_sock_drv.h

diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index 2a037ec244b9..d6b2db4f2c65 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -11,7 +11,7 @@
 #include "i40e_diag.h"
 #include "i40e_xsk.h"
 #include <net/udp_tunnel.h>
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 /* All i40e tracepoints are defined by the include below, which
  * must be included exactly once across the whole kernel with
  * CREATE_TRACE_POINTS defined
diff --git a/drivers/net/ethernet/intel/i40e/i40e_xsk.c b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
index 2b9184aead5f..d8b0be29099a 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_xsk.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
@@ -2,7 +2,7 @@
 /* Copyright(c) 2018 Intel Corporation. */
 
 #include <linux/bpf_trace.h>
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 #include <net/xdp.h>
 
 #include "i40e.h"
diff --git a/drivers/net/ethernet/intel/ice/ice_xsk.c b/drivers/net/ethernet/intel/ice/ice_xsk.c
index 23e5515d4527..70e204307a93 100644
--- a/drivers/net/ethernet/intel/ice/ice_xsk.c
+++ b/drivers/net/ethernet/intel/ice/ice_xsk.c
@@ -2,7 +2,7 @@
 /* Copyright (c) 2019, Intel Corporation. */
 
 #include <linux/bpf_trace.h>
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 #include <net/xdp.h>
 #include "ice.h"
 #include "ice_base.h"
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
index a656ee9a1fae..82e4effae704 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
@@ -2,7 +2,7 @@
 /* Copyright(c) 2018 Intel Corporation. */
 
 #include <linux/bpf_trace.h>
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 #include <net/xdp.h>
 
 #include "ixgbe.h"
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
index 761c8979bd41..3507d23f0eb8 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
@@ -31,7 +31,7 @@
  */
 
 #include <linux/bpf_trace.h>
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 #include "en/xdp.h"
 #include "en/params.h"
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.h b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.h
index cab0e93497ae..a8e11adbf426 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.h
@@ -5,7 +5,7 @@
 #define __MLX5_EN_XSK_RX_H__
 
 #include "en.h"
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 
 /* RX data path */
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/tx.h b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/tx.h
index 79b487d89757..39fa0a705856 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/tx.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/tx.h
@@ -5,7 +5,7 @@
 #define __MLX5_EN_XSK_TX_H__
 
 #include "en.h"
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 
 /* TX data path */
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/umem.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/umem.c
index 4baaa5788320..5e49fdb564b3 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/umem.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/umem.c
@@ -1,7 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
 /* Copyright (c) 2019 Mellanox Technologies. */
 
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 #include "umem.h"
 #include "setup.h"
 #include "en/params.h"
diff --git a/include/net/xdp_sock.h b/include/net/xdp_sock.h
index 8f3f6f5b0dfe..6a986dcbc336 100644
--- a/include/net/xdp_sock.h
+++ b/include/net/xdp_sock.h
@@ -15,6 +15,7 @@
 
 struct net_device;
 struct xsk_queue;
+struct xdp_buff;
 
 /* Masks for xdp_umem_page flags.
  * The low 12-bits of the addr will be 0 since this is the page address, so we
@@ -101,27 +102,9 @@ struct xdp_sock {
 	spinlock_t map_list_lock;
 };
 
-struct xdp_buff;
 #ifdef CONFIG_XDP_SOCKETS
-int xsk_generic_rcv(struct xdp_sock *xs, struct xdp_buff *xdp);
-/* Used from netdev driver */
-bool xsk_umem_has_addrs(struct xdp_umem *umem, u32 cnt);
-bool xsk_umem_peek_addr(struct xdp_umem *umem, u64 *addr);
-void xsk_umem_release_addr(struct xdp_umem *umem);
-void xsk_umem_complete_tx(struct xdp_umem *umem, u32 nb_entries);
-bool xsk_umem_consume_tx(struct xdp_umem *umem, struct xdp_desc *desc);
-void xsk_umem_consume_tx_done(struct xdp_umem *umem);
-struct xdp_umem_fq_reuse *xsk_reuseq_prepare(u32 nentries);
-struct xdp_umem_fq_reuse *xsk_reuseq_swap(struct xdp_umem *umem,
-					  struct xdp_umem_fq_reuse *newq);
-void xsk_reuseq_free(struct xdp_umem_fq_reuse *rq);
-struct xdp_umem *xdp_get_umem_from_qid(struct net_device *dev, u16 queue_id);
-void xsk_set_rx_need_wakeup(struct xdp_umem *umem);
-void xsk_set_tx_need_wakeup(struct xdp_umem *umem);
-void xsk_clear_rx_need_wakeup(struct xdp_umem *umem);
-void xsk_clear_tx_need_wakeup(struct xdp_umem *umem);
-bool xsk_umem_uses_need_wakeup(struct xdp_umem *umem);
 
+int xsk_generic_rcv(struct xdp_sock *xs, struct xdp_buff *xdp);
 int __xsk_map_redirect(struct xdp_sock *xs, struct xdp_buff *xdp);
 void __xsk_map_flush(void);
 
@@ -153,131 +136,24 @@ static inline u64 xsk_umem_add_offset_to_addr(u64 addr)
 	return xsk_umem_extract_addr(addr) + xsk_umem_extract_offset(addr);
 }
 
-static inline char *xdp_umem_get_data(struct xdp_umem *umem, u64 addr)
-{
-	unsigned long page_addr;
-
-	addr = xsk_umem_add_offset_to_addr(addr);
-	page_addr = (unsigned long)umem->pages[addr >> PAGE_SHIFT].addr;
-
-	return (char *)(page_addr & PAGE_MASK) + (addr & ~PAGE_MASK);
-}
-
-static inline dma_addr_t xdp_umem_get_dma(struct xdp_umem *umem, u64 addr)
-{
-	addr = xsk_umem_add_offset_to_addr(addr);
-
-	return umem->pages[addr >> PAGE_SHIFT].dma + (addr & ~PAGE_MASK);
-}
-
-/* Reuse-queue aware version of FILL queue helpers */
-static inline bool xsk_umem_has_addrs_rq(struct xdp_umem *umem, u32 cnt)
-{
-	struct xdp_umem_fq_reuse *rq = umem->fq_reuse;
-
-	if (rq->length >= cnt)
-		return true;
-
-	return xsk_umem_has_addrs(umem, cnt - rq->length);
-}
-
-static inline bool xsk_umem_peek_addr_rq(struct xdp_umem *umem, u64 *addr)
-{
-	struct xdp_umem_fq_reuse *rq = umem->fq_reuse;
-
-	if (!rq->length)
-		return xsk_umem_peek_addr(umem, addr);
-
-	*addr = rq->handles[rq->length - 1];
-	return addr;
-}
-
-static inline void xsk_umem_release_addr_rq(struct xdp_umem *umem)
-{
-	struct xdp_umem_fq_reuse *rq = umem->fq_reuse;
-
-	if (!rq->length)
-		xsk_umem_release_addr(umem);
-	else
-		rq->length--;
-}
-
-static inline void xsk_umem_fq_reuse(struct xdp_umem *umem, u64 addr)
-{
-	struct xdp_umem_fq_reuse *rq = umem->fq_reuse;
-
-	rq->handles[rq->length++] = addr;
-}
-
-/* Handle the offset appropriately depending on aligned or unaligned mode.
- * For unaligned mode, we store the offset in the upper 16-bits of the address.
- * For aligned mode, we simply add the offset to the address.
- */
-static inline u64 xsk_umem_adjust_offset(struct xdp_umem *umem, u64 address,
-					 u64 offset)
-{
-	if (umem->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG)
-		return address + (offset << XSK_UNALIGNED_BUF_OFFSET_SHIFT);
-	else
-		return address + offset;
-}
-
-static inline u32 xsk_umem_xdp_frame_sz(struct xdp_umem *umem)
-{
-	return umem->chunk_size_nohr;
-}
-
 #else
+
 static inline int xsk_generic_rcv(struct xdp_sock *xs, struct xdp_buff *xdp)
 {
 	return -ENOTSUPP;
 }
 
-static inline bool xsk_umem_has_addrs(struct xdp_umem *umem, u32 cnt)
-{
-	return false;
-}
-
-static inline u64 *xsk_umem_peek_addr(struct xdp_umem *umem, u64 *addr)
-{
-	return NULL;
-}
-
-static inline void xsk_umem_release_addr(struct xdp_umem *umem)
-{
-}
-
-static inline void xsk_umem_complete_tx(struct xdp_umem *umem, u32 nb_entries)
-{
-}
-
-static inline bool xsk_umem_consume_tx(struct xdp_umem *umem,
-				       struct xdp_desc *desc)
-{
-	return false;
-}
-
-static inline void xsk_umem_consume_tx_done(struct xdp_umem *umem)
-{
-}
-
-static inline struct xdp_umem_fq_reuse *xsk_reuseq_prepare(u32 nentries)
+static inline int __xsk_map_redirect(struct xdp_sock *xs, struct xdp_buff *xdp)
 {
-	return NULL;
+	return -EOPNOTSUPP;
 }
 
-static inline struct xdp_umem_fq_reuse *xsk_reuseq_swap(
-	struct xdp_umem *umem,
-	struct xdp_umem_fq_reuse *newq)
-{
-	return NULL;
-}
-static inline void xsk_reuseq_free(struct xdp_umem_fq_reuse *rq)
+static inline void __xsk_map_flush(void)
 {
 }
 
-static inline struct xdp_umem *xdp_get_umem_from_qid(struct net_device *dev,
-						     u16 queue_id)
+static inline struct xdp_sock *__xsk_map_lookup_elem(struct bpf_map *map,
+						     u32 key)
 {
 	return NULL;
 }
@@ -297,80 +173,6 @@ static inline u64 xsk_umem_add_offset_to_addr(u64 addr)
 	return 0;
 }
 
-static inline char *xdp_umem_get_data(struct xdp_umem *umem, u64 addr)
-{
-	return NULL;
-}
-
-static inline dma_addr_t xdp_umem_get_dma(struct xdp_umem *umem, u64 addr)
-{
-	return 0;
-}
-
-static inline bool xsk_umem_has_addrs_rq(struct xdp_umem *umem, u32 cnt)
-{
-	return false;
-}
-
-static inline u64 *xsk_umem_peek_addr_rq(struct xdp_umem *umem, u64 *addr)
-{
-	return NULL;
-}
-
-static inline void xsk_umem_release_addr_rq(struct xdp_umem *umem)
-{
-}
-
-static inline void xsk_umem_fq_reuse(struct xdp_umem *umem, u64 addr)
-{
-}
-
-static inline void xsk_set_rx_need_wakeup(struct xdp_umem *umem)
-{
-}
-
-static inline void xsk_set_tx_need_wakeup(struct xdp_umem *umem)
-{
-}
-
-static inline void xsk_clear_rx_need_wakeup(struct xdp_umem *umem)
-{
-}
-
-static inline void xsk_clear_tx_need_wakeup(struct xdp_umem *umem)
-{
-}
-
-static inline bool xsk_umem_uses_need_wakeup(struct xdp_umem *umem)
-{
-	return false;
-}
-
-static inline u64 xsk_umem_adjust_offset(struct xdp_umem *umem, u64 handle,
-					 u64 offset)
-{
-	return 0;
-}
-
-static inline u32 xsk_umem_xdp_frame_sz(struct xdp_umem *umem)
-{
-	return 0;
-}
-
-static inline int __xsk_map_redirect(struct xdp_sock *xs, struct xdp_buff *xdp)
-{
-	return -EOPNOTSUPP;
-}
-
-static inline void __xsk_map_flush(void)
-{
-}
-
-static inline struct xdp_sock *__xsk_map_lookup_elem(struct bpf_map *map,
-						     u32 key)
-{
-	return NULL;
-}
 #endif /* CONFIG_XDP_SOCKETS */
 
 #endif /* _LINUX_XDP_SOCK_H */
diff --git a/include/net/xdp_sock_drv.h b/include/net/xdp_sock_drv.h
new file mode 100644
index 000000000000..d67f2361937a
--- /dev/null
+++ b/include/net/xdp_sock_drv.h
@@ -0,0 +1,217 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Interface for implementing AF_XDP zero-copy support in drivers.
+ * Copyright(c) 2020 Intel Corporation.
+ */
+
+#ifndef _LINUX_XDP_SOCK_DRV_H
+#define _LINUX_XDP_SOCK_DRV_H
+
+#include <net/xdp_sock.h>
+
+#ifdef CONFIG_XDP_SOCKETS
+
+bool xsk_umem_has_addrs(struct xdp_umem *umem, u32 cnt);
+bool xsk_umem_peek_addr(struct xdp_umem *umem, u64 *addr);
+void xsk_umem_release_addr(struct xdp_umem *umem);
+void xsk_umem_complete_tx(struct xdp_umem *umem, u32 nb_entries);
+bool xsk_umem_consume_tx(struct xdp_umem *umem, struct xdp_desc *desc);
+void xsk_umem_consume_tx_done(struct xdp_umem *umem);
+struct xdp_umem_fq_reuse *xsk_reuseq_prepare(u32 nentries);
+struct xdp_umem_fq_reuse *xsk_reuseq_swap(struct xdp_umem *umem,
+					  struct xdp_umem_fq_reuse *newq);
+void xsk_reuseq_free(struct xdp_umem_fq_reuse *rq);
+struct xdp_umem *xdp_get_umem_from_qid(struct net_device *dev, u16 queue_id);
+void xsk_set_rx_need_wakeup(struct xdp_umem *umem);
+void xsk_set_tx_need_wakeup(struct xdp_umem *umem);
+void xsk_clear_rx_need_wakeup(struct xdp_umem *umem);
+void xsk_clear_tx_need_wakeup(struct xdp_umem *umem);
+bool xsk_umem_uses_need_wakeup(struct xdp_umem *umem);
+
+static inline char *xdp_umem_get_data(struct xdp_umem *umem, u64 addr)
+{
+	unsigned long page_addr;
+
+	addr = xsk_umem_add_offset_to_addr(addr);
+	page_addr = (unsigned long)umem->pages[addr >> PAGE_SHIFT].addr;
+
+	return (char *)(page_addr & PAGE_MASK) + (addr & ~PAGE_MASK);
+}
+
+static inline dma_addr_t xdp_umem_get_dma(struct xdp_umem *umem, u64 addr)
+{
+	addr = xsk_umem_add_offset_to_addr(addr);
+
+	return umem->pages[addr >> PAGE_SHIFT].dma + (addr & ~PAGE_MASK);
+}
+
+/* Reuse-queue aware version of FILL queue helpers */
+static inline bool xsk_umem_has_addrs_rq(struct xdp_umem *umem, u32 cnt)
+{
+	struct xdp_umem_fq_reuse *rq = umem->fq_reuse;
+
+	if (rq->length >= cnt)
+		return true;
+
+	return xsk_umem_has_addrs(umem, cnt - rq->length);
+}
+
+static inline bool xsk_umem_peek_addr_rq(struct xdp_umem *umem, u64 *addr)
+{
+	struct xdp_umem_fq_reuse *rq = umem->fq_reuse;
+
+	if (!rq->length)
+		return xsk_umem_peek_addr(umem, addr);
+
+	*addr = rq->handles[rq->length - 1];
+	return addr;
+}
+
+static inline void xsk_umem_release_addr_rq(struct xdp_umem *umem)
+{
+	struct xdp_umem_fq_reuse *rq = umem->fq_reuse;
+
+	if (!rq->length)
+		xsk_umem_release_addr(umem);
+	else
+		rq->length--;
+}
+
+static inline void xsk_umem_fq_reuse(struct xdp_umem *umem, u64 addr)
+{
+	struct xdp_umem_fq_reuse *rq = umem->fq_reuse;
+
+	rq->handles[rq->length++] = addr;
+}
+
+/* Handle the offset appropriately depending on aligned or unaligned mode.
+ * For unaligned mode, we store the offset in the upper 16-bits of the address.
+ * For aligned mode, we simply add the offset to the address.
+ */
+static inline u64 xsk_umem_adjust_offset(struct xdp_umem *umem, u64 address,
+					 u64 offset)
+{
+	if (umem->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG)
+		return address + (offset << XSK_UNALIGNED_BUF_OFFSET_SHIFT);
+	else
+		return address + offset;
+}
+
+static inline u32 xsk_umem_xdp_frame_sz(struct xdp_umem *umem)
+{
+	return umem->chunk_size_nohr;
+}
+
+#else
+
+static inline bool xsk_umem_has_addrs(struct xdp_umem *umem, u32 cnt)
+{
+	return false;
+}
+
+static inline u64 *xsk_umem_peek_addr(struct xdp_umem *umem, u64 *addr)
+{
+	return NULL;
+}
+
+static inline void xsk_umem_release_addr(struct xdp_umem *umem)
+{
+}
+
+static inline void xsk_umem_complete_tx(struct xdp_umem *umem, u32 nb_entries)
+{
+}
+
+static inline bool xsk_umem_consume_tx(struct xdp_umem *umem,
+				       struct xdp_desc *desc)
+{
+	return false;
+}
+
+static inline void xsk_umem_consume_tx_done(struct xdp_umem *umem)
+{
+}
+
+static inline struct xdp_umem_fq_reuse *xsk_reuseq_prepare(u32 nentries)
+{
+	return NULL;
+}
+
+static inline struct xdp_umem_fq_reuse *xsk_reuseq_swap(
+	struct xdp_umem *umem, struct xdp_umem_fq_reuse *newq)
+{
+	return NULL;
+}
+
+static inline void xsk_reuseq_free(struct xdp_umem_fq_reuse *rq)
+{
+}
+
+static inline struct xdp_umem *xdp_get_umem_from_qid(struct net_device *dev,
+						     u16 queue_id)
+{
+	return NULL;
+}
+
+static inline char *xdp_umem_get_data(struct xdp_umem *umem, u64 addr)
+{
+	return NULL;
+}
+
+static inline dma_addr_t xdp_umem_get_dma(struct xdp_umem *umem, u64 addr)
+{
+	return 0;
+}
+
+static inline bool xsk_umem_has_addrs_rq(struct xdp_umem *umem, u32 cnt)
+{
+	return false;
+}
+
+static inline u64 *xsk_umem_peek_addr_rq(struct xdp_umem *umem, u64 *addr)
+{
+	return NULL;
+}
+
+static inline void xsk_umem_release_addr_rq(struct xdp_umem *umem)
+{
+}
+
+static inline void xsk_umem_fq_reuse(struct xdp_umem *umem, u64 addr)
+{
+}
+
+static inline void xsk_set_rx_need_wakeup(struct xdp_umem *umem)
+{
+}
+
+static inline void xsk_set_tx_need_wakeup(struct xdp_umem *umem)
+{
+}
+
+static inline void xsk_clear_rx_need_wakeup(struct xdp_umem *umem)
+{
+}
+
+static inline void xsk_clear_tx_need_wakeup(struct xdp_umem *umem)
+{
+}
+
+static inline bool xsk_umem_uses_need_wakeup(struct xdp_umem *umem)
+{
+	return false;
+}
+
+static inline u64 xsk_umem_adjust_offset(struct xdp_umem *umem, u64 handle,
+					 u64 offset)
+{
+	return 0;
+}
+
+static inline u32 xsk_umem_xdp_frame_sz(struct xdp_umem *umem)
+{
+	return 0;
+}
+
+#endif /* CONFIG_XDP_SOCKETS */
+
+#endif /* _LINUX_XDP_SOCK_DRV_H */
diff --git a/net/ethtool/channels.c b/net/ethtool/channels.c
index 389924b65d05..658a8580b464 100644
--- a/net/ethtool/channels.c
+++ b/net/ethtool/channels.c
@@ -1,6 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0-only
 
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 
 #include "netlink.h"
 #include "common.h"
diff --git a/net/ethtool/ioctl.c b/net/ethtool/ioctl.c
index 52102ab1709b..74892623bacd 100644
--- a/net/ethtool/ioctl.c
+++ b/net/ethtool/ioctl.c
@@ -24,7 +24,7 @@
 #include <linux/sched/signal.h>
 #include <linux/net.h>
 #include <net/devlink.h>
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 #include <net/flow_offload.h>
 #include <linux/ethtool_netlink.h>
 #include <generated/utsrelease.h>
diff --git a/net/xdp/xdp_umem.h b/net/xdp/xdp_umem.h
index a63a9fb251f5..32067fe98f65 100644
--- a/net/xdp/xdp_umem.h
+++ b/net/xdp/xdp_umem.h
@@ -6,7 +6,7 @@
 #ifndef XDP_UMEM_H_
 #define XDP_UMEM_H_
 
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 
 int xdp_umem_assign_dev(struct xdp_umem *umem, struct net_device *dev,
 			u16 queue_id, u16 flags);
diff --git a/net/xdp/xsk.c b/net/xdp/xsk.c
index 45ffd67b367d..8bda654e82ec 100644
--- a/net/xdp/xsk.c
+++ b/net/xdp/xsk.c
@@ -22,7 +22,7 @@
 #include <linux/net.h>
 #include <linux/netdevice.h>
 #include <linux/rculist.h>
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 #include <net/xdp.h>
 
 #include "xsk_queue.h"
-- 
2.25.1

[PATCH bpf-next v4 04/15] xsk: move defines only used by AF_XDP internals to xsk.h

[Björn Töpel]

From: Björn Töpel <bjorn.topel@intel.com>

Move the XSK_NEXT_PG_CONTIG_{MASK,SHIFT}, and
XDP_UMEM_USES_NEED_WAKEUP defines from xdp_sock.h to the AF_XDP
internal xsk.h file. Also, start using the BIT{,_ULL} macro instead of
explicit shifts.

Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
 include/net/xdp_sock.h | 14 --------------
 net/xdp/xsk.h          | 14 ++++++++++++++
 net/xdp/xsk_queue.h    |  2 ++
 3 files changed, 16 insertions(+), 14 deletions(-)

diff --git a/include/net/xdp_sock.h b/include/net/xdp_sock.h
index 6a986dcbc336..fb7fe3060175 100644
--- a/include/net/xdp_sock.h
+++ b/include/net/xdp_sock.h
@@ -17,13 +17,6 @@ struct net_device;
 struct xsk_queue;
 struct xdp_buff;
 
-/* Masks for xdp_umem_page flags.
- * The low 12-bits of the addr will be 0 since this is the page address, so we
- * can use them for flags.
- */
-#define XSK_NEXT_PG_CONTIG_SHIFT 0
-#define XSK_NEXT_PG_CONTIG_MASK (1ULL << XSK_NEXT_PG_CONTIG_SHIFT)
-
 struct xdp_umem_page {
 	void *addr;
 	dma_addr_t dma;
@@ -35,13 +28,6 @@ struct xdp_umem_fq_reuse {
 	u64 handles[];
 };
 
-/* Flags for the umem flags field.
- *
- * The NEED_WAKEUP flag is 1 due to the reuse of the flags field for public
- * flags. See inlude/uapi/include/linux/if_xdp.h.
- */
-#define XDP_UMEM_USES_NEED_WAKEUP (1 << 1)
-
 struct xdp_umem {
 	struct xsk_queue *fq;
 	struct xsk_queue *cq;
diff --git a/net/xdp/xsk.h b/net/xdp/xsk.h
index d6a0979050e6..455ddd480f3d 100644
--- a/net/xdp/xsk.h
+++ b/net/xdp/xsk.h
@@ -4,6 +4,20 @@
 #ifndef XSK_H_
 #define XSK_H_
 
+/* Masks for xdp_umem_page flags.
+ * The low 12-bits of the addr will be 0 since this is the page address, so we
+ * can use them for flags.
+ */
+#define XSK_NEXT_PG_CONTIG_SHIFT 0
+#define XSK_NEXT_PG_CONTIG_MASK BIT_ULL(XSK_NEXT_PG_CONTIG_SHIFT)
+
+/* Flags for the umem flags field.
+ *
+ * The NEED_WAKEUP flag is 1 due to the reuse of the flags field for public
+ * flags. See inlude/uapi/include/linux/if_xdp.h.
+ */
+#define XDP_UMEM_USES_NEED_WAKEUP BIT(1)
+
 struct xdp_ring_offset_v1 {
 	__u64 producer;
 	__u64 consumer;
diff --git a/net/xdp/xsk_queue.h b/net/xdp/xsk_queue.h
index 648733ec24ac..a322a7dac58c 100644
--- a/net/xdp/xsk_queue.h
+++ b/net/xdp/xsk_queue.h
@@ -10,6 +10,8 @@
 #include <linux/if_xdp.h>
 #include <net/xdp_sock.h>
 
+#include "xsk.h"
+
 struct xdp_ring {
 	u32 producer ____cacheline_aligned_in_smp;
 	u32 consumer ____cacheline_aligned_in_smp;
-- 
2.25.1

[PATCH bpf-next v4 05/15] xsk: introduce AF_XDP buffer allocation API

[Björn Töpel]

From: Björn Töpel <bjorn.topel@intel.com>

In order to simplify AF_XDP zero-copy enablement for NIC driver
developers, a new AF_XDP buffer allocation API is added. The
implementation is based on a single core (single producer/consumer)
buffer pool for the AF_XDP UMEM.

A buffer is allocated using the xsk_buff_alloc() function, and
returned using xsk_buff_free(). If a buffer is disassociated with the
pool, e.g. when a buffer is passed to an AF_XDP socket, a buffer is
said to be released. Currently, the release function is only used by
the AF_XDP internals and not visible to the driver.

Drivers using this API should register the XDP memory model with the
new MEM_TYPE_XSK_BUFF_POOL type.

The API is defined in net/xdp_sock_drv.h.

The buffer type is struct xdp_buff, and follows the lifetime of
regular xdp_buffs, i.e.  the lifetime of an xdp_buff is restricted to
a NAPI context. In other words, the API is not replacing xdp_frames.

In addition to introducing the API and implementations, the AF_XDP
core is migrated to use the new APIs.

rfc->v1: Fixed build errors/warnings for m68k and riscv. (kbuild test
         robot)
         Added headroom/chunk size getter. (Maxim/Björn)

v1->v2: Swapped SoBs. (Maxim)

v2->v3: Initialize struct xdp_buff member frame_sz. (Björn)
        Add API to query the DMA address of a frame. (Maxim)
        Do DMA sync for CPU till the end of the frame to handle
        possible growth (frame_sz). (Maxim)

Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
---
 include/net/xdp.h           |   4 +-
 include/net/xdp_sock.h      |   2 +
 include/net/xdp_sock_drv.h  | 164 +++++++++++++
 include/net/xsk_buff_pool.h |  56 +++++
 include/trace/events/xdp.h  |   3 +-
 net/core/xdp.c              |  14 +-
 net/xdp/Makefile            |   1 +
 net/xdp/xdp_umem.c          |  19 +-
 net/xdp/xsk.c               | 147 +++++-------
 net/xdp/xsk_buff_pool.c     | 467 ++++++++++++++++++++++++++++++++++++
 net/xdp/xsk_diag.c          |   2 +-
 net/xdp/xsk_queue.h         |  59 +++--
 12 files changed, 819 insertions(+), 119 deletions(-)
 create mode 100644 include/net/xsk_buff_pool.h
 create mode 100644 net/xdp/xsk_buff_pool.c

diff --git a/include/net/xdp.h b/include/net/xdp.h
index 3094fccf5a88..f432134c7c00 100644
--- a/include/net/xdp.h
+++ b/include/net/xdp.h
@@ -40,6 +40,7 @@ enum xdp_mem_type {
 	MEM_TYPE_PAGE_ORDER0,     /* Orig XDP full page model */
 	MEM_TYPE_PAGE_POOL,
 	MEM_TYPE_ZERO_COPY,
+	MEM_TYPE_XSK_BUFF_POOL,
 	MEM_TYPE_MAX,
 };
 
@@ -119,7 +120,8 @@ struct xdp_frame *convert_to_xdp_frame(struct xdp_buff *xdp)
 	int metasize;
 	int headroom;
 
-	if (xdp->rxq->mem.type == MEM_TYPE_ZERO_COPY)
+	if (xdp->rxq->mem.type == MEM_TYPE_ZERO_COPY ||
+	    xdp->rxq->mem.type == MEM_TYPE_XSK_BUFF_POOL)
 		return xdp_convert_zc_to_xdp_frame(xdp);
 
 	/* Assure headroom is available for storing info */
diff --git a/include/net/xdp_sock.h b/include/net/xdp_sock.h
index fb7fe3060175..6e7265f63c04 100644
--- a/include/net/xdp_sock.h
+++ b/include/net/xdp_sock.h
@@ -31,11 +31,13 @@ struct xdp_umem_fq_reuse {
 struct xdp_umem {
 	struct xsk_queue *fq;
 	struct xsk_queue *cq;
+	struct xsk_buff_pool *pool;
 	struct xdp_umem_page *pages;
 	u64 chunk_mask;
 	u64 size;
 	u32 headroom;
 	u32 chunk_size_nohr;
+	u32 chunk_size;
 	struct user_struct *user;
 	refcount_t users;
 	struct work_struct work;
diff --git a/include/net/xdp_sock_drv.h b/include/net/xdp_sock_drv.h
index d67f2361937a..7752c8663d1b 100644
--- a/include/net/xdp_sock_drv.h
+++ b/include/net/xdp_sock_drv.h
@@ -7,6 +7,7 @@
 #define _LINUX_XDP_SOCK_DRV_H
 
 #include <net/xdp_sock.h>
+#include <net/xsk_buff_pool.h>
 
 #ifdef CONFIG_XDP_SOCKETS
 
@@ -101,6 +102,94 @@ static inline u32 xsk_umem_xdp_frame_sz(struct xdp_umem *umem)
 	return umem->chunk_size_nohr;
 }
 
+static inline u32 xsk_umem_get_headroom(struct xdp_umem *umem)
+{
+	return XDP_PACKET_HEADROOM + umem->headroom;
+}
+
+static inline u32 xsk_umem_get_chunk_size(struct xdp_umem *umem)
+{
+	return umem->chunk_size;
+}
+
+static inline u32 xsk_umem_get_rx_frame_size(struct xdp_umem *umem)
+{
+	return xsk_umem_get_chunk_size(umem) - xsk_umem_get_headroom(umem);
+}
+
+static inline void xsk_buff_set_rxq_info(struct xdp_umem *umem,
+					 struct xdp_rxq_info *rxq)
+{
+	xp_set_rxq_info(umem->pool, rxq);
+}
+
+static inline void xsk_buff_dma_unmap(struct xdp_umem *umem,
+				      unsigned long attrs)
+{
+	xp_dma_unmap(umem->pool, attrs);
+}
+
+static inline int xsk_buff_dma_map(struct xdp_umem *umem, struct device *dev,
+				   unsigned long attrs)
+{
+	return xp_dma_map(umem->pool, dev, attrs, umem->pgs, umem->npgs);
+}
+
+static inline dma_addr_t xsk_buff_xdp_get_dma(struct xdp_buff *xdp)
+{
+	struct xdp_buff_xsk *xskb = container_of(xdp, struct xdp_buff_xsk, xdp);
+
+	return xp_get_dma(xskb);
+}
+
+static inline dma_addr_t xsk_buff_xdp_get_frame_dma(struct xdp_buff *xdp)
+{
+	struct xdp_buff_xsk *xskb = container_of(xdp, struct xdp_buff_xsk, xdp);
+
+	return xp_get_frame_dma(xskb);
+}
+
+static inline struct xdp_buff *xsk_buff_alloc(struct xdp_umem *umem)
+{
+	return xp_alloc(umem->pool);
+}
+
+static inline bool xsk_buff_can_alloc(struct xdp_umem *umem, u32 count)
+{
+	return xp_can_alloc(umem->pool, count);
+}
+
+static inline void xsk_buff_free(struct xdp_buff *xdp)
+{
+	struct xdp_buff_xsk *xskb = container_of(xdp, struct xdp_buff_xsk, xdp);
+
+	xp_free(xskb);
+}
+
+static inline dma_addr_t xsk_buff_raw_get_dma(struct xdp_umem *umem, u64 addr)
+{
+	return xp_raw_get_dma(umem->pool, addr);
+}
+
+static inline void *xsk_buff_raw_get_data(struct xdp_umem *umem, u64 addr)
+{
+	return xp_raw_get_data(umem->pool, addr);
+}
+
+static inline void xsk_buff_dma_sync_for_cpu(struct xdp_buff *xdp)
+{
+	struct xdp_buff_xsk *xskb = container_of(xdp, struct xdp_buff_xsk, xdp);
+
+	xp_dma_sync_for_cpu(xskb);
+}
+
+static inline void xsk_buff_raw_dma_sync_for_device(struct xdp_umem *umem,
+						    dma_addr_t dma,
+						    size_t size)
+{
+	xp_dma_sync_for_device(umem->pool, dma, size);
+}
+
 #else
 
 static inline bool xsk_umem_has_addrs(struct xdp_umem *umem, u32 cnt)
@@ -212,6 +301,81 @@ static inline u32 xsk_umem_xdp_frame_sz(struct xdp_umem *umem)
 	return 0;
 }
 
+static inline u32 xsk_umem_get_headroom(struct xdp_umem *umem)
+{
+	return 0;
+}
+
+static inline u32 xsk_umem_get_chunk_size(struct xdp_umem *umem)
+{
+	return 0;
+}
+
+static inline u32 xsk_umem_get_rx_frame_size(struct xdp_umem *umem)
+{
+	return 0;
+}
+
+static inline void xsk_buff_set_rxq_info(struct xdp_umem *umem,
+					 struct xdp_rxq_info *rxq)
+{
+}
+
+static inline void xsk_buff_dma_unmap(struct xdp_umem *umem,
+				      unsigned long attrs)
+{
+}
+
+static inline int xsk_buff_dma_map(struct xdp_umem *umem, struct device *dev,
+				   unsigned long attrs)
+{
+	return 0;
+}
+
+static inline dma_addr_t xsk_buff_xdp_get_dma(struct xdp_buff *xdp)
+{
+	return 0;
+}
+
+static inline dma_addr_t xsk_buff_xdp_get_frame_dma(struct xdp_buff *xdp)
+{
+	return 0;
+}
+
+static inline struct xdp_buff *xsk_buff_alloc(struct xdp_umem *umem)
+{
+	return NULL;
+}
+
+static inline bool xsk_buff_can_alloc(struct xdp_umem *umem, u32 count)
+{
+	return false;
+}
+
+static inline void xsk_buff_free(struct xdp_buff *xdp)
+{
+}
+
+static inline dma_addr_t xsk_buff_raw_get_dma(struct xdp_umem *umem, u64 addr)
+{
+	return 0;
+}
+
+static inline void *xsk_buff_raw_get_data(struct xdp_umem *umem, u64 addr)
+{
+	return NULL;
+}
+
+static inline void xsk_buff_dma_sync_for_cpu(struct xdp_buff *xdp)
+{
+}
+
+static inline void xsk_buff_raw_dma_sync_for_device(struct xdp_umem *umem,
+						    dma_addr_t dma,
+						    size_t size)
+{
+}
+
 #endif /* CONFIG_XDP_SOCKETS */
 
 #endif /* _LINUX_XDP_SOCK_DRV_H */
diff --git a/include/net/xsk_buff_pool.h b/include/net/xsk_buff_pool.h
new file mode 100644
index 000000000000..9f221b36e405
--- /dev/null
+++ b/include/net/xsk_buff_pool.h
@@ -0,0 +1,56 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright(c) 2020 Intel Corporation. */
+
+#ifndef XSK_BUFF_POOL_H_
+#define XSK_BUFF_POOL_H_
+
+#include <linux/types.h>
+#include <linux/dma-mapping.h>
+#include <net/xdp.h>
+
+struct xsk_buff_pool;
+struct xdp_rxq_info;
+struct xsk_queue;
+struct xdp_desc;
+struct device;
+struct page;
+
+struct xdp_buff_xsk {
+	struct xdp_buff xdp;
+	dma_addr_t dma;
+	dma_addr_t frame_dma;
+	struct xsk_buff_pool *pool;
+	bool unaligned;
+	u64 orig_addr;
+	struct list_head free_list_node;
+};
+
+/* AF_XDP core. */
+struct xsk_buff_pool *xp_create(struct page **pages, u32 nr_pages, u32 chunks,
+				u32 chunk_size, u32 headroom, u64 size,
+				bool unaligned);
+void xp_set_fq(struct xsk_buff_pool *pool, struct xsk_queue *fq);
+void xp_destroy(struct xsk_buff_pool *pool);
+void xp_release(struct xdp_buff_xsk *xskb);
+u64 xp_get_handle(struct xdp_buff_xsk *xskb);
+bool xp_validate_desc(struct xsk_buff_pool *pool, struct xdp_desc *desc);
+
+/* AF_XDP, and XDP core. */
+void xp_free(struct xdp_buff_xsk *xskb);
+
+/* AF_XDP ZC drivers, via xdp_sock_buff.h */
+void xp_set_rxq_info(struct xsk_buff_pool *pool, struct xdp_rxq_info *rxq);
+int xp_dma_map(struct xsk_buff_pool *pool, struct device *dev,
+	       unsigned long attrs, struct page **pages, u32 nr_pages);
+void xp_dma_unmap(struct xsk_buff_pool *pool, unsigned long attrs);
+struct xdp_buff *xp_alloc(struct xsk_buff_pool *pool);
+bool xp_can_alloc(struct xsk_buff_pool *pool, u32 count);
+void *xp_raw_get_data(struct xsk_buff_pool *pool, u64 addr);
+dma_addr_t xp_raw_get_dma(struct xsk_buff_pool *pool, u64 addr);
+dma_addr_t xp_get_dma(struct xdp_buff_xsk *xskb);
+dma_addr_t xp_get_frame_dma(struct xdp_buff_xsk *xskb);
+void xp_dma_sync_for_cpu(struct xdp_buff_xsk *xskb);
+void xp_dma_sync_for_device(struct xsk_buff_pool *pool, dma_addr_t dma,
+			    size_t size);
+
+#endif /* XSK_BUFF_POOL_H_ */
diff --git a/include/trace/events/xdp.h b/include/trace/events/xdp.h
index b95d65e8c628..48547a12fa27 100644
--- a/include/trace/events/xdp.h
+++ b/include/trace/events/xdp.h
@@ -287,7 +287,8 @@ TRACE_EVENT(xdp_devmap_xmit,
 	FN(PAGE_SHARED)		\
 	FN(PAGE_ORDER0)		\
 	FN(PAGE_POOL)		\
-	FN(ZERO_COPY)
+	FN(ZERO_COPY)		\
+	FN(XSK_BUFF_POOL)
 
 #define __MEM_TYPE_TP_FN(x)	\
 	TRACE_DEFINE_ENUM(MEM_TYPE_##x);
diff --git a/net/core/xdp.c b/net/core/xdp.c
index 490b8f5fa8ee..f0ce8b195193 100644
--- a/net/core/xdp.c
+++ b/net/core/xdp.c
@@ -17,6 +17,7 @@
 #include <net/xdp.h>
 #include <net/xdp_priv.h> /* struct xdp_mem_allocator */
 #include <trace/events/xdp.h>
+#include <net/xdp_sock_drv.h>
 
 #define REG_STATE_NEW		0x0
 #define REG_STATE_REGISTERED	0x1
@@ -361,7 +362,7 @@ EXPORT_SYMBOL_GPL(xdp_rxq_info_reg_mem_model);
  * of xdp_frames/pages in those cases.
  */
 static void __xdp_return(void *data, struct xdp_mem_info *mem, bool napi_direct,
-			 unsigned long handle)
+			 unsigned long handle, struct xdp_buff *xdp)
 {
 	struct xdp_mem_allocator *xa;
 	struct page *page;
@@ -390,6 +391,11 @@ static void __xdp_return(void *data, struct xdp_mem_info *mem, bool napi_direct,
 		xa = rhashtable_lookup(mem_id_ht, &mem->id, mem_id_rht_params);
 		xa->zc_alloc->free(xa->zc_alloc, handle);
 		rcu_read_unlock();
+		break;
+	case MEM_TYPE_XSK_BUFF_POOL:
+		/* NB! Only valid from an xdp_buff! */
+		xsk_buff_free(xdp);
+		break;
 	default:
 		/* Not possible, checked in xdp_rxq_info_reg_mem_model() */
 		break;
@@ -398,19 +404,19 @@ static void __xdp_return(void *data, struct xdp_mem_info *mem, bool napi_direct,
 
 void xdp_return_frame(struct xdp_frame *xdpf)
 {
-	__xdp_return(xdpf->data, &xdpf->mem, false, 0);
+	__xdp_return(xdpf->data, &xdpf->mem, false, 0, NULL);
 }
 EXPORT_SYMBOL_GPL(xdp_return_frame);
 
 void xdp_return_frame_rx_napi(struct xdp_frame *xdpf)
 {
-	__xdp_return(xdpf->data, &xdpf->mem, true, 0);
+	__xdp_return(xdpf->data, &xdpf->mem, true, 0, NULL);
 }
 EXPORT_SYMBOL_GPL(xdp_return_frame_rx_napi);
 
 void xdp_return_buff(struct xdp_buff *xdp)
 {
-	__xdp_return(xdp->data, &xdp->rxq->mem, true, xdp->handle);
+	__xdp_return(xdp->data, &xdp->rxq->mem, true, xdp->handle, xdp);
 }
 EXPORT_SYMBOL_GPL(xdp_return_buff);
 
diff --git a/net/xdp/Makefile b/net/xdp/Makefile
index 90b5460d6166..30cdc4315f42 100644
--- a/net/xdp/Makefile
+++ b/net/xdp/Makefile
@@ -1,3 +1,4 @@
 # SPDX-License-Identifier: GPL-2.0-only
 obj-$(CONFIG_XDP_SOCKETS) += xsk.o xdp_umem.o xsk_queue.o xskmap.o
+obj-$(CONFIG_XDP_SOCKETS) += xsk_buff_pool.o
 obj-$(CONFIG_XDP_SOCKETS_DIAG) += xsk_diag.o
diff --git a/net/xdp/xdp_umem.c b/net/xdp/xdp_umem.c
index 37ace3bc0d48..7f04688045d5 100644
--- a/net/xdp/xdp_umem.c
+++ b/net/xdp/xdp_umem.c
@@ -245,7 +245,7 @@ static void xdp_umem_release(struct xdp_umem *umem)
 	}
 
 	xsk_reuseq_destroy(umem);
-
+	xp_destroy(umem->pool);
 	xdp_umem_unmap_pages(umem);
 	xdp_umem_unpin_pages(umem);
 
@@ -390,6 +390,7 @@ static int xdp_umem_reg(struct xdp_umem *umem, struct xdp_umem_reg *mr)
 	umem->size = size;
 	umem->headroom = headroom;
 	umem->chunk_size_nohr = chunk_size - headroom;
+	umem->chunk_size = chunk_size;
 	umem->npgs = size / PAGE_SIZE;
 	umem->pgs = NULL;
 	umem->user = NULL;
@@ -415,11 +416,21 @@ static int xdp_umem_reg(struct xdp_umem *umem, struct xdp_umem_reg *mr)
 	}
 
 	err = xdp_umem_map_pages(umem);
-	if (!err)
-		return 0;
+	if (err)
+		goto out_pages;
 
-	kvfree(umem->pages);
+	umem->pool = xp_create(umem->pgs, umem->npgs, chunks, chunk_size,
+			       headroom, size, unaligned_chunks);
+	if (!umem->pool) {
+		err = -ENOMEM;
+		goto out_unmap;
+	}
+	return 0;
 
+out_unmap:
+	xdp_umem_unmap_pages(umem);
+out_pages:
+	kvfree(umem->pages);
 out_pin:
 	xdp_umem_unpin_pages(umem);
 out_account:
diff --git a/net/xdp/xsk.c b/net/xdp/xsk.c
index 8bda654e82ec..6933f0d494ba 100644
--- a/net/xdp/xsk.c
+++ b/net/xdp/xsk.c
@@ -117,76 +117,67 @@ bool xsk_umem_uses_need_wakeup(struct xdp_umem *umem)
 }
 EXPORT_SYMBOL(xsk_umem_uses_need_wakeup);
 
-/* If a buffer crosses a page boundary, we need to do 2 memcpy's, one for
- * each page. This is only required in copy mode.
- */
-static void __xsk_rcv_memcpy(struct xdp_umem *umem, u64 addr, void *from_buf,
-			     u32 len, u32 metalen)
+static int __xsk_rcv_zc(struct xdp_sock *xs, struct xdp_buff *xdp, u32 len)
 {
-	void *to_buf = xdp_umem_get_data(umem, addr);
-
-	addr = xsk_umem_add_offset_to_addr(addr);
-	if (xskq_cons_crosses_non_contig_pg(umem, addr, len + metalen)) {
-		void *next_pg_addr = umem->pages[(addr >> PAGE_SHIFT) + 1].addr;
-		u64 page_start = addr & ~(PAGE_SIZE - 1);
-		u64 first_len = PAGE_SIZE - (addr - page_start);
-
-		memcpy(to_buf, from_buf, first_len);
-		memcpy(next_pg_addr, from_buf + first_len,
-		       len + metalen - first_len);
+	struct xdp_buff_xsk *xskb = container_of(xdp, struct xdp_buff_xsk, xdp);
+	u64 addr;
+	int err;
 
-		return;
+	addr = xp_get_handle(xskb);
+	err = xskq_prod_reserve_desc(xs->rx, addr, len);
+	if (err) {
+		xs->rx_dropped++;
+		return err;
 	}
 
-	memcpy(to_buf, from_buf, len + metalen);
+	xp_release(xskb);
+	return 0;
 }
 
-static int __xsk_rcv(struct xdp_sock *xs, struct xdp_buff *xdp, u32 len)
+static void xsk_copy_xdp(struct xdp_buff *to, struct xdp_buff *from, u32 len)
 {
-	u64 offset = xs->umem->headroom;
-	u64 addr, memcpy_addr;
-	void *from_buf;
+	void *from_buf, *to_buf;
 	u32 metalen;
-	int err;
-
-	if (!xskq_cons_peek_addr(xs->umem->fq, &addr, xs->umem) ||
-	    len > xs->umem->chunk_size_nohr - XDP_PACKET_HEADROOM) {
-		xs->rx_dropped++;
-		return -ENOSPC;
-	}
 
-	if (unlikely(xdp_data_meta_unsupported(xdp))) {
-		from_buf = xdp->data;
+	if (unlikely(xdp_data_meta_unsupported(from))) {
+		from_buf = from->data;
+		to_buf = to->data;
 		metalen = 0;
 	} else {
-		from_buf = xdp->data_meta;
-		metalen = xdp->data - xdp->data_meta;
+		from_buf = from->data_meta;
+		metalen = from->data - from->data_meta;
+		to_buf = to->data - metalen;
 	}
 
-	memcpy_addr = xsk_umem_adjust_offset(xs->umem, addr, offset);
-	__xsk_rcv_memcpy(xs->umem, memcpy_addr, from_buf, len, metalen);
-
-	offset += metalen;
-	addr = xsk_umem_adjust_offset(xs->umem, addr, offset);
-	err = xskq_prod_reserve_desc(xs->rx, addr, len);
-	if (!err) {
-		xskq_cons_release(xs->umem->fq);
-		xdp_return_buff(xdp);
-		return 0;
-	}
-
-	xs->rx_dropped++;
-	return err;
+	memcpy(to_buf, from_buf, len + metalen);
 }
 
-static int __xsk_rcv_zc(struct xdp_sock *xs, struct xdp_buff *xdp, u32 len)
+static int __xsk_rcv(struct xdp_sock *xs, struct xdp_buff *xdp, u32 len,
+		     bool explicit_free)
 {
-	int err = xskq_prod_reserve_desc(xs->rx, xdp->handle, len);
+	struct xdp_buff *xsk_xdp;
+	int err;
 
-	if (err)
+	if (len > xsk_umem_get_rx_frame_size(xs->umem)) {
+		xs->rx_dropped++;
+		return -ENOSPC;
+	}
+
+	xsk_xdp = xsk_buff_alloc(xs->umem);
+	if (!xsk_xdp) {
 		xs->rx_dropped++;
+		return -ENOSPC;
+	}
 
-	return err;
+	xsk_copy_xdp(xsk_xdp, xdp, len);
+	err = __xsk_rcv_zc(xs, xsk_xdp, len);
+	if (err) {
+		xsk_buff_free(xsk_xdp);
+		return err;
+	}
+	if (explicit_free)
+		xdp_return_buff(xdp);
+	return 0;
 }
 
 static bool xsk_is_bound(struct xdp_sock *xs)
@@ -199,7 +190,8 @@ static bool xsk_is_bound(struct xdp_sock *xs)
 	return false;
 }
 
-static int xsk_rcv(struct xdp_sock *xs, struct xdp_buff *xdp)
+static int xsk_rcv(struct xdp_sock *xs, struct xdp_buff *xdp,
+		   bool explicit_free)
 {
 	u32 len;
 
@@ -211,8 +203,10 @@ static int xsk_rcv(struct xdp_sock *xs, struct xdp_buff *xdp)
 
 	len = xdp->data_end - xdp->data;
 
-	return (xdp->rxq->mem.type == MEM_TYPE_ZERO_COPY) ?
-		__xsk_rcv_zc(xs, xdp, len) : __xsk_rcv(xs, xdp, len);
+	return xdp->rxq->mem.type == MEM_TYPE_ZERO_COPY ||
+		xdp->rxq->mem.type == MEM_TYPE_XSK_BUFF_POOL ?
+		__xsk_rcv_zc(xs, xdp, len) :
+		__xsk_rcv(xs, xdp, len, explicit_free);
 }
 
 static void xsk_flush(struct xdp_sock *xs)
@@ -224,46 +218,11 @@ static void xsk_flush(struct xdp_sock *xs)
 
 int xsk_generic_rcv(struct xdp_sock *xs, struct xdp_buff *xdp)
 {
-	u32 metalen = xdp->data - xdp->data_meta;
-	u32 len = xdp->data_end - xdp->data;
-	u64 offset = xs->umem->headroom;
-	void *buffer;
-	u64 addr;
 	int err;
 
 	spin_lock_bh(&xs->rx_lock);
-
-	if (xs->dev != xdp->rxq->dev || xs->queue_id != xdp->rxq->queue_index) {
-		err = -EINVAL;
-		goto out_unlock;
-	}
-
-	if (!xskq_cons_peek_addr(xs->umem->fq, &addr, xs->umem) ||
-	    len > xs->umem->chunk_size_nohr - XDP_PACKET_HEADROOM) {
-		err = -ENOSPC;
-		goto out_drop;
-	}
-
-	addr = xsk_umem_adjust_offset(xs->umem, addr, offset);
-	buffer = xdp_umem_get_data(xs->umem, addr);
-	memcpy(buffer, xdp->data_meta, len + metalen);
-
-	addr = xsk_umem_adjust_offset(xs->umem, addr, metalen);
-	err = xskq_prod_reserve_desc(xs->rx, addr, len);
-	if (err)
-		goto out_drop;
-
-	xskq_cons_release(xs->umem->fq);
-	xskq_prod_submit(xs->rx);
-
-	spin_unlock_bh(&xs->rx_lock);
-
-	xs->sk.sk_data_ready(&xs->sk);
-	return 0;
-
-out_drop:
-	xs->rx_dropped++;
-out_unlock:
+	err = xsk_rcv(xs, xdp, false);
+	xsk_flush(xs);
 	spin_unlock_bh(&xs->rx_lock);
 	return err;
 }
@@ -273,7 +232,7 @@ int __xsk_map_redirect(struct xdp_sock *xs, struct xdp_buff *xdp)
 	struct list_head *flush_list = this_cpu_ptr(&xskmap_flush_list);
 	int err;
 
-	err = xsk_rcv(xs, xdp);
+	err = xsk_rcv(xs, xdp, true);
 	if (err)
 		return err;
 
@@ -404,7 +363,7 @@ static int xsk_generic_xmit(struct sock *sk)
 
 		skb_put(skb, len);
 		addr = desc.addr;
-		buffer = xdp_umem_get_data(xs->umem, addr);
+		buffer = xsk_buff_raw_get_data(xs->umem, addr);
 		err = skb_store_bits(skb, 0, buffer, len);
 		/* This is the backpressure mechanism for the Tx path.
 		 * Reserve space in the completion queue and only proceed
@@ -860,6 +819,8 @@ static int xsk_setsockopt(struct socket *sock, int level, int optname,
 		q = (optname == XDP_UMEM_FILL_RING) ? &xs->umem->fq :
 			&xs->umem->cq;
 		err = xsk_init_queue(entries, q, true);
+		if (optname == XDP_UMEM_FILL_RING)
+			xp_set_fq(xs->umem->pool, *q);
 		mutex_unlock(&xs->mutex);
 		return err;
 	}
diff --git a/net/xdp/xsk_buff_pool.c b/net/xdp/xsk_buff_pool.c
new file mode 100644
index 000000000000..e214a5795a62
--- /dev/null
+++ b/net/xdp/xsk_buff_pool.c
@@ -0,0 +1,467 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <net/xsk_buff_pool.h>
+#include <net/xdp_sock.h>
+#include <linux/dma-direct.h>
+#include <linux/dma-noncoherent.h>
+#include <linux/swiotlb.h>
+
+#include "xsk_queue.h"
+
+struct xsk_buff_pool {
+	struct xsk_queue *fq;
+	struct list_head free_list;
+	dma_addr_t *dma_pages;
+	struct xdp_buff_xsk *heads;
+	u64 chunk_mask;
+	u64 addrs_cnt;
+	u32 free_list_cnt;
+	u32 dma_pages_cnt;
+	u32 heads_cnt;
+	u32 free_heads_cnt;
+	u32 headroom;
+	u32 chunk_size;
+	u32 frame_len;
+	bool cheap_dma;
+	bool unaligned;
+	void *addrs;
+	struct device *dev;
+	struct xdp_buff_xsk *free_heads[];
+};
+
+static void xp_addr_unmap(struct xsk_buff_pool *pool)
+{
+	vunmap(pool->addrs);
+}
+
+static int xp_addr_map(struct xsk_buff_pool *pool,
+		       struct page **pages, u32 nr_pages)
+{
+	pool->addrs = vmap(pages, nr_pages, VM_MAP, PAGE_KERNEL);
+	if (!pool->addrs)
+		return -ENOMEM;
+	return 0;
+}
+
+void xp_destroy(struct xsk_buff_pool *pool)
+{
+	if (!pool)
+		return;
+
+	xp_addr_unmap(pool);
+	kvfree(pool->heads);
+	kvfree(pool);
+}
+
+struct xsk_buff_pool *xp_create(struct page **pages, u32 nr_pages, u32 chunks,
+				u32 chunk_size, u32 headroom, u64 size,
+				bool unaligned)
+{
+	struct xsk_buff_pool *pool;
+	struct xdp_buff_xsk *xskb;
+	int err;
+	u32 i;
+
+	pool = kvzalloc(struct_size(pool, free_heads, chunks), GFP_KERNEL);
+	if (!pool)
+		goto out;
+
+	pool->heads = kvcalloc(chunks, sizeof(*pool->heads), GFP_KERNEL);
+	if (!pool->heads)
+		goto out;
+
+	pool->chunk_mask = ~((u64)chunk_size - 1);
+	pool->addrs_cnt = size;
+	pool->heads_cnt = chunks;
+	pool->free_heads_cnt = chunks;
+	pool->headroom = headroom;
+	pool->chunk_size = chunk_size;
+	pool->cheap_dma = true;
+	pool->unaligned = unaligned;
+	pool->frame_len = chunk_size - headroom - XDP_PACKET_HEADROOM;
+	INIT_LIST_HEAD(&pool->free_list);
+
+	for (i = 0; i < pool->free_heads_cnt; i++) {
+		xskb = &pool->heads[i];
+		xskb->pool = pool;
+		xskb->xdp.frame_sz = chunk_size - headroom;
+		pool->free_heads[i] = xskb;
+	}
+
+	err = xp_addr_map(pool, pages, nr_pages);
+	if (!err)
+		return pool;
+
+out:
+	xp_destroy(pool);
+	return NULL;
+}
+
+void xp_set_fq(struct xsk_buff_pool *pool, struct xsk_queue *fq)
+{
+	pool->fq = fq;
+}
+
+void xp_set_rxq_info(struct xsk_buff_pool *pool, struct xdp_rxq_info *rxq)
+{
+	u32 i;
+
+	for (i = 0; i < pool->heads_cnt; i++)
+		pool->heads[i].xdp.rxq = rxq;
+}
+EXPORT_SYMBOL(xp_set_rxq_info);
+
+void xp_dma_unmap(struct xsk_buff_pool *pool, unsigned long attrs)
+{
+	dma_addr_t *dma;
+	u32 i;
+
+	if (pool->dma_pages_cnt == 0)
+		return;
+
+	for (i = 0; i < pool->dma_pages_cnt; i++) {
+		dma = &pool->dma_pages[i];
+		if (*dma) {
+			dma_unmap_page_attrs(pool->dev, *dma, PAGE_SIZE,
+					     DMA_BIDIRECTIONAL, attrs);
+			*dma = 0;
+		}
+	}
+
+	kvfree(pool->dma_pages);
+	pool->dma_pages_cnt = 0;
+	pool->dev = NULL;
+}
+EXPORT_SYMBOL(xp_dma_unmap);
+
+static void xp_check_dma_contiguity(struct xsk_buff_pool *pool)
+{
+	u32 i;
+
+	for (i = 0; i < pool->dma_pages_cnt - 1; i++) {
+		if (pool->dma_pages[i] + PAGE_SIZE == pool->dma_pages[i + 1])
+			pool->dma_pages[i] |= XSK_NEXT_PG_CONTIG_MASK;
+		else
+			pool->dma_pages[i] &= ~XSK_NEXT_PG_CONTIG_MASK;
+	}
+}
+
+static bool __maybe_unused xp_check_swiotlb_dma(struct xsk_buff_pool *pool)
+{
+#if defined(CONFIG_SWIOTLB)
+	phys_addr_t paddr;
+	u32 i;
+
+	for (i = 0; i < pool->dma_pages_cnt; i++) {
+		paddr = dma_to_phys(pool->dev, pool->dma_pages[i]);
+		if (is_swiotlb_buffer(paddr))
+			return false;
+	}
+#endif
+	return true;
+}
+
+static bool xp_check_cheap_dma(struct xsk_buff_pool *pool)
+{
+#if defined(CONFIG_HAS_DMA)
+	const struct dma_map_ops *ops = get_dma_ops(pool->dev);
+
+	if (ops) {
+		return !ops->sync_single_for_cpu &&
+			!ops->sync_single_for_device;
+	}
+
+	if (!dma_is_direct(ops))
+		return false;
+
+	if (!xp_check_swiotlb_dma(pool))
+		return false;
+
+	if (!dev_is_dma_coherent(pool->dev)) {
+#if defined(CONFIG_ARCH_HAS_SYNC_DMA_FOR_CPU) ||		\
+	defined(CONFIG_ARCH_HAS_SYNC_DMA_FOR_CPU_ALL) ||	\
+	defined(CONFIG_ARCH_HAS_SYNC_DMA_FOR_DEVICE)
+		return false;
+#endif
+	}
+#endif
+	return true;
+}
+
+int xp_dma_map(struct xsk_buff_pool *pool, struct device *dev,
+	       unsigned long attrs, struct page **pages, u32 nr_pages)
+{
+	dma_addr_t dma;
+	u32 i;
+
+	pool->dma_pages = kvcalloc(nr_pages, sizeof(*pool->dma_pages),
+				   GFP_KERNEL);
+	if (!pool->dma_pages)
+		return -ENOMEM;
+
+	pool->dev = dev;
+	pool->dma_pages_cnt = nr_pages;
+
+	for (i = 0; i < pool->dma_pages_cnt; i++) {
+		dma = dma_map_page_attrs(dev, pages[i], 0, PAGE_SIZE,
+					 DMA_BIDIRECTIONAL, attrs);
+		if (dma_mapping_error(dev, dma)) {
+			xp_dma_unmap(pool, attrs);
+			return -ENOMEM;
+		}
+		pool->dma_pages[i] = dma;
+	}
+
+	if (pool->unaligned)
+		xp_check_dma_contiguity(pool);
+
+	pool->dev = dev;
+	pool->cheap_dma = xp_check_cheap_dma(pool);
+	return 0;
+}
+EXPORT_SYMBOL(xp_dma_map);
+
+static bool xp_desc_crosses_non_contig_pg(struct xsk_buff_pool *pool,
+					  u64 addr, u32 len)
+{
+	bool cross_pg = (addr & (PAGE_SIZE - 1)) + len > PAGE_SIZE;
+
+	if (pool->dma_pages_cnt && cross_pg) {
+		return !(pool->dma_pages[addr >> PAGE_SHIFT] &
+			 XSK_NEXT_PG_CONTIG_MASK);
+	}
+	return false;
+}
+
+static bool xp_addr_crosses_non_contig_pg(struct xsk_buff_pool *pool,
+					  u64 addr)
+{
+	return xp_desc_crosses_non_contig_pg(pool, addr, pool->chunk_size);
+}
+
+void xp_release(struct xdp_buff_xsk *xskb)
+{
+	xskb->pool->free_heads[xskb->pool->free_heads_cnt++] = xskb;
+}
+
+static u64 xp_aligned_extract_addr(struct xsk_buff_pool *pool, u64 addr)
+{
+	return addr & pool->chunk_mask;
+}
+
+static u64 xp_unaligned_extract_addr(u64 addr)
+{
+	return addr & XSK_UNALIGNED_BUF_ADDR_MASK;
+}
+
+static u64 xp_unaligned_extract_offset(u64 addr)
+{
+	return addr >> XSK_UNALIGNED_BUF_OFFSET_SHIFT;
+}
+
+static u64 xp_unaligned_add_offset_to_addr(u64 addr)
+{
+	return xp_unaligned_extract_addr(addr) +
+		xp_unaligned_extract_offset(addr);
+}
+
+static bool xp_check_unaligned(struct xsk_buff_pool *pool, u64 *addr)
+{
+	*addr = xp_unaligned_extract_addr(*addr);
+	if (*addr >= pool->addrs_cnt ||
+	    *addr + pool->chunk_size > pool->addrs_cnt ||
+	    xp_addr_crosses_non_contig_pg(pool, *addr))
+		return false;
+	return true;
+}
+
+static bool xp_check_aligned(struct xsk_buff_pool *pool, u64 *addr)
+{
+	*addr = xp_aligned_extract_addr(pool, *addr);
+	return *addr < pool->addrs_cnt;
+}
+
+static struct xdp_buff_xsk *__xp_alloc(struct xsk_buff_pool *pool)
+{
+	struct xdp_buff_xsk *xskb;
+	u64 addr;
+	bool ok;
+
+	if (pool->free_heads_cnt == 0)
+		return NULL;
+
+	xskb = pool->free_heads[--pool->free_heads_cnt];
+
+	for (;;) {
+		if (!xskq_cons_peek_addr_unchecked(pool->fq, &addr)) {
+			xp_release(xskb);
+			return NULL;
+		}
+
+		ok = pool->unaligned ? xp_check_unaligned(pool, &addr) :
+		     xp_check_aligned(pool, &addr);
+		if (!ok) {
+			pool->fq->invalid_descs++;
+			xskq_cons_release(pool->fq);
+			continue;
+		}
+		break;
+	}
+	xskq_cons_release(pool->fq);
+
+	xskb->orig_addr = addr;
+	xskb->xdp.data_hard_start = pool->addrs + addr + pool->headroom;
+	if (pool->dma_pages_cnt) {
+		xskb->frame_dma = (pool->dma_pages[addr >> PAGE_SHIFT] &
+				   ~XSK_NEXT_PG_CONTIG_MASK) +
+				  (addr & ~PAGE_MASK);
+		xskb->dma = xskb->frame_dma + pool->headroom +
+			    XDP_PACKET_HEADROOM;
+	}
+	return xskb;
+}
+
+struct xdp_buff *xp_alloc(struct xsk_buff_pool *pool)
+{
+	struct xdp_buff_xsk *xskb;
+
+	if (!pool->free_list_cnt) {
+		xskb = __xp_alloc(pool);
+		if (!xskb)
+			return NULL;
+	} else {
+		pool->free_list_cnt--;
+		xskb = list_first_entry(&pool->free_list, struct xdp_buff_xsk,
+					free_list_node);
+		list_del(&xskb->free_list_node);
+	}
+
+	xskb->xdp.data = xskb->xdp.data_hard_start + XDP_PACKET_HEADROOM;
+	xskb->xdp.data_meta = xskb->xdp.data;
+
+	if (!pool->cheap_dma) {
+		dma_sync_single_range_for_device(pool->dev, xskb->dma, 0,
+						 pool->frame_len,
+						 DMA_BIDIRECTIONAL);
+	}
+	return &xskb->xdp;
+}
+EXPORT_SYMBOL(xp_alloc);
+
+bool xp_can_alloc(struct xsk_buff_pool *pool, u32 count)
+{
+	if (pool->free_list_cnt >= count)
+		return true;
+	return xskq_cons_has_entries(pool->fq, count - pool->free_list_cnt);
+}
+EXPORT_SYMBOL(xp_can_alloc);
+
+void xp_free(struct xdp_buff_xsk *xskb)
+{
+	xskb->pool->free_list_cnt++;
+	list_add(&xskb->free_list_node, &xskb->pool->free_list);
+}
+EXPORT_SYMBOL(xp_free);
+
+static bool xp_aligned_validate_desc(struct xsk_buff_pool *pool,
+				     struct xdp_desc *desc)
+{
+	u64 chunk, chunk_end;
+
+	chunk = xp_aligned_extract_addr(pool, desc->addr);
+	chunk_end = xp_aligned_extract_addr(pool, desc->addr + desc->len);
+	if (chunk != chunk_end)
+		return false;
+
+	if (chunk >= pool->addrs_cnt)
+		return false;
+
+	if (desc->options)
+		return false;
+	return true;
+}
+
+static bool xp_unaligned_validate_desc(struct xsk_buff_pool *pool,
+				       struct xdp_desc *desc)
+{
+	u64 addr, base_addr;
+
+	base_addr = xp_unaligned_extract_addr(desc->addr);
+	addr = xp_unaligned_add_offset_to_addr(desc->addr);
+
+	if (desc->len > pool->chunk_size)
+		return false;
+
+	if (base_addr >= pool->addrs_cnt || addr >= pool->addrs_cnt ||
+	    xp_desc_crosses_non_contig_pg(pool, addr, desc->len))
+		return false;
+
+	if (desc->options)
+		return false;
+	return true;
+}
+
+bool xp_validate_desc(struct xsk_buff_pool *pool, struct xdp_desc *desc)
+{
+	return pool->unaligned ? xp_unaligned_validate_desc(pool, desc) :
+		xp_aligned_validate_desc(pool, desc);
+}
+
+u64 xp_get_handle(struct xdp_buff_xsk *xskb)
+{
+	u64 offset = xskb->xdp.data - xskb->xdp.data_hard_start;
+
+	offset += xskb->pool->headroom;
+	if (!xskb->pool->unaligned)
+		return xskb->orig_addr + offset;
+	return xskb->orig_addr + (offset << XSK_UNALIGNED_BUF_OFFSET_SHIFT);
+}
+
+void *xp_raw_get_data(struct xsk_buff_pool *pool, u64 addr)
+{
+	addr = pool->unaligned ? xp_unaligned_add_offset_to_addr(addr) : addr;
+	return pool->addrs + addr;
+}
+EXPORT_SYMBOL(xp_raw_get_data);
+
+dma_addr_t xp_raw_get_dma(struct xsk_buff_pool *pool, u64 addr)
+{
+	addr = pool->unaligned ? xp_unaligned_add_offset_to_addr(addr) : addr;
+	return (pool->dma_pages[addr >> PAGE_SHIFT] &
+		~XSK_NEXT_PG_CONTIG_MASK) +
+		(addr & ~PAGE_MASK);
+}
+EXPORT_SYMBOL(xp_raw_get_dma);
+
+dma_addr_t xp_get_dma(struct xdp_buff_xsk *xskb)
+{
+	return xskb->dma;
+}
+EXPORT_SYMBOL(xp_get_dma);
+
+dma_addr_t xp_get_frame_dma(struct xdp_buff_xsk *xskb)
+{
+	return xskb->frame_dma;
+}
+EXPORT_SYMBOL(xp_get_frame_dma);
+
+void xp_dma_sync_for_cpu(struct xdp_buff_xsk *xskb)
+{
+	if (xskb->pool->cheap_dma)
+		return;
+
+	dma_sync_single_range_for_cpu(xskb->pool->dev, xskb->dma, 0,
+				      xskb->pool->frame_len, DMA_BIDIRECTIONAL);
+}
+EXPORT_SYMBOL(xp_dma_sync_for_cpu);
+
+void xp_dma_sync_for_device(struct xsk_buff_pool *pool, dma_addr_t dma,
+			    size_t size)
+{
+	if (pool->cheap_dma)
+		return;
+
+	dma_sync_single_range_for_device(pool->dev, dma, 0,
+					 size, DMA_BIDIRECTIONAL);
+}
+EXPORT_SYMBOL(xp_dma_sync_for_device);
diff --git a/net/xdp/xsk_diag.c b/net/xdp/xsk_diag.c
index f59791ba43a0..0163b26aaf63 100644
--- a/net/xdp/xsk_diag.c
+++ b/net/xdp/xsk_diag.c
@@ -56,7 +56,7 @@ static int xsk_diag_put_umem(const struct xdp_sock *xs, struct sk_buff *nlskb)
 	du.id = umem->id;
 	du.size = umem->size;
 	du.num_pages = umem->npgs;
-	du.chunk_size = umem->chunk_size_nohr + umem->headroom;
+	du.chunk_size = umem->chunk_size;
 	du.headroom = umem->headroom;
 	du.ifindex = umem->dev ? umem->dev->ifindex : 0;
 	du.queue_id = umem->queue_id;
diff --git a/net/xdp/xsk_queue.h b/net/xdp/xsk_queue.h
index a322a7dac58c..9151aef7dbca 100644
--- a/net/xdp/xsk_queue.h
+++ b/net/xdp/xsk_queue.h
@@ -9,6 +9,7 @@
 #include <linux/types.h>
 #include <linux/if_xdp.h>
 #include <net/xdp_sock.h>
+#include <net/xsk_buff_pool.h>
 
 #include "xsk.h"
 
@@ -172,31 +173,45 @@ static inline bool xskq_cons_read_addr(struct xsk_queue *q, u64 *addr,
 	return false;
 }
 
-static inline bool xskq_cons_is_valid_desc(struct xsk_queue *q,
-					   struct xdp_desc *d,
-					   struct xdp_umem *umem)
+static inline bool xskq_cons_read_addr_aligned(struct xsk_queue *q, u64 *addr)
 {
-	if (umem->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG) {
-		if (!xskq_cons_is_valid_unaligned(q, d->addr, d->len, umem))
-			return false;
+	struct xdp_umem_ring *ring = (struct xdp_umem_ring *)q->ring;
 
-		if (d->len > umem->chunk_size_nohr || d->options) {
-			q->invalid_descs++;
-			return false;
-		}
+	while (q->cached_cons != q->cached_prod) {
+		u32 idx = q->cached_cons & q->ring_mask;
+
+		*addr = ring->desc[idx];
+		if (xskq_cons_is_valid_addr(q, *addr))
+			return true;
 
+		q->cached_cons++;
+	}
+
+	return false;
+}
+
+static inline bool xskq_cons_read_addr_unchecked(struct xsk_queue *q, u64 *addr)
+{
+	struct xdp_umem_ring *ring = (struct xdp_umem_ring *)q->ring;
+
+	if (q->cached_cons != q->cached_prod) {
+		u32 idx = q->cached_cons & q->ring_mask;
+
+		*addr = ring->desc[idx];
 		return true;
 	}
 
-	if (!xskq_cons_is_valid_addr(q, d->addr))
-		return false;
+	return false;
+}
 
-	if (((d->addr + d->len) & q->chunk_mask) != (d->addr & q->chunk_mask) ||
-	    d->options) {
+static inline bool xskq_cons_is_valid_desc(struct xsk_queue *q,
+					   struct xdp_desc *d,
+					   struct xdp_umem *umem)
+{
+	if (!xp_validate_desc(umem->pool, d)) {
 		q->invalid_descs++;
 		return false;
 	}
-
 	return true;
 }
 
@@ -260,6 +275,20 @@ static inline bool xskq_cons_peek_addr(struct xsk_queue *q, u64 *addr,
 	return xskq_cons_read_addr(q, addr, umem);
 }
 
+static inline bool xskq_cons_peek_addr_aligned(struct xsk_queue *q, u64 *addr)
+{
+	if (q->cached_prod == q->cached_cons)
+		xskq_cons_get_entries(q);
+	return xskq_cons_read_addr_aligned(q, addr);
+}
+
+static inline bool xskq_cons_peek_addr_unchecked(struct xsk_queue *q, u64 *addr)
+{
+	if (q->cached_prod == q->cached_cons)
+		xskq_cons_get_entries(q);
+	return xskq_cons_read_addr_unchecked(q, addr);
+}
+
 static inline bool xskq_cons_peek_desc(struct xsk_queue *q,
 				       struct xdp_desc *desc,
 				       struct xdp_umem *umem)
-- 
2.25.1

[PATCH bpf-next v4 06/15] i40e: refactor rx_bi accesses

[Björn Töpel]

From: Björn Töpel <bjorn.topel@intel.com>

As a first step to migrate i40e to the new MEM_TYPE_XSK_BUFF_POOL
APIs, code that accesses the rx_bi (SW/shadow ring) is refactored to
use an accessor function.

Cc: intel-wired-lan@lists.osuosl.org
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_txrx.c | 17 +++++++++++------
 drivers/net/ethernet/intel/i40e/i40e_xsk.c  | 18 ++++++++++++------
 2 files changed, 23 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
index a3772beffe02..9b9ef951f9ce 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
@@ -1195,6 +1195,11 @@ static void i40e_update_itr(struct i40e_q_vector *q_vector,
 	rc->total_packets = 0;
 }
 
+static struct i40e_rx_buffer *i40e_rx_bi(struct i40e_ring *rx_ring, u32 idx)
+{
+	return &rx_ring->rx_bi[idx];
+}
+
 /**
  * i40e_reuse_rx_page - page flip buffer and store it back on the ring
  * @rx_ring: rx descriptor ring to store buffers on
@@ -1208,7 +1213,7 @@ static void i40e_reuse_rx_page(struct i40e_ring *rx_ring,
 	struct i40e_rx_buffer *new_buff;
 	u16 nta = rx_ring->next_to_alloc;
 
-	new_buff = &rx_ring->rx_bi[nta];
+	new_buff = i40e_rx_bi(rx_ring, nta);
 
 	/* update, and store next to alloc */
 	nta++;
@@ -1272,7 +1277,7 @@ struct i40e_rx_buffer *i40e_clean_programming_status(
 	ntc = rx_ring->next_to_clean;
 
 	/* fetch, update, and store next to clean */
-	rx_buffer = &rx_ring->rx_bi[ntc++];
+	rx_buffer = i40e_rx_bi(rx_ring, ntc++);
 	ntc = (ntc < rx_ring->count) ? ntc : 0;
 	rx_ring->next_to_clean = ntc;
 
@@ -1361,7 +1366,7 @@ void i40e_clean_rx_ring(struct i40e_ring *rx_ring)
 
 	/* Free all the Rx ring sk_buffs */
 	for (i = 0; i < rx_ring->count; i++) {
-		struct i40e_rx_buffer *rx_bi = &rx_ring->rx_bi[i];
+		struct i40e_rx_buffer *rx_bi = i40e_rx_bi(rx_ring, i);
 
 		if (!rx_bi->page)
 			continue;
@@ -1592,7 +1597,7 @@ bool i40e_alloc_rx_buffers(struct i40e_ring *rx_ring, u16 cleaned_count)
 		return false;
 
 	rx_desc = I40E_RX_DESC(rx_ring, ntu);
-	bi = &rx_ring->rx_bi[ntu];
+	bi = i40e_rx_bi(rx_ring, ntu);
 
 	do {
 		if (!i40e_alloc_mapped_page(rx_ring, bi))
@@ -1614,7 +1619,7 @@ bool i40e_alloc_rx_buffers(struct i40e_ring *rx_ring, u16 cleaned_count)
 		ntu++;
 		if (unlikely(ntu == rx_ring->count)) {
 			rx_desc = I40E_RX_DESC(rx_ring, 0);
-			bi = rx_ring->rx_bi;
+			bi = i40e_rx_bi(rx_ring, 0);
 			ntu = 0;
 		}
 
@@ -1981,7 +1986,7 @@ static struct i40e_rx_buffer *i40e_get_rx_buffer(struct i40e_ring *rx_ring,
 {
 	struct i40e_rx_buffer *rx_buffer;
 
-	rx_buffer = &rx_ring->rx_bi[rx_ring->next_to_clean];
+	rx_buffer = i40e_rx_bi(rx_ring, rx_ring->next_to_clean);
 	prefetchw(rx_buffer->page);
 
 	/* we are reusing so sync this buffer for CPU use */
diff --git a/drivers/net/ethernet/intel/i40e/i40e_xsk.c b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
index d8b0be29099a..d84ec92f8538 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_xsk.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
@@ -9,6 +9,11 @@
 #include "i40e_txrx_common.h"
 #include "i40e_xsk.h"
 
+static struct i40e_rx_buffer *i40e_rx_bi(struct i40e_ring *rx_ring, u32 idx)
+{
+	return &rx_ring->rx_bi[idx];
+}
+
 /**
  * i40e_xsk_umem_dma_map - DMA maps all UMEM memory for the netdev
  * @vsi: Current VSI
@@ -321,7 +326,7 @@ __i40e_alloc_rx_buffers_zc(struct i40e_ring *rx_ring, u16 count,
 	bool ok = true;
 
 	rx_desc = I40E_RX_DESC(rx_ring, ntu);
-	bi = &rx_ring->rx_bi[ntu];
+	bi = i40e_rx_bi(rx_ring, ntu);
 	do {
 		if (!alloc(rx_ring, bi)) {
 			ok = false;
@@ -340,7 +345,7 @@ __i40e_alloc_rx_buffers_zc(struct i40e_ring *rx_ring, u16 count,
 
 		if (unlikely(ntu == rx_ring->count)) {
 			rx_desc = I40E_RX_DESC(rx_ring, 0);
-			bi = rx_ring->rx_bi;
+			bi = i40e_rx_bi(rx_ring, 0);
 			ntu = 0;
 		}
 
@@ -402,7 +407,7 @@ static struct i40e_rx_buffer *i40e_get_rx_buffer_zc(struct i40e_ring *rx_ring,
 {
 	struct i40e_rx_buffer *bi;
 
-	bi = &rx_ring->rx_bi[rx_ring->next_to_clean];
+	bi = i40e_rx_bi(rx_ring, rx_ring->next_to_clean);
 
 	/* we are reusing so sync this buffer for CPU use */
 	dma_sync_single_range_for_cpu(rx_ring->dev,
@@ -424,7 +429,8 @@ static struct i40e_rx_buffer *i40e_get_rx_buffer_zc(struct i40e_ring *rx_ring,
 static void i40e_reuse_rx_buffer_zc(struct i40e_ring *rx_ring,
 				    struct i40e_rx_buffer *old_bi)
 {
-	struct i40e_rx_buffer *new_bi = &rx_ring->rx_bi[rx_ring->next_to_alloc];
+	struct i40e_rx_buffer *new_bi = i40e_rx_bi(rx_ring,
+						   rx_ring->next_to_alloc);
 	u16 nta = rx_ring->next_to_alloc;
 
 	/* update, and store next to alloc */
@@ -456,7 +462,7 @@ void i40e_zca_free(struct zero_copy_allocator *alloc, unsigned long handle)
 	mask = rx_ring->xsk_umem->chunk_mask;
 
 	nta = rx_ring->next_to_alloc;
-	bi = &rx_ring->rx_bi[nta];
+	bi = i40e_rx_bi(rx_ring, nta);
 
 	nta++;
 	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
@@ -826,7 +832,7 @@ void i40e_xsk_clean_rx_ring(struct i40e_ring *rx_ring)
 	u16 i;
 
 	for (i = 0; i < rx_ring->count; i++) {
-		struct i40e_rx_buffer *rx_bi = &rx_ring->rx_bi[i];
+		struct i40e_rx_buffer *rx_bi = i40e_rx_bi(rx_ring, i);
 
 		if (!rx_bi->addr)
 			continue;
-- 
2.25.1

[PATCH bpf-next v4 07/15] i40e: separate kernel allocated rx_bi rings from AF_XDP rings

[Björn Töpel]

From: Björn Töpel <bjorn.topel@intel.com>

Continuing the path to support MEM_TYPE_XSK_BUFF_POOL, the AF_XDP
zero-copy/sk_buff rx_bi rings are now separate. Functions to properly
allocate the different rings are added as well.

v3->v4: Made i40e_fd_handle_status() static. (kbuild test robot)

Cc: intel-wired-lan@lists.osuosl.org
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_main.c   |   7 ++
 drivers/net/ethernet/intel/i40e/i40e_txrx.c   | 119 +++++++-----------
 drivers/net/ethernet/intel/i40e/i40e_txrx.h   |  22 ++--
 .../ethernet/intel/i40e/i40e_txrx_common.h    |  40 +++++-
 drivers/net/ethernet/intel/i40e/i40e_type.h   |   5 +-
 drivers/net/ethernet/intel/i40e/i40e_xsk.c    |  74 ++++++-----
 drivers/net/ethernet/intel/i40e/i40e_xsk.h    |   2 +
 7 files changed, 142 insertions(+), 127 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index d6b2db4f2c65..3e1695bb8262 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -3260,8 +3260,12 @@ static int i40e_configure_rx_ring(struct i40e_ring *ring)
 	if (ring->vsi->type == I40E_VSI_MAIN)
 		xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
 
+	kfree(ring->rx_bi);
 	ring->xsk_umem = i40e_xsk_umem(ring);
 	if (ring->xsk_umem) {
+		ret = i40e_alloc_rx_bi_zc(ring);
+		if (ret)
+			return ret;
 		ring->rx_buf_len = ring->xsk_umem->chunk_size_nohr -
 				   XDP_PACKET_HEADROOM;
 		/* For AF_XDP ZC, we disallow packets to span on
@@ -3280,6 +3284,9 @@ static int i40e_configure_rx_ring(struct i40e_ring *ring)
 			 ring->queue_index);
 
 	} else {
+		ret = i40e_alloc_rx_bi(ring);
+		if (ret)
+			return ret;
 		ring->rx_buf_len = vsi->rx_buf_len;
 		if (ring->vsi->type == I40E_VSI_MAIN) {
 			ret = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
index 9b9ef951f9ce..e1439a6c8062 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
@@ -521,28 +521,29 @@ int i40e_add_del_fdir(struct i40e_vsi *vsi,
 /**
  * i40e_fd_handle_status - check the Programming Status for FD
  * @rx_ring: the Rx ring for this descriptor
- * @rx_desc: the Rx descriptor for programming Status, not a packet descriptor.
+ * @qword0_raw: qword0
+ * @qword1: qword1 after le_to_cpu
  * @prog_id: the id originally used for programming
  *
  * This is used to verify if the FD programming or invalidation
  * requested by SW to the HW is successful or not and take actions accordingly.
  **/
-void i40e_fd_handle_status(struct i40e_ring *rx_ring,
-			   union i40e_rx_desc *rx_desc, u8 prog_id)
+static void i40e_fd_handle_status(struct i40e_ring *rx_ring, u64 qword0_raw,
+				  u64 qword1, u8 prog_id)
 {
 	struct i40e_pf *pf = rx_ring->vsi->back;
 	struct pci_dev *pdev = pf->pdev;
+	struct i40e_32b_rx_wb_qw0 *qw0;
 	u32 fcnt_prog, fcnt_avail;
 	u32 error;
-	u64 qw;
 
-	qw = le64_to_cpu(rx_desc->wb.qword1.status_error_len);
-	error = (qw & I40E_RX_PROG_STATUS_DESC_QW1_ERROR_MASK) >>
+	qw0 = (struct i40e_32b_rx_wb_qw0 *)&qword0_raw;
+	error = (qword1 & I40E_RX_PROG_STATUS_DESC_QW1_ERROR_MASK) >>
 		I40E_RX_PROG_STATUS_DESC_QW1_ERROR_SHIFT;
 
 	if (error == BIT(I40E_RX_PROG_STATUS_DESC_FD_TBL_FULL_SHIFT)) {
-		pf->fd_inv = le32_to_cpu(rx_desc->wb.qword0.hi_dword.fd_id);
-		if ((rx_desc->wb.qword0.hi_dword.fd_id != 0) ||
+		pf->fd_inv = le32_to_cpu(qw0->hi_dword.fd_id);
+		if (qw0->hi_dword.fd_id != 0 ||
 		    (I40E_DEBUG_FD & pf->hw.debug_mask))
 			dev_warn(&pdev->dev, "ntuple filter loc = %d, could not be added\n",
 				 pf->fd_inv);
@@ -560,7 +561,7 @@ void i40e_fd_handle_status(struct i40e_ring *rx_ring,
 		/* store the current atr filter count */
 		pf->fd_atr_cnt = i40e_get_current_atr_cnt(pf);
 
-		if ((rx_desc->wb.qword0.hi_dword.fd_id == 0) &&
+		if (qw0->hi_dword.fd_id == 0 &&
 		    test_bit(__I40E_FD_SB_AUTO_DISABLED, pf->state)) {
 			/* These set_bit() calls aren't atomic with the
 			 * test_bit() here, but worse case we potentially
@@ -589,7 +590,7 @@ void i40e_fd_handle_status(struct i40e_ring *rx_ring,
 	} else if (error == BIT(I40E_RX_PROG_STATUS_DESC_NO_FD_ENTRY_SHIFT)) {
 		if (I40E_DEBUG_FD & pf->hw.debug_mask)
 			dev_info(&pdev->dev, "ntuple filter fd_id = %d, could not be removed\n",
-				 rx_desc->wb.qword0.hi_dword.fd_id);
+				 qw0->hi_dword.fd_id);
 	}
 }
 
@@ -1232,29 +1233,10 @@ static void i40e_reuse_rx_page(struct i40e_ring *rx_ring,
 }
 
 /**
- * i40e_rx_is_programming_status - check for programming status descriptor
- * @qw: qword representing status_error_len in CPU ordering
- *
- * The value of in the descriptor length field indicate if this
- * is a programming status descriptor for flow director or FCoE
- * by the value of I40E_RX_PROG_STATUS_DESC_LENGTH, otherwise
- * it is a packet descriptor.
- **/
-static inline bool i40e_rx_is_programming_status(u64 qw)
-{
-	/* The Rx filter programming status and SPH bit occupy the same
-	 * spot in the descriptor. Since we don't support packet split we
-	 * can just reuse the bit as an indication that this is a
-	 * programming status descriptor.
-	 */
-	return qw & I40E_RXD_QW1_LENGTH_SPH_MASK;
-}
-
-/**
- * i40e_clean_programming_status - try clean the programming status descriptor
+ * i40e_clean_programming_status - clean the programming status descriptor
  * @rx_ring: the rx ring that has this descriptor
- * @rx_desc: the rx descriptor written back by HW
- * @qw: qword representing status_error_len in CPU ordering
+ * @qword0: qword0
+ * @qword1: qword1 representing status_error_len in CPU ordering
  *
  * Flow director should handle FD_FILTER_STATUS to check its filter programming
  * status being successful or not and take actions accordingly. FCoE should
@@ -1262,34 +1244,16 @@ static inline bool i40e_rx_is_programming_status(u64 qw)
  *
  * Returns an i40e_rx_buffer to reuse if the cleanup occurred, otherwise NULL.
  **/
-struct i40e_rx_buffer *i40e_clean_programming_status(
-	struct i40e_ring *rx_ring,
-	union i40e_rx_desc *rx_desc,
-	u64 qw)
+void i40e_clean_programming_status(struct i40e_ring *rx_ring, u64 qword0_raw,
+				   u64 qword1)
 {
-	struct i40e_rx_buffer *rx_buffer;
-	u32 ntc;
 	u8 id;
 
-	if (!i40e_rx_is_programming_status(qw))
-		return NULL;
-
-	ntc = rx_ring->next_to_clean;
-
-	/* fetch, update, and store next to clean */
-	rx_buffer = i40e_rx_bi(rx_ring, ntc++);
-	ntc = (ntc < rx_ring->count) ? ntc : 0;
-	rx_ring->next_to_clean = ntc;
-
-	prefetch(I40E_RX_DESC(rx_ring, ntc));
-
-	id = (qw & I40E_RX_PROG_STATUS_DESC_QW1_PROGID_MASK) >>
+	id = (qword1 & I40E_RX_PROG_STATUS_DESC_QW1_PROGID_MASK) >>
 		  I40E_RX_PROG_STATUS_DESC_QW1_PROGID_SHIFT;
 
 	if (id == I40E_RX_PROG_STATUS_DESC_FD_FILTER_STATUS)
-		i40e_fd_handle_status(rx_ring, rx_desc, id);
-
-	return rx_buffer;
+		i40e_fd_handle_status(rx_ring, qword0_raw, qword1, id);
 }
 
 /**
@@ -1341,13 +1305,25 @@ int i40e_setup_tx_descriptors(struct i40e_ring *tx_ring)
 	return -ENOMEM;
 }
 
+int i40e_alloc_rx_bi(struct i40e_ring *rx_ring)
+{
+	unsigned long sz = sizeof(*rx_ring->rx_bi) * rx_ring->count;
+
+	rx_ring->rx_bi = kzalloc(sz, GFP_KERNEL);
+	return rx_ring->rx_bi ? 0 : -ENOMEM;
+}
+
+static void i40e_clear_rx_bi(struct i40e_ring *rx_ring)
+{
+	memset(rx_ring->rx_bi, 0, sizeof(*rx_ring->rx_bi) * rx_ring->count);
+}
+
 /**
  * i40e_clean_rx_ring - Free Rx buffers
  * @rx_ring: ring to be cleaned
  **/
 void i40e_clean_rx_ring(struct i40e_ring *rx_ring)
 {
-	unsigned long bi_size;
 	u16 i;
 
 	/* ring already cleared, nothing to do */
@@ -1393,8 +1369,10 @@ void i40e_clean_rx_ring(struct i40e_ring *rx_ring)
 	}
 
 skip_free:
-	bi_size = sizeof(struct i40e_rx_buffer) * rx_ring->count;
-	memset(rx_ring->rx_bi, 0, bi_size);
+	if (rx_ring->xsk_umem)
+		i40e_clear_rx_bi_zc(rx_ring);
+	else
+		i40e_clear_rx_bi(rx_ring);
 
 	/* Zero out the descriptor ring */
 	memset(rx_ring->desc, 0, rx_ring->size);
@@ -1435,15 +1413,7 @@ void i40e_free_rx_resources(struct i40e_ring *rx_ring)
 int i40e_setup_rx_descriptors(struct i40e_ring *rx_ring)
 {
 	struct device *dev = rx_ring->dev;
-	int err = -ENOMEM;
-	int bi_size;
-
-	/* warn if we are about to overwrite the pointer */
-	WARN_ON(rx_ring->rx_bi);
-	bi_size = sizeof(struct i40e_rx_buffer) * rx_ring->count;
-	rx_ring->rx_bi = kzalloc(bi_size, GFP_KERNEL);
-	if (!rx_ring->rx_bi)
-		goto err;
+	int err;
 
 	u64_stats_init(&rx_ring->syncp);
 
@@ -1456,7 +1426,7 @@ int i40e_setup_rx_descriptors(struct i40e_ring *rx_ring)
 	if (!rx_ring->desc) {
 		dev_info(dev, "Unable to allocate memory for the Rx descriptor ring, size=%d\n",
 			 rx_ring->size);
-		goto err;
+		return -ENOMEM;
 	}
 
 	rx_ring->next_to_alloc = 0;
@@ -1468,16 +1438,12 @@ int i40e_setup_rx_descriptors(struct i40e_ring *rx_ring)
 		err = xdp_rxq_info_reg(&rx_ring->xdp_rxq, rx_ring->netdev,
 				       rx_ring->queue_index);
 		if (err < 0)
-			goto err;
+			return err;
 	}
 
 	rx_ring->xdp_prog = rx_ring->vsi->xdp_prog;
 
 	return 0;
-err:
-	kfree(rx_ring->rx_bi);
-	rx_ring->rx_bi = NULL;
-	return err;
 }
 
 /**
@@ -2387,9 +2353,12 @@ static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget)
 		 */
 		dma_rmb();
 
-		rx_buffer = i40e_clean_programming_status(rx_ring, rx_desc,
-							  qword);
-		if (unlikely(rx_buffer)) {
+		if (i40e_rx_is_programming_status(qword)) {
+			i40e_clean_programming_status(rx_ring,
+						      rx_desc->raw.qword[0],
+						      qword);
+			rx_buffer = i40e_rx_bi(rx_ring, rx_ring->next_to_clean);
+			i40e_inc_ntc(rx_ring);
 			i40e_reuse_rx_page(rx_ring, rx_buffer);
 			cleaned_count++;
 			continue;
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.h b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
index 36d37f31a287..d343498e8de5 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
@@ -296,17 +296,15 @@ struct i40e_tx_buffer {
 
 struct i40e_rx_buffer {
 	dma_addr_t dma;
-	union {
-		struct {
-			struct page *page;
-			__u32 page_offset;
-			__u16 pagecnt_bias;
-		};
-		struct {
-			void *addr;
-			u64 handle;
-		};
-	};
+	struct page *page;
+	__u32 page_offset;
+	__u16 pagecnt_bias;
+};
+
+struct i40e_rx_buffer_zc {
+	dma_addr_t dma;
+	void *addr;
+	u64 handle;
 };
 
 struct i40e_queue_stats {
@@ -358,6 +356,7 @@ struct i40e_ring {
 	union {
 		struct i40e_tx_buffer *tx_bi;
 		struct i40e_rx_buffer *rx_bi;
+		struct i40e_rx_buffer_zc *rx_bi_zc;
 	};
 	DECLARE_BITMAP(state, __I40E_RING_STATE_NBITS);
 	u16 queue_index;		/* Queue number of ring */
@@ -495,6 +494,7 @@ int __i40e_maybe_stop_tx(struct i40e_ring *tx_ring, int size);
 bool __i40e_chk_linearize(struct sk_buff *skb);
 int i40e_xdp_xmit(struct net_device *dev, int n, struct xdp_frame **frames,
 		  u32 flags);
+int i40e_alloc_rx_bi(struct i40e_ring *rx_ring);
 
 /**
  * i40e_get_head - Retrieve head from head writeback
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx_common.h b/drivers/net/ethernet/intel/i40e/i40e_txrx_common.h
index 8af0e99c6c0d..667c4dc4b39f 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx_common.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx_common.h
@@ -4,13 +4,9 @@
 #ifndef I40E_TXRX_COMMON_
 #define I40E_TXRX_COMMON_
 
-void i40e_fd_handle_status(struct i40e_ring *rx_ring,
-			   union i40e_rx_desc *rx_desc, u8 prog_id);
 int i40e_xmit_xdp_tx_ring(struct xdp_buff *xdp, struct i40e_ring *xdp_ring);
-struct i40e_rx_buffer *i40e_clean_programming_status(
-	struct i40e_ring *rx_ring,
-	union i40e_rx_desc *rx_desc,
-	u64 qw);
+void i40e_clean_programming_status(struct i40e_ring *rx_ring, u64 qword0_raw,
+				   u64 qword1);
 void i40e_process_skb_fields(struct i40e_ring *rx_ring,
 			     union i40e_rx_desc *rx_desc, struct sk_buff *skb);
 void i40e_xdp_ring_update_tail(struct i40e_ring *xdp_ring);
@@ -84,6 +80,38 @@ static inline void i40e_arm_wb(struct i40e_ring *tx_ring,
 	}
 }
 
+/**
+ * i40e_rx_is_programming_status - check for programming status descriptor
+ * @qword1: qword1 representing status_error_len in CPU ordering
+ *
+ * The value of in the descriptor length field indicate if this
+ * is a programming status descriptor for flow director or FCoE
+ * by the value of I40E_RX_PROG_STATUS_DESC_LENGTH, otherwise
+ * it is a packet descriptor.
+ **/
+static inline bool i40e_rx_is_programming_status(u64 qword1)
+{
+	/* The Rx filter programming status and SPH bit occupy the same
+	 * spot in the descriptor. Since we don't support packet split we
+	 * can just reuse the bit as an indication that this is a
+	 * programming status descriptor.
+	 */
+	return qword1 & I40E_RXD_QW1_LENGTH_SPH_MASK;
+}
+
+/**
+ * i40e_inc_ntc: Advance the next_to_clean index
+ * @rx_ring: Rx ring
+ **/
+static inline void i40e_inc_ntc(struct i40e_ring *rx_ring)
+{
+	u32 ntc = rx_ring->next_to_clean + 1;
+
+	ntc = (ntc < rx_ring->count) ? ntc : 0;
+	rx_ring->next_to_clean = ntc;
+	prefetch(I40E_RX_DESC(rx_ring, ntc));
+}
+
 void i40e_xsk_clean_rx_ring(struct i40e_ring *rx_ring);
 void i40e_xsk_clean_tx_ring(struct i40e_ring *tx_ring);
 bool i40e_xsk_any_rx_ring_enabled(struct i40e_vsi *vsi);
diff --git a/drivers/net/ethernet/intel/i40e/i40e_type.h b/drivers/net/ethernet/intel/i40e/i40e_type.h
index 6ea2867ff60f..63e098f7cb63 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_type.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_type.h
@@ -689,7 +689,7 @@ union i40e_32byte_rx_desc {
 		__le64  rsvd2;
 	} read;
 	struct {
-		struct {
+		struct i40e_32b_rx_wb_qw0 {
 			struct {
 				union {
 					__le16 mirroring_status;
@@ -727,6 +727,9 @@ union i40e_32byte_rx_desc {
 			} hi_dword;
 		} qword3;
 	} wb;  /* writeback */
+	struct {
+		u64 qword[4];
+	} raw;
 };
 
 enum i40e_rx_desc_status_bits {
diff --git a/drivers/net/ethernet/intel/i40e/i40e_xsk.c b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
index d84ec92f8538..4fca52a30ea4 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_xsk.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
@@ -9,9 +9,23 @@
 #include "i40e_txrx_common.h"
 #include "i40e_xsk.h"
 
-static struct i40e_rx_buffer *i40e_rx_bi(struct i40e_ring *rx_ring, u32 idx)
+int i40e_alloc_rx_bi_zc(struct i40e_ring *rx_ring)
 {
-	return &rx_ring->rx_bi[idx];
+	unsigned long sz = sizeof(*rx_ring->rx_bi_zc) * rx_ring->count;
+
+	rx_ring->rx_bi_zc = kzalloc(sz, GFP_KERNEL);
+	return rx_ring->rx_bi_zc ? 0 : -ENOMEM;
+}
+
+void i40e_clear_rx_bi_zc(struct i40e_ring *rx_ring)
+{
+	memset(rx_ring->rx_bi_zc, 0,
+	       sizeof(*rx_ring->rx_bi_zc) * rx_ring->count);
+}
+
+static struct i40e_rx_buffer_zc *i40e_rx_bi(struct i40e_ring *rx_ring, u32 idx)
+{
+	return &rx_ring->rx_bi_zc[idx];
 }
 
 /**
@@ -238,7 +252,7 @@ static int i40e_run_xdp_zc(struct i40e_ring *rx_ring, struct xdp_buff *xdp)
 }
 
 /**
- * i40e_alloc_buffer_zc - Allocates an i40e_rx_buffer
+ * i40e_alloc_buffer_zc - Allocates an i40e_rx_buffer_zc
  * @rx_ring: Rx ring
  * @bi: Rx buffer to populate
  *
@@ -248,7 +262,7 @@ static int i40e_run_xdp_zc(struct i40e_ring *rx_ring, struct xdp_buff *xdp)
  * Returns true for a successful allocation, false otherwise
  **/
 static bool i40e_alloc_buffer_zc(struct i40e_ring *rx_ring,
-				 struct i40e_rx_buffer *bi)
+				 struct i40e_rx_buffer_zc *bi)
 {
 	struct xdp_umem *umem = rx_ring->xsk_umem;
 	void *addr = bi->addr;
@@ -279,7 +293,7 @@ static bool i40e_alloc_buffer_zc(struct i40e_ring *rx_ring,
 }
 
 /**
- * i40e_alloc_buffer_slow_zc - Allocates an i40e_rx_buffer
+ * i40e_alloc_buffer_slow_zc - Allocates an i40e_rx_buffer_zc
  * @rx_ring: Rx ring
  * @bi: Rx buffer to populate
  *
@@ -289,7 +303,7 @@ static bool i40e_alloc_buffer_zc(struct i40e_ring *rx_ring,
  * Returns true for a successful allocation, false otherwise
  **/
 static bool i40e_alloc_buffer_slow_zc(struct i40e_ring *rx_ring,
-				      struct i40e_rx_buffer *bi)
+				      struct i40e_rx_buffer_zc *bi)
 {
 	struct xdp_umem *umem = rx_ring->xsk_umem;
 	u64 handle, hr;
@@ -318,11 +332,11 @@ static bool i40e_alloc_buffer_slow_zc(struct i40e_ring *rx_ring,
 static __always_inline bool
 __i40e_alloc_rx_buffers_zc(struct i40e_ring *rx_ring, u16 count,
 			   bool alloc(struct i40e_ring *rx_ring,
-				      struct i40e_rx_buffer *bi))
+				      struct i40e_rx_buffer_zc *bi))
 {
 	u16 ntu = rx_ring->next_to_use;
 	union i40e_rx_desc *rx_desc;
-	struct i40e_rx_buffer *bi;
+	struct i40e_rx_buffer_zc *bi;
 	bool ok = true;
 
 	rx_desc = I40E_RX_DESC(rx_ring, ntu);
@@ -402,10 +416,11 @@ static bool i40e_alloc_rx_buffers_fast_zc(struct i40e_ring *rx_ring, u16 count)
  *
  * Returns the received Rx buffer
  **/
-static struct i40e_rx_buffer *i40e_get_rx_buffer_zc(struct i40e_ring *rx_ring,
-						    const unsigned int size)
+static struct i40e_rx_buffer_zc *i40e_get_rx_buffer_zc(
+	struct i40e_ring *rx_ring,
+	const unsigned int size)
 {
-	struct i40e_rx_buffer *bi;
+	struct i40e_rx_buffer_zc *bi;
 
 	bi = i40e_rx_bi(rx_ring, rx_ring->next_to_clean);
 
@@ -427,10 +442,10 @@ static struct i40e_rx_buffer *i40e_get_rx_buffer_zc(struct i40e_ring *rx_ring,
  * recycle queue (next_to_alloc).
  **/
 static void i40e_reuse_rx_buffer_zc(struct i40e_ring *rx_ring,
-				    struct i40e_rx_buffer *old_bi)
+				    struct i40e_rx_buffer_zc *old_bi)
 {
-	struct i40e_rx_buffer *new_bi = i40e_rx_bi(rx_ring,
-						   rx_ring->next_to_alloc);
+	struct i40e_rx_buffer_zc *new_bi = i40e_rx_bi(rx_ring,
+						      rx_ring->next_to_alloc);
 	u16 nta = rx_ring->next_to_alloc;
 
 	/* update, and store next to alloc */
@@ -452,7 +467,7 @@ static void i40e_reuse_rx_buffer_zc(struct i40e_ring *rx_ring,
  **/
 void i40e_zca_free(struct zero_copy_allocator *alloc, unsigned long handle)
 {
-	struct i40e_rx_buffer *bi;
+	struct i40e_rx_buffer_zc *bi;
 	struct i40e_ring *rx_ring;
 	u64 hr, mask;
 	u16 nta;
@@ -490,7 +505,7 @@ void i40e_zca_free(struct zero_copy_allocator *alloc, unsigned long handle)
  * Returns the skb, or NULL on failure.
  **/
 static struct sk_buff *i40e_construct_skb_zc(struct i40e_ring *rx_ring,
-					     struct i40e_rx_buffer *bi,
+					     struct i40e_rx_buffer_zc *bi,
 					     struct xdp_buff *xdp)
 {
 	unsigned int metasize = xdp->data - xdp->data_meta;
@@ -513,19 +528,6 @@ static struct sk_buff *i40e_construct_skb_zc(struct i40e_ring *rx_ring,
 	return skb;
 }
 
-/**
- * i40e_inc_ntc: Advance the next_to_clean index
- * @rx_ring: Rx ring
- **/
-static void i40e_inc_ntc(struct i40e_ring *rx_ring)
-{
-	u32 ntc = rx_ring->next_to_clean + 1;
-
-	ntc = (ntc < rx_ring->count) ? ntc : 0;
-	rx_ring->next_to_clean = ntc;
-	prefetch(I40E_RX_DESC(rx_ring, ntc));
-}
-
 /**
  * i40e_clean_rx_irq_zc - Consumes Rx packets from the hardware ring
  * @rx_ring: Rx ring
@@ -547,7 +549,7 @@ int i40e_clean_rx_irq_zc(struct i40e_ring *rx_ring, int budget)
 	xdp.frame_sz = xsk_umem_xdp_frame_sz(umem);
 
 	while (likely(total_rx_packets < (unsigned int)budget)) {
-		struct i40e_rx_buffer *bi;
+		struct i40e_rx_buffer_zc *bi;
 		union i40e_rx_desc *rx_desc;
 		unsigned int size;
 		u64 qword;
@@ -568,14 +570,18 @@ int i40e_clean_rx_irq_zc(struct i40e_ring *rx_ring, int budget)
 		 */
 		dma_rmb();
 
-		bi = i40e_clean_programming_status(rx_ring, rx_desc,
-						   qword);
-		if (unlikely(bi)) {
+		if (i40e_rx_is_programming_status(qword)) {
+			i40e_clean_programming_status(rx_ring,
+						      rx_desc->raw.qword[0],
+						      qword);
+			bi = i40e_rx_bi(rx_ring, rx_ring->next_to_clean);
+			i40e_inc_ntc(rx_ring);
 			i40e_reuse_rx_buffer_zc(rx_ring, bi);
 			cleaned_count++;
 			continue;
 		}
 
+		bi = i40e_rx_bi(rx_ring, rx_ring->next_to_clean);
 		size = (qword & I40E_RXD_QW1_LENGTH_PBUF_MASK) >>
 		       I40E_RXD_QW1_LENGTH_PBUF_SHIFT;
 		if (!size)
@@ -832,7 +838,7 @@ void i40e_xsk_clean_rx_ring(struct i40e_ring *rx_ring)
 	u16 i;
 
 	for (i = 0; i < rx_ring->count; i++) {
-		struct i40e_rx_buffer *rx_bi = i40e_rx_bi(rx_ring, i);
+		struct i40e_rx_buffer_zc *rx_bi = i40e_rx_bi(rx_ring, i);
 
 		if (!rx_bi->addr)
 			continue;
diff --git a/drivers/net/ethernet/intel/i40e/i40e_xsk.h b/drivers/net/ethernet/intel/i40e/i40e_xsk.h
index 9ed59c14eb55..f5e292c218ee 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_xsk.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_xsk.h
@@ -19,5 +19,7 @@ int i40e_clean_rx_irq_zc(struct i40e_ring *rx_ring, int budget);
 bool i40e_clean_xdp_tx_irq(struct i40e_vsi *vsi,
 			   struct i40e_ring *tx_ring, int napi_budget);
 int i40e_xsk_wakeup(struct net_device *dev, u32 queue_id, u32 flags);
+int i40e_alloc_rx_bi_zc(struct i40e_ring *rx_ring);
+void i40e_clear_rx_bi_zc(struct i40e_ring *rx_ring);
 
 #endif /* _I40E_XSK_H_ */
-- 
2.25.1

[PATCH bpf-next v4 08/15] i40e, xsk: migrate to new MEM_TYPE_XSK_BUFF_POOL

[Björn Töpel]

From: Björn Töpel <bjorn.topel@intel.com>

Remove MEM_TYPE_ZERO_COPY in favor of the new MEM_TYPE_XSK_BUFF_POOL
APIs. The AF_XDP zero-copy rx_bi ring is now simply a struct xdp_buff
pointer.

Cc: intel-wired-lan@lists.osuosl.org
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_main.c |  19 +-
 drivers/net/ethernet/intel/i40e/i40e_txrx.h |   9 +-
 drivers/net/ethernet/intel/i40e/i40e_xsk.c  | 352 ++------------------
 drivers/net/ethernet/intel/i40e/i40e_xsk.h  |   1 -
 4 files changed, 47 insertions(+), 334 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index 3e1695bb8262..ea7395b391e5 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -3266,21 +3266,19 @@ static int i40e_configure_rx_ring(struct i40e_ring *ring)
 		ret = i40e_alloc_rx_bi_zc(ring);
 		if (ret)
 			return ret;
-		ring->rx_buf_len = ring->xsk_umem->chunk_size_nohr -
-				   XDP_PACKET_HEADROOM;
+		ring->rx_buf_len = xsk_umem_get_rx_frame_size(ring->xsk_umem);
 		/* For AF_XDP ZC, we disallow packets to span on
 		 * multiple buffers, thus letting us skip that
 		 * handling in the fast-path.
 		 */
 		chain_len = 1;
-		ring->zca.free = i40e_zca_free;
 		ret = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
-						 MEM_TYPE_ZERO_COPY,
-						 &ring->zca);
+						 MEM_TYPE_XSK_BUFF_POOL,
+						 NULL);
 		if (ret)
 			return ret;
 		dev_info(&vsi->back->pdev->dev,
-			 "Registered XDP mem model MEM_TYPE_ZERO_COPY on Rx ring %d\n",
+			 "Registered XDP mem model MEM_TYPE_XSK_BUFF_POOL on Rx ring %d\n",
 			 ring->queue_index);
 
 	} else {
@@ -3351,9 +3349,12 @@ static int i40e_configure_rx_ring(struct i40e_ring *ring)
 	ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
 	writel(0, ring->tail);
 
-	ok = ring->xsk_umem ?
-	     i40e_alloc_rx_buffers_zc(ring, I40E_DESC_UNUSED(ring)) :
-	     !i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
+	if (ring->xsk_umem) {
+		xsk_buff_set_rxq_info(ring->xsk_umem, &ring->xdp_rxq);
+		ok = i40e_alloc_rx_buffers_zc(ring, I40E_DESC_UNUSED(ring));
+	} else {
+		ok = !i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
+	}
 	if (!ok) {
 		/* Log this in case the user has forgotten to give the kernel
 		 * any buffers, even later in the application.
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.h b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
index d343498e8de5..5c255977fd58 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
@@ -301,12 +301,6 @@ struct i40e_rx_buffer {
 	__u16 pagecnt_bias;
 };
 
-struct i40e_rx_buffer_zc {
-	dma_addr_t dma;
-	void *addr;
-	u64 handle;
-};
-
 struct i40e_queue_stats {
 	u64 packets;
 	u64 bytes;
@@ -356,7 +350,7 @@ struct i40e_ring {
 	union {
 		struct i40e_tx_buffer *tx_bi;
 		struct i40e_rx_buffer *rx_bi;
-		struct i40e_rx_buffer_zc *rx_bi_zc;
+		struct xdp_buff **rx_bi_zc;
 	};
 	DECLARE_BITMAP(state, __I40E_RING_STATE_NBITS);
 	u16 queue_index;		/* Queue number of ring */
@@ -418,7 +412,6 @@ struct i40e_ring {
 	struct i40e_channel *ch;
 	struct xdp_rxq_info xdp_rxq;
 	struct xdp_umem *xsk_umem;
-	struct zero_copy_allocator zca; /* ZC allocator anchor */
 } ____cacheline_internodealigned_in_smp;
 
 static inline bool ring_uses_build_skb(struct i40e_ring *ring)
diff --git a/drivers/net/ethernet/intel/i40e/i40e_xsk.c b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
index 4fca52a30ea4..460f5052e1db 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_xsk.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_xsk.c
@@ -23,68 +23,11 @@ void i40e_clear_rx_bi_zc(struct i40e_ring *rx_ring)
 	       sizeof(*rx_ring->rx_bi_zc) * rx_ring->count);
 }
 
-static struct i40e_rx_buffer_zc *i40e_rx_bi(struct i40e_ring *rx_ring, u32 idx)
+static struct xdp_buff **i40e_rx_bi(struct i40e_ring *rx_ring, u32 idx)
 {
 	return &rx_ring->rx_bi_zc[idx];
 }
 
-/**
- * i40e_xsk_umem_dma_map - DMA maps all UMEM memory for the netdev
- * @vsi: Current VSI
- * @umem: UMEM to DMA map
- *
- * Returns 0 on success, <0 on failure
- **/
-static int i40e_xsk_umem_dma_map(struct i40e_vsi *vsi, struct xdp_umem *umem)
-{
-	struct i40e_pf *pf = vsi->back;
-	struct device *dev;
-	unsigned int i, j;
-	dma_addr_t dma;
-
-	dev = &pf->pdev->dev;
-	for (i = 0; i < umem->npgs; i++) {
-		dma = dma_map_page_attrs(dev, umem->pgs[i], 0, PAGE_SIZE,
-					 DMA_BIDIRECTIONAL, I40E_RX_DMA_ATTR);
-		if (dma_mapping_error(dev, dma))
-			goto out_unmap;
-
-		umem->pages[i].dma = dma;
-	}
-
-	return 0;
-
-out_unmap:
-	for (j = 0; j < i; j++) {
-		dma_unmap_page_attrs(dev, umem->pages[i].dma, PAGE_SIZE,
-				     DMA_BIDIRECTIONAL, I40E_RX_DMA_ATTR);
-		umem->pages[i].dma = 0;
-	}
-
-	return -1;
-}
-
-/**
- * i40e_xsk_umem_dma_unmap - DMA unmaps all UMEM memory for the netdev
- * @vsi: Current VSI
- * @umem: UMEM to DMA map
- **/
-static void i40e_xsk_umem_dma_unmap(struct i40e_vsi *vsi, struct xdp_umem *umem)
-{
-	struct i40e_pf *pf = vsi->back;
-	struct device *dev;
-	unsigned int i;
-
-	dev = &pf->pdev->dev;
-
-	for (i = 0; i < umem->npgs; i++) {
-		dma_unmap_page_attrs(dev, umem->pages[i].dma, PAGE_SIZE,
-				     DMA_BIDIRECTIONAL, I40E_RX_DMA_ATTR);
-
-		umem->pages[i].dma = 0;
-	}
-}
-
 /**
  * i40e_xsk_umem_enable - Enable/associate a UMEM to a certain ring/qid
  * @vsi: Current VSI
@@ -97,7 +40,6 @@ static int i40e_xsk_umem_enable(struct i40e_vsi *vsi, struct xdp_umem *umem,
 				u16 qid)
 {
 	struct net_device *netdev = vsi->netdev;
-	struct xdp_umem_fq_reuse *reuseq;
 	bool if_running;
 	int err;
 
@@ -111,13 +53,7 @@ static int i40e_xsk_umem_enable(struct i40e_vsi *vsi, struct xdp_umem *umem,
 	    qid >= netdev->real_num_tx_queues)
 		return -EINVAL;
 
-	reuseq = xsk_reuseq_prepare(vsi->rx_rings[0]->count);
-	if (!reuseq)
-		return -ENOMEM;
-
-	xsk_reuseq_free(xsk_reuseq_swap(umem, reuseq));
-
-	err = i40e_xsk_umem_dma_map(vsi, umem);
+	err = xsk_buff_dma_map(umem, &vsi->back->pdev->dev, I40E_RX_DMA_ATTR);
 	if (err)
 		return err;
 
@@ -170,7 +106,7 @@ static int i40e_xsk_umem_disable(struct i40e_vsi *vsi, u16 qid)
 	}
 
 	clear_bit(qid, vsi->af_xdp_zc_qps);
-	i40e_xsk_umem_dma_unmap(vsi, umem);
+	xsk_buff_dma_unmap(umem, I40E_RX_DMA_ATTR);
 
 	if (if_running) {
 		err = i40e_queue_pair_enable(vsi, qid);
@@ -209,11 +145,9 @@ int i40e_xsk_umem_setup(struct i40e_vsi *vsi, struct xdp_umem *umem,
  **/
 static int i40e_run_xdp_zc(struct i40e_ring *rx_ring, struct xdp_buff *xdp)
 {
-	struct xdp_umem *umem = rx_ring->xsk_umem;
 	int err, result = I40E_XDP_PASS;
 	struct i40e_ring *xdp_ring;
 	struct bpf_prog *xdp_prog;
-	u64 offset;
 	u32 act;
 
 	rcu_read_lock();
@@ -222,9 +156,6 @@ static int i40e_run_xdp_zc(struct i40e_ring *rx_ring, struct xdp_buff *xdp)
 	 */
 	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
 	act = bpf_prog_run_xdp(xdp_prog, xdp);
-	offset = xdp->data - xdp->data_hard_start;
-
-	xdp->handle = xsk_umem_adjust_offset(umem, xdp->handle, offset);
 
 	switch (act) {
 	case XDP_PASS:
@@ -251,107 +182,26 @@ static int i40e_run_xdp_zc(struct i40e_ring *rx_ring, struct xdp_buff *xdp)
 	return result;
 }
 
-/**
- * i40e_alloc_buffer_zc - Allocates an i40e_rx_buffer_zc
- * @rx_ring: Rx ring
- * @bi: Rx buffer to populate
- *
- * This function allocates an Rx buffer. The buffer can come from fill
- * queue, or via the recycle queue (next_to_alloc).
- *
- * Returns true for a successful allocation, false otherwise
- **/
-static bool i40e_alloc_buffer_zc(struct i40e_ring *rx_ring,
-				 struct i40e_rx_buffer_zc *bi)
-{
-	struct xdp_umem *umem = rx_ring->xsk_umem;
-	void *addr = bi->addr;
-	u64 handle, hr;
-
-	if (addr) {
-		rx_ring->rx_stats.page_reuse_count++;
-		return true;
-	}
-
-	if (!xsk_umem_peek_addr(umem, &handle)) {
-		rx_ring->rx_stats.alloc_page_failed++;
-		return false;
-	}
-
-	hr = umem->headroom + XDP_PACKET_HEADROOM;
-
-	bi->dma = xdp_umem_get_dma(umem, handle);
-	bi->dma += hr;
-
-	bi->addr = xdp_umem_get_data(umem, handle);
-	bi->addr += hr;
-
-	bi->handle = xsk_umem_adjust_offset(umem, handle, umem->headroom);
-
-	xsk_umem_release_addr(umem);
-	return true;
-}
-
-/**
- * i40e_alloc_buffer_slow_zc - Allocates an i40e_rx_buffer_zc
- * @rx_ring: Rx ring
- * @bi: Rx buffer to populate
- *
- * This function allocates an Rx buffer. The buffer can come from fill
- * queue, or via the reuse queue.
- *
- * Returns true for a successful allocation, false otherwise
- **/
-static bool i40e_alloc_buffer_slow_zc(struct i40e_ring *rx_ring,
-				      struct i40e_rx_buffer_zc *bi)
-{
-	struct xdp_umem *umem = rx_ring->xsk_umem;
-	u64 handle, hr;
-
-	if (!xsk_umem_peek_addr_rq(umem, &handle)) {
-		rx_ring->rx_stats.alloc_page_failed++;
-		return false;
-	}
-
-	handle &= rx_ring->xsk_umem->chunk_mask;
-
-	hr = umem->headroom + XDP_PACKET_HEADROOM;
-
-	bi->dma = xdp_umem_get_dma(umem, handle);
-	bi->dma += hr;
-
-	bi->addr = xdp_umem_get_data(umem, handle);
-	bi->addr += hr;
-
-	bi->handle = xsk_umem_adjust_offset(umem, handle, umem->headroom);
-
-	xsk_umem_release_addr_rq(umem);
-	return true;
-}
-
-static __always_inline bool
-__i40e_alloc_rx_buffers_zc(struct i40e_ring *rx_ring, u16 count,
-			   bool alloc(struct i40e_ring *rx_ring,
-				      struct i40e_rx_buffer_zc *bi))
+bool i40e_alloc_rx_buffers_zc(struct i40e_ring *rx_ring, u16 count)
 {
 	u16 ntu = rx_ring->next_to_use;
 	union i40e_rx_desc *rx_desc;
-	struct i40e_rx_buffer_zc *bi;
+	struct xdp_buff **bi, *xdp;
+	dma_addr_t dma;
 	bool ok = true;
 
 	rx_desc = I40E_RX_DESC(rx_ring, ntu);
 	bi = i40e_rx_bi(rx_ring, ntu);
 	do {
-		if (!alloc(rx_ring, bi)) {
+		xdp = xsk_buff_alloc(rx_ring->xsk_umem);
+		if (!xdp) {
 			ok = false;
 			goto no_buffers;
 		}
-
-		dma_sync_single_range_for_device(rx_ring->dev, bi->dma, 0,
-						 rx_ring->rx_buf_len,
-						 DMA_BIDIRECTIONAL);
-
-		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma);
+		*bi = xdp;
+		dma = xsk_buff_xdp_get_dma(xdp);
+		rx_desc->read.pkt_addr = cpu_to_le64(dma);
+		rx_desc->read.hdr_addr = 0;
 
 		rx_desc++;
 		bi++;
@@ -363,7 +213,6 @@ __i40e_alloc_rx_buffers_zc(struct i40e_ring *rx_ring, u16 count,
 			ntu = 0;
 		}
 
-		rx_desc->wb.qword1.status_error_len = 0;
 		count--;
 	} while (count);
 
@@ -374,126 +223,6 @@ __i40e_alloc_rx_buffers_zc(struct i40e_ring *rx_ring, u16 count,
 	return ok;
 }
 
-/**
- * i40e_alloc_rx_buffers_zc - Allocates a number of Rx buffers
- * @rx_ring: Rx ring
- * @count: The number of buffers to allocate
- *
- * This function allocates a number of Rx buffers from the reuse queue
- * or fill ring and places them on the Rx ring.
- *
- * Returns true for a successful allocation, false otherwise
- **/
-bool i40e_alloc_rx_buffers_zc(struct i40e_ring *rx_ring, u16 count)
-{
-	return __i40e_alloc_rx_buffers_zc(rx_ring, count,
-					  i40e_alloc_buffer_slow_zc);
-}
-
-/**
- * i40e_alloc_rx_buffers_fast_zc - Allocates a number of Rx buffers
- * @rx_ring: Rx ring
- * @count: The number of buffers to allocate
- *
- * This function allocates a number of Rx buffers from the fill ring
- * or the internal recycle mechanism and places them on the Rx ring.
- *
- * Returns true for a successful allocation, false otherwise
- **/
-static bool i40e_alloc_rx_buffers_fast_zc(struct i40e_ring *rx_ring, u16 count)
-{
-	return __i40e_alloc_rx_buffers_zc(rx_ring, count,
-					  i40e_alloc_buffer_zc);
-}
-
-/**
- * i40e_get_rx_buffer_zc - Return the current Rx buffer
- * @rx_ring: Rx ring
- * @size: The size of the rx buffer (read from descriptor)
- *
- * This function returns the current, received Rx buffer, and also
- * does DMA synchronization.  the Rx ring.
- *
- * Returns the received Rx buffer
- **/
-static struct i40e_rx_buffer_zc *i40e_get_rx_buffer_zc(
-	struct i40e_ring *rx_ring,
-	const unsigned int size)
-{
-	struct i40e_rx_buffer_zc *bi;
-
-	bi = i40e_rx_bi(rx_ring, rx_ring->next_to_clean);
-
-	/* we are reusing so sync this buffer for CPU use */
-	dma_sync_single_range_for_cpu(rx_ring->dev,
-				      bi->dma, 0,
-				      size,
-				      DMA_BIDIRECTIONAL);
-
-	return bi;
-}
-
-/**
- * i40e_reuse_rx_buffer_zc - Recycle an Rx buffer
- * @rx_ring: Rx ring
- * @old_bi: The Rx buffer to recycle
- *
- * This function recycles a finished Rx buffer, and places it on the
- * recycle queue (next_to_alloc).
- **/
-static void i40e_reuse_rx_buffer_zc(struct i40e_ring *rx_ring,
-				    struct i40e_rx_buffer_zc *old_bi)
-{
-	struct i40e_rx_buffer_zc *new_bi = i40e_rx_bi(rx_ring,
-						      rx_ring->next_to_alloc);
-	u16 nta = rx_ring->next_to_alloc;
-
-	/* 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 */
-	new_bi->dma = old_bi->dma;
-	new_bi->addr = old_bi->addr;
-	new_bi->handle = old_bi->handle;
-
-	old_bi->addr = NULL;
-}
-
-/**
- * i40e_zca_free - Free callback for MEM_TYPE_ZERO_COPY allocations
- * @alloc: Zero-copy allocator
- * @handle: Buffer handle
- **/
-void i40e_zca_free(struct zero_copy_allocator *alloc, unsigned long handle)
-{
-	struct i40e_rx_buffer_zc *bi;
-	struct i40e_ring *rx_ring;
-	u64 hr, mask;
-	u16 nta;
-
-	rx_ring = container_of(alloc, struct i40e_ring, zca);
-	hr = rx_ring->xsk_umem->headroom + XDP_PACKET_HEADROOM;
-	mask = rx_ring->xsk_umem->chunk_mask;
-
-	nta = rx_ring->next_to_alloc;
-	bi = i40e_rx_bi(rx_ring, nta);
-
-	nta++;
-	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
-
-	handle &= mask;
-
-	bi->dma = xdp_umem_get_dma(rx_ring->xsk_umem, handle);
-	bi->dma += hr;
-
-	bi->addr = xdp_umem_get_data(rx_ring->xsk_umem, handle);
-	bi->addr += hr;
-
-	bi->handle = xsk_umem_adjust_offset(rx_ring->xsk_umem, (u64)handle,
-					    rx_ring->xsk_umem->headroom);
-}
-
 /**
  * i40e_construct_skb_zc - Create skbufff from zero-copy Rx buffer
  * @rx_ring: Rx ring
@@ -505,7 +234,6 @@ void i40e_zca_free(struct zero_copy_allocator *alloc, unsigned long handle)
  * Returns the skb, or NULL on failure.
  **/
 static struct sk_buff *i40e_construct_skb_zc(struct i40e_ring *rx_ring,
-					     struct i40e_rx_buffer_zc *bi,
 					     struct xdp_buff *xdp)
 {
 	unsigned int metasize = xdp->data - xdp->data_meta;
@@ -524,7 +252,7 @@ static struct sk_buff *i40e_construct_skb_zc(struct i40e_ring *rx_ring,
 	if (metasize)
 		skb_metadata_set(skb, metasize);
 
-	i40e_reuse_rx_buffer_zc(rx_ring, bi);
+	xsk_buff_free(xdp);
 	return skb;
 }
 
@@ -539,25 +267,20 @@ int i40e_clean_rx_irq_zc(struct i40e_ring *rx_ring, int budget)
 {
 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
 	u16 cleaned_count = I40E_DESC_UNUSED(rx_ring);
-	struct xdp_umem *umem = rx_ring->xsk_umem;
 	unsigned int xdp_res, xdp_xmit = 0;
 	bool failure = false;
 	struct sk_buff *skb;
-	struct xdp_buff xdp;
-
-	xdp.rxq = &rx_ring->xdp_rxq;
-	xdp.frame_sz = xsk_umem_xdp_frame_sz(umem);
 
 	while (likely(total_rx_packets < (unsigned int)budget)) {
-		struct i40e_rx_buffer_zc *bi;
 		union i40e_rx_desc *rx_desc;
+		struct xdp_buff **bi;
 		unsigned int size;
 		u64 qword;
 
 		if (cleaned_count >= I40E_RX_BUFFER_WRITE) {
 			failure = failure ||
-				  !i40e_alloc_rx_buffers_fast_zc(rx_ring,
-								 cleaned_count);
+				  !i40e_alloc_rx_buffers_zc(rx_ring,
+							    cleaned_count);
 			cleaned_count = 0;
 		}
 
@@ -575,9 +298,10 @@ int i40e_clean_rx_irq_zc(struct i40e_ring *rx_ring, int budget)
 						      rx_desc->raw.qword[0],
 						      qword);
 			bi = i40e_rx_bi(rx_ring, rx_ring->next_to_clean);
-			i40e_inc_ntc(rx_ring);
-			i40e_reuse_rx_buffer_zc(rx_ring, bi);
+			xsk_buff_free(*bi);
+			*bi = NULL;
 			cleaned_count++;
+			i40e_inc_ntc(rx_ring);
 			continue;
 		}
 
@@ -587,22 +311,18 @@ int i40e_clean_rx_irq_zc(struct i40e_ring *rx_ring, int budget)
 		if (!size)
 			break;
 
-		bi = i40e_get_rx_buffer_zc(rx_ring, size);
-		xdp.data = bi->addr;
-		xdp.data_meta = xdp.data;
-		xdp.data_hard_start = xdp.data - XDP_PACKET_HEADROOM;
-		xdp.data_end = xdp.data + size;
-		xdp.handle = bi->handle;
+		bi = i40e_rx_bi(rx_ring, rx_ring->next_to_clean);
+		(*bi)->data_end = (*bi)->data + size;
+		xsk_buff_dma_sync_for_cpu(*bi);
 
-		xdp_res = i40e_run_xdp_zc(rx_ring, &xdp);
+		xdp_res = i40e_run_xdp_zc(rx_ring, *bi);
 		if (xdp_res) {
-			if (xdp_res & (I40E_XDP_TX | I40E_XDP_REDIR)) {
+			if (xdp_res & (I40E_XDP_TX | I40E_XDP_REDIR))
 				xdp_xmit |= xdp_res;
-				bi->addr = NULL;
-			} else {
-				i40e_reuse_rx_buffer_zc(rx_ring, bi);
-			}
+			else
+				xsk_buff_free(*bi);
 
+			*bi = NULL;
 			total_rx_bytes += size;
 			total_rx_packets++;
 
@@ -618,7 +338,8 @@ int i40e_clean_rx_irq_zc(struct i40e_ring *rx_ring, int budget)
 		 * BIT(I40E_RXD_QW1_ERROR_SHIFT). This is due to that
 		 * SBP is *not* set in PRT_SBPVSI (default not set).
 		 */
-		skb = i40e_construct_skb_zc(rx_ring, bi, &xdp);
+		skb = i40e_construct_skb_zc(rx_ring, *bi);
+		*bi = NULL;
 		if (!skb) {
 			rx_ring->rx_stats.alloc_buff_failed++;
 			break;
@@ -676,10 +397,9 @@ static bool i40e_xmit_zc(struct i40e_ring *xdp_ring, unsigned int budget)
 		if (!xsk_umem_consume_tx(xdp_ring->xsk_umem, &desc))
 			break;
 
-		dma = xdp_umem_get_dma(xdp_ring->xsk_umem, desc.addr);
-
-		dma_sync_single_for_device(xdp_ring->dev, dma, desc.len,
-					   DMA_BIDIRECTIONAL);
+		dma = xsk_buff_raw_get_dma(xdp_ring->xsk_umem, desc.addr);
+		xsk_buff_raw_dma_sync_for_device(xdp_ring->xsk_umem, dma,
+						 desc.len);
 
 		tx_bi = &xdp_ring->tx_bi[xdp_ring->next_to_use];
 		tx_bi->bytecount = desc.len;
@@ -838,13 +558,13 @@ void i40e_xsk_clean_rx_ring(struct i40e_ring *rx_ring)
 	u16 i;
 
 	for (i = 0; i < rx_ring->count; i++) {
-		struct i40e_rx_buffer_zc *rx_bi = i40e_rx_bi(rx_ring, i);
+		struct xdp_buff *rx_bi = *i40e_rx_bi(rx_ring, i);
 
-		if (!rx_bi->addr)
+		if (!rx_bi)
 			continue;
 
-		xsk_umem_fq_reuse(rx_ring->xsk_umem, rx_bi->handle);
-		rx_bi->addr = NULL;
+		xsk_buff_free(rx_bi);
+		rx_bi = NULL;
 	}
 }
 
diff --git a/drivers/net/ethernet/intel/i40e/i40e_xsk.h b/drivers/net/ethernet/intel/i40e/i40e_xsk.h
index f5e292c218ee..ea919a7d60ec 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_xsk.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_xsk.h
@@ -12,7 +12,6 @@ int i40e_queue_pair_disable(struct i40e_vsi *vsi, int queue_pair);
 int i40e_queue_pair_enable(struct i40e_vsi *vsi, int queue_pair);
 int i40e_xsk_umem_setup(struct i40e_vsi *vsi, struct xdp_umem *umem,
 			u16 qid);
-void i40e_zca_free(struct zero_copy_allocator *alloc, unsigned long handle);
 bool i40e_alloc_rx_buffers_zc(struct i40e_ring *rx_ring, u16 cleaned_count);
 int i40e_clean_rx_irq_zc(struct i40e_ring *rx_ring, int budget);
 
-- 
2.25.1

[PATCH bpf-next v4 09/15] ice, xsk: migrate to new MEM_TYPE_XSK_BUFF_POOL

[Björn Töpel]

From: Björn Töpel <bjorn.topel@intel.com>

Remove MEM_TYPE_ZERO_COPY in favor of the new MEM_TYPE_XSK_BUFF_POOL
APIs.

Cc: intel-wired-lan@lists.osuosl.org
Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
 drivers/net/ethernet/intel/ice/ice_base.c |  16 +-
 drivers/net/ethernet/intel/ice/ice_txrx.h |   8 +-
 drivers/net/ethernet/intel/ice/ice_xsk.c  | 374 +++-------------------
 drivers/net/ethernet/intel/ice/ice_xsk.h  |  13 +-
 4 files changed, 54 insertions(+), 357 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_base.c b/drivers/net/ethernet/intel/ice/ice_base.c
index a19cd6f5436b..433eb72b1c85 100644
--- a/drivers/net/ethernet/intel/ice/ice_base.c
+++ b/drivers/net/ethernet/intel/ice/ice_base.c
@@ -1,6 +1,7 @@
 // SPDX-License-Identifier: GPL-2.0
 /* Copyright (c) 2019, Intel Corporation. */
 
+#include <net/xdp_sock_drv.h>
 #include "ice_base.h"
 #include "ice_dcb_lib.h"
 
@@ -308,24 +309,23 @@ int ice_setup_rx_ctx(struct ice_ring *ring)
 		if (ring->xsk_umem) {
 			xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
 
-			ring->rx_buf_len = ring->xsk_umem->chunk_size_nohr -
-					   XDP_PACKET_HEADROOM;
+			ring->rx_buf_len =
+				xsk_umem_get_rx_frame_size(ring->xsk_umem);
 			/* For AF_XDP ZC, we disallow packets to span on
 			 * multiple buffers, thus letting us skip that
 			 * handling in the fast-path.
 			 */
 			chain_len = 1;
-			ring->zca.free = ice_zca_free;
 			err = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
-							 MEM_TYPE_ZERO_COPY,
-							 &ring->zca);
+							 MEM_TYPE_XSK_BUFF_POOL,
+							 NULL);
 			if (err)
 				return err;
+			xsk_buff_set_rxq_info(ring->xsk_umem, &ring->xdp_rxq);
 
-			dev_info(ice_pf_to_dev(vsi->back), "Registered XDP mem model MEM_TYPE_ZERO_COPY on Rx ring %d\n",
+			dev_info(ice_pf_to_dev(vsi->back), "Registered XDP mem model MEM_TYPE_XSK_BUFF_POOL on Rx ring %d\n",
 				 ring->q_index);
 		} else {
-			ring->zca.free = NULL;
 			if (!xdp_rxq_info_is_reg(&ring->xdp_rxq))
 				/* coverity[check_return] */
 				xdp_rxq_info_reg(&ring->xdp_rxq,
@@ -426,7 +426,7 @@ int ice_setup_rx_ctx(struct ice_ring *ring)
 	writel(0, ring->tail);
 
 	err = ring->xsk_umem ?
-	      ice_alloc_rx_bufs_slow_zc(ring, ICE_DESC_UNUSED(ring)) :
+	      ice_alloc_rx_bufs_zc(ring, ICE_DESC_UNUSED(ring)) :
 	      ice_alloc_rx_bufs(ring, ICE_DESC_UNUSED(ring));
 	if (err)
 		dev_info(ice_pf_to_dev(vsi->back), "Failed allocate some buffers on %sRx ring %d (pf_q %d)\n",
diff --git a/drivers/net/ethernet/intel/ice/ice_txrx.h b/drivers/net/ethernet/intel/ice/ice_txrx.h
index 7ee00a128663..d0fd2173854f 100644
--- a/drivers/net/ethernet/intel/ice/ice_txrx.h
+++ b/drivers/net/ethernet/intel/ice/ice_txrx.h
@@ -155,17 +155,16 @@ struct ice_tx_offload_params {
 };
 
 struct ice_rx_buf {
-	struct sk_buff *skb;
-	dma_addr_t dma;
 	union {
 		struct {
+			struct sk_buff *skb;
+			dma_addr_t dma;
 			struct page *page;
 			unsigned int page_offset;
 			u16 pagecnt_bias;
 		};
 		struct {
-			void *addr;
-			u64 handle;
+			struct xdp_buff *xdp;
 		};
 	};
 };
@@ -289,7 +288,6 @@ struct ice_ring {
 	struct rcu_head rcu;		/* to avoid race on free */
 	struct bpf_prog *xdp_prog;
 	struct xdp_umem *xsk_umem;
-	struct zero_copy_allocator zca;
 	/* CL3 - 3rd cacheline starts here */
 	struct xdp_rxq_info xdp_rxq;
 	/* CLX - the below items are only accessed infrequently and should be
diff --git a/drivers/net/ethernet/intel/ice/ice_xsk.c b/drivers/net/ethernet/intel/ice/ice_xsk.c
index 70e204307a93..da89589c3137 100644
--- a/drivers/net/ethernet/intel/ice/ice_xsk.c
+++ b/drivers/net/ethernet/intel/ice/ice_xsk.c
@@ -279,28 +279,6 @@ static int ice_xsk_alloc_umems(struct ice_vsi *vsi)
 	return 0;
 }
 
-/**
- * ice_xsk_add_umem - add a UMEM region for XDP sockets
- * @vsi: VSI to which the UMEM will be added
- * @umem: pointer to a requested UMEM region
- * @qid: queue ID
- *
- * Returns 0 on success, negative on error
- */
-static int ice_xsk_add_umem(struct ice_vsi *vsi, struct xdp_umem *umem, u16 qid)
-{
-	int err;
-
-	err = ice_xsk_alloc_umems(vsi);
-	if (err)
-		return err;
-
-	vsi->xsk_umems[qid] = umem;
-	vsi->num_xsk_umems_used++;
-
-	return 0;
-}
-
 /**
  * ice_xsk_remove_umem - Remove an UMEM for a certain ring/qid
  * @vsi: VSI from which the VSI will be removed
@@ -318,65 +296,6 @@ static void ice_xsk_remove_umem(struct ice_vsi *vsi, u16 qid)
 	}
 }
 
-/**
- * ice_xsk_umem_dma_map - DMA map UMEM region for XDP sockets
- * @vsi: VSI to map the UMEM region
- * @umem: UMEM to map
- *
- * Returns 0 on success, negative on error
- */
-static int ice_xsk_umem_dma_map(struct ice_vsi *vsi, struct xdp_umem *umem)
-{
-	struct ice_pf *pf = vsi->back;
-	struct device *dev;
-	unsigned int i;
-
-	dev = ice_pf_to_dev(pf);
-	for (i = 0; i < umem->npgs; i++) {
-		dma_addr_t dma = dma_map_page_attrs(dev, umem->pgs[i], 0,
-						    PAGE_SIZE,
-						    DMA_BIDIRECTIONAL,
-						    ICE_RX_DMA_ATTR);
-		if (dma_mapping_error(dev, dma)) {
-			dev_dbg(dev, "XSK UMEM DMA mapping error on page num %d\n",
-				i);
-			goto out_unmap;
-		}
-
-		umem->pages[i].dma = dma;
-	}
-
-	return 0;
-
-out_unmap:
-	for (; i > 0; i--) {
-		dma_unmap_page_attrs(dev, umem->pages[i].dma, PAGE_SIZE,
-				     DMA_BIDIRECTIONAL, ICE_RX_DMA_ATTR);
-		umem->pages[i].dma = 0;
-	}
-
-	return -EFAULT;
-}
-
-/**
- * ice_xsk_umem_dma_unmap - DMA unmap UMEM region for XDP sockets
- * @vsi: VSI from which the UMEM will be unmapped
- * @umem: UMEM to unmap
- */
-static void ice_xsk_umem_dma_unmap(struct ice_vsi *vsi, struct xdp_umem *umem)
-{
-	struct ice_pf *pf = vsi->back;
-	struct device *dev;
-	unsigned int i;
-
-	dev = ice_pf_to_dev(pf);
-	for (i = 0; i < umem->npgs; i++) {
-		dma_unmap_page_attrs(dev, umem->pages[i].dma, PAGE_SIZE,
-				     DMA_BIDIRECTIONAL, ICE_RX_DMA_ATTR);
-
-		umem->pages[i].dma = 0;
-	}
-}
 
 /**
  * ice_xsk_umem_disable - disable a UMEM region
@@ -391,7 +310,7 @@ static int ice_xsk_umem_disable(struct ice_vsi *vsi, u16 qid)
 	    !vsi->xsk_umems[qid])
 		return -EINVAL;
 
-	ice_xsk_umem_dma_unmap(vsi, vsi->xsk_umems[qid]);
+	xsk_buff_dma_unmap(vsi->xsk_umems[qid], ICE_RX_DMA_ATTR);
 	ice_xsk_remove_umem(vsi, qid);
 
 	return 0;
@@ -408,7 +327,6 @@ static int ice_xsk_umem_disable(struct ice_vsi *vsi, u16 qid)
 static int
 ice_xsk_umem_enable(struct ice_vsi *vsi, struct xdp_umem *umem, u16 qid)
 {
-	struct xdp_umem_fq_reuse *reuseq;
 	int err;
 
 	if (vsi->type != ICE_VSI_PF)
@@ -419,20 +337,18 @@ ice_xsk_umem_enable(struct ice_vsi *vsi, struct xdp_umem *umem, u16 qid)
 	if (qid >= vsi->num_xsk_umems)
 		return -EINVAL;
 
+	err = ice_xsk_alloc_umems(vsi);
+	if (err)
+		return err;
+
 	if (vsi->xsk_umems && vsi->xsk_umems[qid])
 		return -EBUSY;
 
-	reuseq = xsk_reuseq_prepare(vsi->rx_rings[0]->count);
-	if (!reuseq)
-		return -ENOMEM;
-
-	xsk_reuseq_free(xsk_reuseq_swap(umem, reuseq));
-
-	err = ice_xsk_umem_dma_map(vsi, umem);
-	if (err)
-		return err;
+	vsi->xsk_umems[qid] = umem;
+	vsi->num_xsk_umems_used++;
 
-	err = ice_xsk_add_umem(vsi, umem, qid);
+	err = xsk_buff_dma_map(vsi->xsk_umems[qid], ice_pf_to_dev(vsi->back),
+			       ICE_RX_DMA_ATTR);
 	if (err)
 		return err;
 
@@ -483,119 +399,6 @@ int ice_xsk_umem_setup(struct ice_vsi *vsi, struct xdp_umem *umem, u16 qid)
 	return ret;
 }
 
-/**
- * ice_zca_free - Callback for MEM_TYPE_ZERO_COPY allocations
- * @zca: zero-cpoy allocator
- * @handle: Buffer handle
- */
-void ice_zca_free(struct zero_copy_allocator *zca, unsigned long handle)
-{
-	struct ice_rx_buf *rx_buf;
-	struct ice_ring *rx_ring;
-	struct xdp_umem *umem;
-	u64 hr, mask;
-	u16 nta;
-
-	rx_ring = container_of(zca, struct ice_ring, zca);
-	umem = rx_ring->xsk_umem;
-	hr = umem->headroom + XDP_PACKET_HEADROOM;
-
-	mask = umem->chunk_mask;
-
-	nta = rx_ring->next_to_alloc;
-	rx_buf = &rx_ring->rx_buf[nta];
-
-	nta++;
-	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
-
-	handle &= mask;
-
-	rx_buf->dma = xdp_umem_get_dma(umem, handle);
-	rx_buf->dma += hr;
-
-	rx_buf->addr = xdp_umem_get_data(umem, handle);
-	rx_buf->addr += hr;
-
-	rx_buf->handle = (u64)handle + umem->headroom;
-}
-
-/**
- * ice_alloc_buf_fast_zc - Retrieve buffer address from XDP umem
- * @rx_ring: ring with an xdp_umem bound to it
- * @rx_buf: buffer to which xsk page address will be assigned
- *
- * This function allocates an Rx buffer in the hot path.
- * The buffer can come from fill queue or recycle queue.
- *
- * Returns true if an assignment was successful, false if not.
- */
-static __always_inline bool
-ice_alloc_buf_fast_zc(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf)
-{
-	struct xdp_umem *umem = rx_ring->xsk_umem;
-	void *addr = rx_buf->addr;
-	u64 handle, hr;
-
-	if (addr) {
-		rx_ring->rx_stats.page_reuse_count++;
-		return true;
-	}
-
-	if (!xsk_umem_peek_addr(umem, &handle)) {
-		rx_ring->rx_stats.alloc_page_failed++;
-		return false;
-	}
-
-	hr = umem->headroom + XDP_PACKET_HEADROOM;
-
-	rx_buf->dma = xdp_umem_get_dma(umem, handle);
-	rx_buf->dma += hr;
-
-	rx_buf->addr = xdp_umem_get_data(umem, handle);
-	rx_buf->addr += hr;
-
-	rx_buf->handle = handle + umem->headroom;
-
-	xsk_umem_release_addr(umem);
-	return true;
-}
-
-/**
- * ice_alloc_buf_slow_zc - Retrieve buffer address from XDP umem
- * @rx_ring: ring with an xdp_umem bound to it
- * @rx_buf: buffer to which xsk page address will be assigned
- *
- * This function allocates an Rx buffer in the slow path.
- * The buffer can come from fill queue or recycle queue.
- *
- * Returns true if an assignment was successful, false if not.
- */
-static __always_inline bool
-ice_alloc_buf_slow_zc(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf)
-{
-	struct xdp_umem *umem = rx_ring->xsk_umem;
-	u64 handle, headroom;
-
-	if (!xsk_umem_peek_addr_rq(umem, &handle)) {
-		rx_ring->rx_stats.alloc_page_failed++;
-		return false;
-	}
-
-	handle &= umem->chunk_mask;
-	headroom = umem->headroom + XDP_PACKET_HEADROOM;
-
-	rx_buf->dma = xdp_umem_get_dma(umem, handle);
-	rx_buf->dma += headroom;
-
-	rx_buf->addr = xdp_umem_get_data(umem, handle);
-	rx_buf->addr += headroom;
-
-	rx_buf->handle = handle + umem->headroom;
-
-	xsk_umem_release_addr_rq(umem);
-	return true;
-}
-
 /**
  * ice_alloc_rx_bufs_zc - allocate a number of Rx buffers
  * @rx_ring: Rx ring
@@ -607,14 +410,13 @@ ice_alloc_buf_slow_zc(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf)
  *
  * Returns false if all allocations were successful, true if any fail.
  */
-static bool
-ice_alloc_rx_bufs_zc(struct ice_ring *rx_ring, int count,
-		     bool (*alloc)(struct ice_ring *, struct ice_rx_buf *))
+bool ice_alloc_rx_bufs_zc(struct ice_ring *rx_ring, u16 count)
 {
 	union ice_32b_rx_flex_desc *rx_desc;
 	u16 ntu = rx_ring->next_to_use;
 	struct ice_rx_buf *rx_buf;
 	bool ret = false;
+	dma_addr_t dma;
 
 	if (!count)
 		return false;
@@ -623,16 +425,14 @@ ice_alloc_rx_bufs_zc(struct ice_ring *rx_ring, int count,
 	rx_buf = &rx_ring->rx_buf[ntu];
 
 	do {
-		if (!alloc(rx_ring, rx_buf)) {
+		rx_buf->xdp = xsk_buff_alloc(rx_ring->xsk_umem);
+		if (!rx_buf->xdp) {
 			ret = true;
 			break;
 		}
 
-		dma_sync_single_range_for_device(rx_ring->dev, rx_buf->dma, 0,
-						 rx_ring->rx_buf_len,
-						 DMA_BIDIRECTIONAL);
-
-		rx_desc->read.pkt_addr = cpu_to_le64(rx_buf->dma);
+		dma = xsk_buff_xdp_get_dma(rx_buf->xdp);
+		rx_desc->read.pkt_addr = cpu_to_le64(dma);
 		rx_desc->wb.status_error0 = 0;
 
 		rx_desc++;
@@ -652,32 +452,6 @@ ice_alloc_rx_bufs_zc(struct ice_ring *rx_ring, int count,
 	return ret;
 }
 
-/**
- * ice_alloc_rx_bufs_fast_zc - allocate zero copy bufs in the hot path
- * @rx_ring: Rx ring
- * @count: number of bufs to allocate
- *
- * Returns false on success, true on failure.
- */
-static bool ice_alloc_rx_bufs_fast_zc(struct ice_ring *rx_ring, u16 count)
-{
-	return ice_alloc_rx_bufs_zc(rx_ring, count,
-				    ice_alloc_buf_fast_zc);
-}
-
-/**
- * ice_alloc_rx_bufs_slow_zc - allocate zero copy bufs in the slow path
- * @rx_ring: Rx ring
- * @count: number of bufs to allocate
- *
- * Returns false on success, true on failure.
- */
-bool ice_alloc_rx_bufs_slow_zc(struct ice_ring *rx_ring, u16 count)
-{
-	return ice_alloc_rx_bufs_zc(rx_ring, count,
-				    ice_alloc_buf_slow_zc);
-}
-
 /**
  * ice_bump_ntc - Bump the next_to_clean counter of an Rx ring
  * @rx_ring: Rx ring
@@ -691,59 +465,6 @@ static void ice_bump_ntc(struct ice_ring *rx_ring)
 	prefetch(ICE_RX_DESC(rx_ring, ntc));
 }
 
-/**
- * ice_get_rx_buf_zc - Fetch the current Rx buffer
- * @rx_ring: Rx ring
- * @size: size of a buffer
- *
- * This function returns the current, received Rx buffer and does
- * DMA synchronization.
- *
- * Returns a pointer to the received Rx buffer.
- */
-static struct ice_rx_buf *ice_get_rx_buf_zc(struct ice_ring *rx_ring, int size)
-{
-	struct ice_rx_buf *rx_buf;
-
-	rx_buf = &rx_ring->rx_buf[rx_ring->next_to_clean];
-
-	dma_sync_single_range_for_cpu(rx_ring->dev, rx_buf->dma, 0,
-				      size, DMA_BIDIRECTIONAL);
-
-	return rx_buf;
-}
-
-/**
- * ice_reuse_rx_buf_zc - reuse an Rx buffer
- * @rx_ring: Rx ring
- * @old_buf: The buffer to recycle
- *
- * This function recycles a finished Rx buffer, and places it on the recycle
- * queue (next_to_alloc).
- */
-static void
-ice_reuse_rx_buf_zc(struct ice_ring *rx_ring, struct ice_rx_buf *old_buf)
-{
-	unsigned long mask = (unsigned long)rx_ring->xsk_umem->chunk_mask;
-	u64 hr = rx_ring->xsk_umem->headroom + XDP_PACKET_HEADROOM;
-	u16 nta = rx_ring->next_to_alloc;
-	struct ice_rx_buf *new_buf;
-
-	new_buf = &rx_ring->rx_buf[nta++];
-	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
-
-	new_buf->dma = old_buf->dma & mask;
-	new_buf->dma += hr;
-
-	new_buf->addr = (void *)((unsigned long)old_buf->addr & mask);
-	new_buf->addr += hr;
-
-	new_buf->handle = old_buf->handle & mask;
-	new_buf->handle += rx_ring->xsk_umem->headroom;
-
-	old_buf->addr = NULL;
-}
-
 /**
  * ice_construct_skb_zc - Create an sk_buff from zero-copy buffer
  * @rx_ring: Rx ring
@@ -755,13 +476,12 @@ ice_reuse_rx_buf_zc(struct ice_ring *rx_ring, struct ice_rx_buf *old_buf)
  * Returns the skb on success, NULL on failure.
  */
 static struct sk_buff *
-ice_construct_skb_zc(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf,
-		     struct xdp_buff *xdp)
+ice_construct_skb_zc(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf)
 {
-	unsigned int metasize = xdp->data - xdp->data_meta;
-	unsigned int datasize = xdp->data_end - xdp->data;
-	unsigned int datasize_hard = xdp->data_end -
-				     xdp->data_hard_start;
+	unsigned int metasize = rx_buf->xdp->data - rx_buf->xdp->data_meta;
+	unsigned int datasize = rx_buf->xdp->data_end - rx_buf->xdp->data;
+	unsigned int datasize_hard = rx_buf->xdp->data_end -
+				     rx_buf->xdp->data_hard_start;
 	struct sk_buff *skb;
 
 	skb = __napi_alloc_skb(&rx_ring->q_vector->napi, datasize_hard,
@@ -769,13 +489,13 @@ ice_construct_skb_zc(struct ice_ring *rx_ring, struct ice_rx_buf *rx_buf,
 	if (unlikely(!skb))
 		return NULL;
 
-	skb_reserve(skb, xdp->data - xdp->data_hard_start);
-	memcpy(__skb_put(skb, datasize), xdp->data, datasize);
+	skb_reserve(skb, rx_buf->xdp->data - rx_buf->xdp->data_hard_start);
+	memcpy(__skb_put(skb, datasize), rx_buf->xdp->data, datasize);
 	if (metasize)
 		skb_metadata_set(skb, metasize);
 
-	ice_reuse_rx_buf_zc(rx_ring, rx_buf);
-
+	xsk_buff_free(rx_buf->xdp);
+	rx_buf->xdp = NULL;
 	return skb;
 }
 
@@ -802,7 +522,6 @@ ice_run_xdp_zc(struct ice_ring *rx_ring, struct xdp_buff *xdp)
 	}
 
 	act = bpf_prog_run_xdp(xdp_prog, xdp);
-	xdp->handle += xdp->data - xdp->data_hard_start;
 	switch (act) {
 	case XDP_PASS:
 		break;
@@ -840,13 +559,8 @@ int ice_clean_rx_irq_zc(struct ice_ring *rx_ring, int budget)
 {
 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
 	u16 cleaned_count = ICE_DESC_UNUSED(rx_ring);
-	struct xdp_umem *umem = rx_ring->xsk_umem;
 	unsigned int xdp_xmit = 0;
 	bool failure = false;
-	struct xdp_buff xdp;
-
-	xdp.rxq = &rx_ring->xdp_rxq;
-	xdp.frame_sz = xsk_umem_xdp_frame_sz(umem);
 
 	while (likely(total_rx_packets < (unsigned int)budget)) {
 		union ice_32b_rx_flex_desc *rx_desc;
@@ -858,8 +572,8 @@ int ice_clean_rx_irq_zc(struct ice_ring *rx_ring, int budget)
 		u8 rx_ptype;
 
 		if (cleaned_count >= ICE_RX_BUF_WRITE) {
-			failure |= ice_alloc_rx_bufs_fast_zc(rx_ring,
-							     cleaned_count);
+			failure |= ice_alloc_rx_bufs_zc(rx_ring,
+							cleaned_count);
 			cleaned_count = 0;
 		}
 
@@ -880,25 +594,19 @@ int ice_clean_rx_irq_zc(struct ice_ring *rx_ring, int budget)
 		if (!size)
 			break;
 
-		rx_buf = ice_get_rx_buf_zc(rx_ring, size);
-		if (!rx_buf->addr)
-			break;
 
-		xdp.data = rx_buf->addr;
-		xdp.data_meta = xdp.data;
-		xdp.data_hard_start = xdp.data - XDP_PACKET_HEADROOM;
-		xdp.data_end = xdp.data + size;
-		xdp.handle = rx_buf->handle;
+		rx_buf = &rx_ring->rx_buf[rx_ring->next_to_clean];
+		rx_buf->xdp->data_end = rx_buf->xdp->data + size;
+		xsk_buff_dma_sync_for_cpu(rx_buf->xdp);
 
-		xdp_res = ice_run_xdp_zc(rx_ring, &xdp);
+		xdp_res = ice_run_xdp_zc(rx_ring, rx_buf->xdp);
 		if (xdp_res) {
-			if (xdp_res & (ICE_XDP_TX | ICE_XDP_REDIR)) {
+			if (xdp_res & (ICE_XDP_TX | ICE_XDP_REDIR))
 				xdp_xmit |= xdp_res;
-				rx_buf->addr = NULL;
-			} else {
-				ice_reuse_rx_buf_zc(rx_ring, rx_buf);
-			}
+			else
+				xsk_buff_free(rx_buf->xdp);
 
+			rx_buf->xdp = NULL;
 			total_rx_bytes += size;
 			total_rx_packets++;
 			cleaned_count++;
@@ -908,7 +616,7 @@ int ice_clean_rx_irq_zc(struct ice_ring *rx_ring, int budget)
 		}
 
 		/* XDP_PASS path */
-		skb = ice_construct_skb_zc(rx_ring, rx_buf, &xdp);
+		skb = ice_construct_skb_zc(rx_ring, rx_buf);
 		if (!skb) {
 			rx_ring->rx_stats.alloc_buf_failed++;
 			break;
@@ -979,10 +687,9 @@ static bool ice_xmit_zc(struct ice_ring *xdp_ring, int budget)
 		if (!xsk_umem_consume_tx(xdp_ring->xsk_umem, &desc))
 			break;
 
-		dma = xdp_umem_get_dma(xdp_ring->xsk_umem, desc.addr);
-
-		dma_sync_single_for_device(xdp_ring->dev, dma, desc.len,
-					   DMA_BIDIRECTIONAL);
+		dma = xsk_buff_raw_get_dma(xdp_ring->xsk_umem, desc.addr);
+		xsk_buff_raw_dma_sync_for_device(xdp_ring->xsk_umem, dma,
+						 desc.len);
 
 		tx_buf->bytecount = desc.len;
 
@@ -1165,11 +872,10 @@ void ice_xsk_clean_rx_ring(struct ice_ring *rx_ring)
 	for (i = 0; i < rx_ring->count; i++) {
 		struct ice_rx_buf *rx_buf = &rx_ring->rx_buf[i];
 
-		if (!rx_buf->addr)
+		if (!rx_buf->xdp)
 			continue;
 
-		xsk_umem_fq_reuse(rx_ring->xsk_umem, rx_buf->handle);
-		rx_buf->addr = NULL;
+		rx_buf->xdp = NULL;
 	}
 }
 
diff --git a/drivers/net/ethernet/intel/ice/ice_xsk.h b/drivers/net/ethernet/intel/ice/ice_xsk.h
index 8a4ba7c6d549..fc1a06b4df36 100644
--- a/drivers/net/ethernet/intel/ice/ice_xsk.h
+++ b/drivers/net/ethernet/intel/ice/ice_xsk.h
@@ -10,11 +10,10 @@ struct ice_vsi;
 
 #ifdef CONFIG_XDP_SOCKETS
 int ice_xsk_umem_setup(struct ice_vsi *vsi, struct xdp_umem *umem, u16 qid);
-void ice_zca_free(struct zero_copy_allocator *zca, unsigned long handle);
 int ice_clean_rx_irq_zc(struct ice_ring *rx_ring, int budget);
 bool ice_clean_tx_irq_zc(struct ice_ring *xdp_ring, int budget);
 int ice_xsk_wakeup(struct net_device *netdev, u32 queue_id, u32 flags);
-bool ice_alloc_rx_bufs_slow_zc(struct ice_ring *rx_ring, u16 count);
+bool ice_alloc_rx_bufs_zc(struct ice_ring *rx_ring, u16 count);
 bool ice_xsk_any_rx_ring_ena(struct ice_vsi *vsi);
 void ice_xsk_clean_rx_ring(struct ice_ring *rx_ring);
 void ice_xsk_clean_xdp_ring(struct ice_ring *xdp_ring);
@@ -27,12 +26,6 @@ ice_xsk_umem_setup(struct ice_vsi __always_unused *vsi,
 	return -EOPNOTSUPP;
 }
 
-static inline void
-ice_zca_free(struct zero_copy_allocator __always_unused *zca,
-	     unsigned long __always_unused handle)
-{
-}
-
 static inline int
 ice_clean_rx_irq_zc(struct ice_ring __always_unused *rx_ring,
 		    int __always_unused budget)
@@ -48,8 +41,8 @@ ice_clean_tx_irq_zc(struct ice_ring __always_unused *xdp_ring,
 }
 
 static inline bool
-ice_alloc_rx_bufs_slow_zc(struct ice_ring __always_unused *rx_ring,
-			  u16 __always_unused count)
+ice_alloc_rx_bufs_zc(struct ice_ring __always_unused *rx_ring,
+		     u16 __always_unused count)
 {
 	return false;
 }
-- 
2.25.1

[PATCH bpf-next v4 10/15] ixgbe, xsk: migrate to new MEM_TYPE_XSK_BUFF_POOL

[Björn Töpel]

From: Björn Töpel <bjorn.topel@intel.com>

Remove MEM_TYPE_ZERO_COPY in favor of the new MEM_TYPE_XSK_BUFF_POOL
APIs.

v1->v2: Fixed xdp_buff data_end update. (Björn)

Cc: intel-wired-lan@lists.osuosl.org
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe.h      |   9 +-
 drivers/net/ethernet/intel/ixgbe/ixgbe_main.c |  15 +-
 .../ethernet/intel/ixgbe/ixgbe_txrx_common.h  |   2 +-
 drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c  | 307 +++---------------
 4 files changed, 62 insertions(+), 271 deletions(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe.h b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
index 2833e4f041ce..5ddfc83a1e46 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe.h
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
@@ -224,17 +224,17 @@ struct ixgbe_tx_buffer {
 };
 
 struct ixgbe_rx_buffer {
-	struct sk_buff *skb;
-	dma_addr_t dma;
 	union {
 		struct {
+			struct sk_buff *skb;
+			dma_addr_t dma;
 			struct page *page;
 			__u32 page_offset;
 			__u16 pagecnt_bias;
 		};
 		struct {
-			void *addr;
-			u64 handle;
+			bool discard;
+			struct xdp_buff *xdp;
 		};
 	};
 };
@@ -351,7 +351,6 @@ struct ixgbe_ring {
 	};
 	struct xdp_rxq_info xdp_rxq;
 	struct xdp_umem *xsk_umem;
-	struct zero_copy_allocator zca; /* ZC allocator anchor */
 	u16 ring_idx;		/* {rx,tx,xdp}_ring back reference idx */
 	u16 rx_buf_len;
 } ____cacheline_internodealigned_in_smp;
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index eab5934b04f5..45fc7ce1a543 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -35,7 +35,7 @@
 #include <net/tc_act/tc_mirred.h>
 #include <net/vxlan.h>
 #include <net/mpls.h>
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 #include <net/xfrm.h>
 
 #include "ixgbe.h"
@@ -3745,8 +3745,7 @@ static void ixgbe_configure_srrctl(struct ixgbe_adapter *adapter,
 
 	/* configure the packet buffer length */
 	if (rx_ring->xsk_umem) {
-		u32 xsk_buf_len = rx_ring->xsk_umem->chunk_size_nohr -
-				  XDP_PACKET_HEADROOM;
+		u32 xsk_buf_len = xsk_umem_get_rx_frame_size(rx_ring->xsk_umem);
 
 		/* If the MAC support setting RXDCTL.RLPML, the
 		 * SRRCTL[n].BSIZEPKT is set to PAGE_SIZE and
@@ -4093,11 +4092,10 @@ void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
 	xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
 	ring->xsk_umem = ixgbe_xsk_umem(adapter, ring);
 	if (ring->xsk_umem) {
-		ring->zca.free = ixgbe_zca_free;
 		WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
-						   MEM_TYPE_ZERO_COPY,
-						   &ring->zca));
-
+						   MEM_TYPE_XSK_BUFF_POOL,
+						   NULL));
+		xsk_buff_set_rxq_info(ring->xsk_umem, &ring->xdp_rxq);
 	} else {
 		WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
 						   MEM_TYPE_PAGE_SHARED, NULL));
@@ -4153,8 +4151,7 @@ void ixgbe_configure_rx_ring(struct ixgbe_adapter *adapter,
 	}
 
 	if (ring->xsk_umem && hw->mac.type != ixgbe_mac_82599EB) {
-		u32 xsk_buf_len = ring->xsk_umem->chunk_size_nohr -
-				  XDP_PACKET_HEADROOM;
+		u32 xsk_buf_len = xsk_umem_get_rx_frame_size(ring->xsk_umem);
 
 		rxdctl &= ~(IXGBE_RXDCTL_RLPMLMASK |
 			    IXGBE_RXDCTL_RLPML_EN);
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_txrx_common.h b/drivers/net/ethernet/intel/ixgbe/ixgbe_txrx_common.h
index 6d01700b46bc..7887ae4aaf4f 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_txrx_common.h
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_txrx_common.h
@@ -35,7 +35,7 @@ int ixgbe_xsk_umem_setup(struct ixgbe_adapter *adapter, struct xdp_umem *umem,
 
 void ixgbe_zca_free(struct zero_copy_allocator *alloc, unsigned long handle);
 
-void ixgbe_alloc_rx_buffers_zc(struct ixgbe_ring *rx_ring, u16 cleaned_count);
+bool ixgbe_alloc_rx_buffers_zc(struct ixgbe_ring *rx_ring, u16 cleaned_count);
 int ixgbe_clean_rx_irq_zc(struct ixgbe_q_vector *q_vector,
 			  struct ixgbe_ring *rx_ring,
 			  const int budget);
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
index 82e4effae704..86add9fbd36c 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_xsk.c
@@ -20,54 +20,11 @@ struct xdp_umem *ixgbe_xsk_umem(struct ixgbe_adapter *adapter,
 	return xdp_get_umem_from_qid(adapter->netdev, qid);
 }
 
-static int ixgbe_xsk_umem_dma_map(struct ixgbe_adapter *adapter,
-				  struct xdp_umem *umem)
-{
-	struct device *dev = &adapter->pdev->dev;
-	unsigned int i, j;
-	dma_addr_t dma;
-
-	for (i = 0; i < umem->npgs; i++) {
-		dma = dma_map_page_attrs(dev, umem->pgs[i], 0, PAGE_SIZE,
-					 DMA_BIDIRECTIONAL, IXGBE_RX_DMA_ATTR);
-		if (dma_mapping_error(dev, dma))
-			goto out_unmap;
-
-		umem->pages[i].dma = dma;
-	}
-
-	return 0;
-
-out_unmap:
-	for (j = 0; j < i; j++) {
-		dma_unmap_page_attrs(dev, umem->pages[i].dma, PAGE_SIZE,
-				     DMA_BIDIRECTIONAL, IXGBE_RX_DMA_ATTR);
-		umem->pages[i].dma = 0;
-	}
-
-	return -1;
-}
-
-static void ixgbe_xsk_umem_dma_unmap(struct ixgbe_adapter *adapter,
-				     struct xdp_umem *umem)
-{
-	struct device *dev = &adapter->pdev->dev;
-	unsigned int i;
-
-	for (i = 0; i < umem->npgs; i++) {
-		dma_unmap_page_attrs(dev, umem->pages[i].dma, PAGE_SIZE,
-				     DMA_BIDIRECTIONAL, IXGBE_RX_DMA_ATTR);
-
-		umem->pages[i].dma = 0;
-	}
-}
-
 static int ixgbe_xsk_umem_enable(struct ixgbe_adapter *adapter,
 				 struct xdp_umem *umem,
 				 u16 qid)
 {
 	struct net_device *netdev = adapter->netdev;
-	struct xdp_umem_fq_reuse *reuseq;
 	bool if_running;
 	int err;
 
@@ -78,13 +35,7 @@ static int ixgbe_xsk_umem_enable(struct ixgbe_adapter *adapter,
 	    qid >= netdev->real_num_tx_queues)
 		return -EINVAL;
 
-	reuseq = xsk_reuseq_prepare(adapter->rx_ring[0]->count);
-	if (!reuseq)
-		return -ENOMEM;
-
-	xsk_reuseq_free(xsk_reuseq_swap(umem, reuseq));
-
-	err = ixgbe_xsk_umem_dma_map(adapter, umem);
+	err = xsk_buff_dma_map(umem, &adapter->pdev->dev, IXGBE_RX_DMA_ATTR);
 	if (err)
 		return err;
 
@@ -124,7 +75,7 @@ static int ixgbe_xsk_umem_disable(struct ixgbe_adapter *adapter, u16 qid)
 		ixgbe_txrx_ring_disable(adapter, qid);
 
 	clear_bit(qid, adapter->af_xdp_zc_qps);
-	ixgbe_xsk_umem_dma_unmap(adapter, umem);
+	xsk_buff_dma_unmap(umem, IXGBE_RX_DMA_ATTR);
 
 	if (if_running)
 		ixgbe_txrx_ring_enable(adapter, qid);
@@ -143,19 +94,14 @@ static int ixgbe_run_xdp_zc(struct ixgbe_adapter *adapter,
 			    struct ixgbe_ring *rx_ring,
 			    struct xdp_buff *xdp)
 {
-	struct xdp_umem *umem = rx_ring->xsk_umem;
 	int err, result = IXGBE_XDP_PASS;
 	struct bpf_prog *xdp_prog;
 	struct xdp_frame *xdpf;
-	u64 offset;
 	u32 act;
 
 	rcu_read_lock();
 	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
 	act = bpf_prog_run_xdp(xdp_prog, xdp);
-	offset = xdp->data - xdp->data_hard_start;
-
-	xdp->handle = xsk_umem_adjust_offset(umem, xdp->handle, offset);
 
 	switch (act) {
 	case XDP_PASS:
@@ -186,140 +132,16 @@ static int ixgbe_run_xdp_zc(struct ixgbe_adapter *adapter,
 	return result;
 }
 
-static struct
-ixgbe_rx_buffer *ixgbe_get_rx_buffer_zc(struct ixgbe_ring *rx_ring,
-					unsigned int size)
-{
-	struct ixgbe_rx_buffer *bi;
-
-	bi = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
-
-	/* we are reusing so sync this buffer for CPU use */
-	dma_sync_single_range_for_cpu(rx_ring->dev,
-				      bi->dma, 0,
-				      size,
-				      DMA_BIDIRECTIONAL);
-
-	return bi;
-}
-
-static void ixgbe_reuse_rx_buffer_zc(struct ixgbe_ring *rx_ring,
-				     struct ixgbe_rx_buffer *obi)
-{
-	u16 nta = rx_ring->next_to_alloc;
-	struct ixgbe_rx_buffer *nbi;
-
-	nbi = &rx_ring->rx_buffer_info[rx_ring->next_to_alloc];
-	/* 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 */
-	nbi->dma = obi->dma;
-	nbi->addr = obi->addr;
-	nbi->handle = obi->handle;
-
-	obi->addr = NULL;
-	obi->skb = NULL;
-}
-
-void ixgbe_zca_free(struct zero_copy_allocator *alloc, unsigned long handle)
-{
-	struct ixgbe_rx_buffer *bi;
-	struct ixgbe_ring *rx_ring;
-	u64 hr, mask;
-	u16 nta;
-
-	rx_ring = container_of(alloc, struct ixgbe_ring, zca);
-	hr = rx_ring->xsk_umem->headroom + XDP_PACKET_HEADROOM;
-	mask = rx_ring->xsk_umem->chunk_mask;
-
-	nta = rx_ring->next_to_alloc;
-	bi = rx_ring->rx_buffer_info;
-
-	nta++;
-	rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
-
-	handle &= mask;
-
-	bi->dma = xdp_umem_get_dma(rx_ring->xsk_umem, handle);
-	bi->dma += hr;
-
-	bi->addr = xdp_umem_get_data(rx_ring->xsk_umem, handle);
-	bi->addr += hr;
-
-	bi->handle = xsk_umem_adjust_offset(rx_ring->xsk_umem, (u64)handle,
-					    rx_ring->xsk_umem->headroom);
-}
-
-static bool ixgbe_alloc_buffer_zc(struct ixgbe_ring *rx_ring,
-				  struct ixgbe_rx_buffer *bi)
-{
-	struct xdp_umem *umem = rx_ring->xsk_umem;
-	void *addr = bi->addr;
-	u64 handle, hr;
-
-	if (addr)
-		return true;
-
-	if (!xsk_umem_peek_addr(umem, &handle)) {
-		rx_ring->rx_stats.alloc_rx_page_failed++;
-		return false;
-	}
-
-	hr = umem->headroom + XDP_PACKET_HEADROOM;
-
-	bi->dma = xdp_umem_get_dma(umem, handle);
-	bi->dma += hr;
-
-	bi->addr = xdp_umem_get_data(umem, handle);
-	bi->addr += hr;
-
-	bi->handle = xsk_umem_adjust_offset(umem, handle, umem->headroom);
-
-	xsk_umem_release_addr(umem);
-	return true;
-}
-
-static bool ixgbe_alloc_buffer_slow_zc(struct ixgbe_ring *rx_ring,
-				       struct ixgbe_rx_buffer *bi)
-{
-	struct xdp_umem *umem = rx_ring->xsk_umem;
-	u64 handle, hr;
-
-	if (!xsk_umem_peek_addr_rq(umem, &handle)) {
-		rx_ring->rx_stats.alloc_rx_page_failed++;
-		return false;
-	}
-
-	handle &= rx_ring->xsk_umem->chunk_mask;
-
-	hr = umem->headroom + XDP_PACKET_HEADROOM;
-
-	bi->dma = xdp_umem_get_dma(umem, handle);
-	bi->dma += hr;
-
-	bi->addr = xdp_umem_get_data(umem, handle);
-	bi->addr += hr;
-
-	bi->handle = xsk_umem_adjust_offset(umem, handle, umem->headroom);
-
-	xsk_umem_release_addr_rq(umem);
-	return true;
-}
-
-static __always_inline bool
-__ixgbe_alloc_rx_buffers_zc(struct ixgbe_ring *rx_ring, u16 cleaned_count,
-			    bool alloc(struct ixgbe_ring *rx_ring,
-				       struct ixgbe_rx_buffer *bi))
+bool ixgbe_alloc_rx_buffers_zc(struct ixgbe_ring *rx_ring, u16 count)
 {
 	union ixgbe_adv_rx_desc *rx_desc;
 	struct ixgbe_rx_buffer *bi;
 	u16 i = rx_ring->next_to_use;
+	dma_addr_t dma;
 	bool ok = true;
 
 	/* nothing to do */
-	if (!cleaned_count)
+	if (!count)
 		return true;
 
 	rx_desc = IXGBE_RX_DESC(rx_ring, i);
@@ -327,21 +149,18 @@ __ixgbe_alloc_rx_buffers_zc(struct ixgbe_ring *rx_ring, u16 cleaned_count,
 	i -= rx_ring->count;
 
 	do {
-		if (!alloc(rx_ring, bi)) {
+		bi->xdp = xsk_buff_alloc(rx_ring->xsk_umem);
+		if (!bi->xdp) {
 			ok = false;
 			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_BIDIRECTIONAL);
+		dma = xsk_buff_xdp_get_dma(bi->xdp);
 
 		/* Refresh the desc even if buffer_addrs didn't change
 		 * because each write-back erases this info.
 		 */
-		rx_desc->read.pkt_addr = cpu_to_le64(bi->dma);
+		rx_desc->read.pkt_addr = cpu_to_le64(dma);
 
 		rx_desc++;
 		bi++;
@@ -355,17 +174,14 @@ __ixgbe_alloc_rx_buffers_zc(struct ixgbe_ring *rx_ring, u16 cleaned_count,
 		/* clear the length for the next_to_use descriptor */
 		rx_desc->wb.upper.length = 0;
 
-		cleaned_count--;
-	} while (cleaned_count);
+		count--;
+	} while (count);
 
 	i += rx_ring->count;
 
 	if (rx_ring->next_to_use != i) {
 		rx_ring->next_to_use = i;
 
-		/* update next to alloc since we have filled the ring */
-		rx_ring->next_to_alloc = i;
-
 		/* Force memory writes to complete before letting h/w
 		 * know there are new descriptors to fetch.  (Only
 		 * applicable for weak-ordered memory model archs,
@@ -378,40 +194,27 @@ __ixgbe_alloc_rx_buffers_zc(struct ixgbe_ring *rx_ring, u16 cleaned_count,
 	return ok;
 }
 
-void ixgbe_alloc_rx_buffers_zc(struct ixgbe_ring *rx_ring, u16 count)
-{
-	__ixgbe_alloc_rx_buffers_zc(rx_ring, count,
-				    ixgbe_alloc_buffer_slow_zc);
-}
-
-static bool ixgbe_alloc_rx_buffers_fast_zc(struct ixgbe_ring *rx_ring,
-					   u16 count)
-{
-	return __ixgbe_alloc_rx_buffers_zc(rx_ring, count,
-					   ixgbe_alloc_buffer_zc);
-}
-
 static struct sk_buff *ixgbe_construct_skb_zc(struct ixgbe_ring *rx_ring,
-					      struct ixgbe_rx_buffer *bi,
-					      struct xdp_buff *xdp)
+					      struct ixgbe_rx_buffer *bi)
 {
-	unsigned int metasize = xdp->data - xdp->data_meta;
-	unsigned int datasize = xdp->data_end - xdp->data;
+	unsigned int metasize = bi->xdp->data - bi->xdp->data_meta;
+	unsigned int datasize = bi->xdp->data_end - bi->xdp->data;
 	struct sk_buff *skb;
 
 	/* allocate a skb to store the frags */
 	skb = __napi_alloc_skb(&rx_ring->q_vector->napi,
-			       xdp->data_end - xdp->data_hard_start,
+			       bi->xdp->data_end - bi->xdp->data_hard_start,
 			       GFP_ATOMIC | __GFP_NOWARN);
 	if (unlikely(!skb))
 		return NULL;
 
-	skb_reserve(skb, xdp->data - xdp->data_hard_start);
-	memcpy(__skb_put(skb, datasize), xdp->data, datasize);
+	skb_reserve(skb, bi->xdp->data - bi->xdp->data_hard_start);
+	memcpy(__skb_put(skb, datasize), bi->xdp->data, datasize);
 	if (metasize)
 		skb_metadata_set(skb, metasize);
 
-	ixgbe_reuse_rx_buffer_zc(rx_ring, bi);
+	xsk_buff_free(bi->xdp);
+	bi->xdp = NULL;
 	return skb;
 }
 
@@ -431,14 +234,9 @@ int ixgbe_clean_rx_irq_zc(struct ixgbe_q_vector *q_vector,
 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
 	struct ixgbe_adapter *adapter = q_vector->adapter;
 	u16 cleaned_count = ixgbe_desc_unused(rx_ring);
-	struct xdp_umem *umem = rx_ring->xsk_umem;
 	unsigned int xdp_res, xdp_xmit = 0;
 	bool failure = false;
 	struct sk_buff *skb;
-	struct xdp_buff xdp;
-
-	xdp.rxq = &rx_ring->xdp_rxq;
-	xdp.frame_sz = xsk_umem_xdp_frame_sz(umem);
 
 	while (likely(total_rx_packets < budget)) {
 		union ixgbe_adv_rx_desc *rx_desc;
@@ -448,8 +246,8 @@ int ixgbe_clean_rx_irq_zc(struct ixgbe_q_vector *q_vector,
 		/* return some buffers to hardware, one at a time is too slow */
 		if (cleaned_count >= IXGBE_RX_BUFFER_WRITE) {
 			failure = failure ||
-				  !ixgbe_alloc_rx_buffers_fast_zc(rx_ring,
-								 cleaned_count);
+				  !ixgbe_alloc_rx_buffers_zc(rx_ring,
+							     cleaned_count);
 			cleaned_count = 0;
 		}
 
@@ -464,42 +262,40 @@ int ixgbe_clean_rx_irq_zc(struct ixgbe_q_vector *q_vector,
 		 */
 		dma_rmb();
 
-		bi = ixgbe_get_rx_buffer_zc(rx_ring, size);
+		bi = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
 
 		if (unlikely(!ixgbe_test_staterr(rx_desc,
 						 IXGBE_RXD_STAT_EOP))) {
 			struct ixgbe_rx_buffer *next_bi;
 
-			ixgbe_reuse_rx_buffer_zc(rx_ring, bi);
+			xsk_buff_free(bi->xdp);
+			bi->xdp = NULL;
 			ixgbe_inc_ntc(rx_ring);
 			next_bi =
 			       &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
-			next_bi->skb = ERR_PTR(-EINVAL);
+			next_bi->discard = true;
 			continue;
 		}
 
-		if (unlikely(bi->skb)) {
-			ixgbe_reuse_rx_buffer_zc(rx_ring, bi);
+		if (unlikely(bi->discard)) {
+			xsk_buff_free(bi->xdp);
+			bi->xdp = NULL;
+			bi->discard = false;
 			ixgbe_inc_ntc(rx_ring);
 			continue;
 		}
 
-		xdp.data = bi->addr;
-		xdp.data_meta = xdp.data;
-		xdp.data_hard_start = xdp.data - XDP_PACKET_HEADROOM;
-		xdp.data_end = xdp.data + size;
-		xdp.handle = bi->handle;
-
-		xdp_res = ixgbe_run_xdp_zc(adapter, rx_ring, &xdp);
+		bi->xdp->data_end = bi->xdp->data + size;
+		xsk_buff_dma_sync_for_cpu(bi->xdp);
+		xdp_res = ixgbe_run_xdp_zc(adapter, rx_ring, bi->xdp);
 
 		if (xdp_res) {
-			if (xdp_res & (IXGBE_XDP_TX | IXGBE_XDP_REDIR)) {
+			if (xdp_res & (IXGBE_XDP_TX | IXGBE_XDP_REDIR))
 				xdp_xmit |= xdp_res;
-				bi->addr = NULL;
-				bi->skb = NULL;
-			} else {
-				ixgbe_reuse_rx_buffer_zc(rx_ring, bi);
-			}
+			else
+				xsk_buff_free(bi->xdp);
+
+			bi->xdp = NULL;
 			total_rx_packets++;
 			total_rx_bytes += size;
 
@@ -509,7 +305,7 @@ int ixgbe_clean_rx_irq_zc(struct ixgbe_q_vector *q_vector,
 		}
 
 		/* XDP_PASS path */
-		skb = ixgbe_construct_skb_zc(rx_ring, bi, &xdp);
+		skb = ixgbe_construct_skb_zc(rx_ring, bi);
 		if (!skb) {
 			rx_ring->rx_stats.alloc_rx_buff_failed++;
 			break;
@@ -561,17 +357,17 @@ int ixgbe_clean_rx_irq_zc(struct ixgbe_q_vector *q_vector,
 
 void ixgbe_xsk_clean_rx_ring(struct ixgbe_ring *rx_ring)
 {
-	u16 i = rx_ring->next_to_clean;
-	struct ixgbe_rx_buffer *bi = &rx_ring->rx_buffer_info[i];
+	struct ixgbe_rx_buffer *bi;
+	u16 i;
 
-	while (i != rx_ring->next_to_alloc) {
-		xsk_umem_fq_reuse(rx_ring->xsk_umem, bi->handle);
-		i++;
-		bi++;
-		if (i == rx_ring->count) {
-			i = 0;
-			bi = rx_ring->rx_buffer_info;
-		}
+	for (i = 0; i < rx_ring->count; i++) {
+		bi = &rx_ring->rx_buffer_info[i];
+
+		if (!bi->xdp)
+			continue;
+
+		xsk_buff_free(bi->xdp);
+		bi->xdp = NULL;
 	}
 }
 
@@ -594,10 +390,9 @@ static bool ixgbe_xmit_zc(struct ixgbe_ring *xdp_ring, unsigned int budget)
 		if (!xsk_umem_consume_tx(xdp_ring->xsk_umem, &desc))
 			break;
 
-		dma = xdp_umem_get_dma(xdp_ring->xsk_umem, desc.addr);
-
-		dma_sync_single_for_device(xdp_ring->dev, dma, desc.len,
-					   DMA_BIDIRECTIONAL);
+		dma = xsk_buff_raw_get_dma(xdp_ring->xsk_umem, desc.addr);
+		xsk_buff_raw_dma_sync_for_device(xdp_ring->xsk_umem, dma,
+						 desc.len);
 
 		tx_bi = &xdp_ring->tx_buffer_info[xdp_ring->next_to_use];
 		tx_bi->bytecount = desc.len;
-- 
2.25.1

[PATCH bpf-next v4 11/15] mlx5, xsk: migrate to new MEM_TYPE_XSK_BUFF_POOL

[Björn Töpel]

From: Björn Töpel <bjorn.topel@intel.com>

Use the new MEM_TYPE_XSK_BUFF_POOL API in lieu of MEM_TYPE_ZERO_COPY in
mlx5e. It allows to drop a lot of code from the driver (which is now
common in AF_XDP core and was related to XSK RX frame allocation, DMA
mapping, etc.) and slightly improve performance (RX +0.8 Mpps, TX +0.4
Mpps).

rfc->v1: Put back the sanity check for XSK params, use XSK API to get
         the total headroom size. (Maxim)

v1->v2: Fix DMA address handling, set XDP metadata to invalid. (Maxim)

v2->v3: Handle frame_sz, use xsk_buff_xdp_get_frame_dma, use xsk_buff
        API for DMA sync on TX, add performance numbers. (Maxim)

v3->v4: Remove unused variable num_xsk_frames. (Jakub)

Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/en.h  |   7 +-
 .../ethernet/mellanox/mlx5/core/en/params.c   |  13 +-
 .../net/ethernet/mellanox/mlx5/core/en/xdp.c  |  31 ++---
 .../net/ethernet/mellanox/mlx5/core/en/xdp.h  |   2 +-
 .../ethernet/mellanox/mlx5/core/en/xsk/rx.c   | 113 ++++--------------
 .../ethernet/mellanox/mlx5/core/en/xsk/rx.h   |  23 +++-
 .../ethernet/mellanox/mlx5/core/en/xsk/tx.c   |   9 +-
 .../ethernet/mellanox/mlx5/core/en/xsk/umem.c |  49 +-------
 .../net/ethernet/mellanox/mlx5/core/en_main.c |  25 +---
 .../net/ethernet/mellanox/mlx5/core/en_rx.c   |  34 +++++-
 10 files changed, 96 insertions(+), 210 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h
index 26911b15f8fe..0a02b804b2fe 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h
@@ -407,10 +407,7 @@ struct mlx5e_dma_info {
 	dma_addr_t addr;
 	union {
 		struct page *page;
-		struct {
-			u64 handle;
-			void *data;
-		} xsk;
+		struct xdp_buff *xsk;
 	};
 };
 
@@ -623,7 +620,6 @@ struct mlx5e_rq {
 		} mpwqe;
 	};
 	struct {
-		u16            umem_headroom;
 		u16            headroom;
 		u32            frame0_sz;
 		u8             map_dir;   /* dma map direction */
@@ -656,7 +652,6 @@ struct mlx5e_rq {
 	struct page_pool      *page_pool;
 
 	/* AF_XDP zero-copy */
-	struct zero_copy_allocator zca;
 	struct xdp_umem       *umem;
 
 	struct work_struct     recover_work;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/params.c b/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
index eb2e1f2138e4..38e4f19d69f8 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/params.c
@@ -12,15 +12,16 @@ static inline bool mlx5e_rx_is_xdp(struct mlx5e_params *params,
 u16 mlx5e_get_linear_rq_headroom(struct mlx5e_params *params,
 				 struct mlx5e_xsk_param *xsk)
 {
-	u16 headroom = NET_IP_ALIGN;
+	u16 headroom;
 
-	if (mlx5e_rx_is_xdp(params, xsk)) {
+	if (xsk)
+		return xsk->headroom;
+
+	headroom = NET_IP_ALIGN;
+	if (mlx5e_rx_is_xdp(params, xsk))
 		headroom += XDP_PACKET_HEADROOM;
-		if (xsk)
-			headroom += xsk->headroom;
-	} else {
+	else
 		headroom += MLX5_RX_HEADROOM;
-	}
 
 	return headroom;
 }
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
index 3507d23f0eb8..a2a194525b15 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
@@ -71,7 +71,7 @@ mlx5e_xmit_xdp_buff(struct mlx5e_xdpsq *sq, struct mlx5e_rq *rq,
 	xdptxd.data = xdpf->data;
 	xdptxd.len  = xdpf->len;
 
-	if (xdp->rxq->mem.type == MEM_TYPE_ZERO_COPY) {
+	if (xdp->rxq->mem.type == MEM_TYPE_XSK_BUFF_POOL) {
 		/* The xdp_buff was in the UMEM and was copied into a newly
 		 * allocated page. The UMEM page was returned via the ZCA, and
 		 * this new page has to be mapped at this point and has to be
@@ -119,50 +119,33 @@ mlx5e_xmit_xdp_buff(struct mlx5e_xdpsq *sq, struct mlx5e_rq *rq,
 
 /* returns true if packet was consumed by xdp */
 bool mlx5e_xdp_handle(struct mlx5e_rq *rq, struct mlx5e_dma_info *di,
-		      void *va, u16 *rx_headroom, u32 *len, bool xsk)
+		      u32 *len, struct xdp_buff *xdp)
 {
 	struct bpf_prog *prog = READ_ONCE(rq->xdp_prog);
-	struct xdp_umem *umem = rq->umem;
-	struct xdp_buff xdp;
 	u32 act;
 	int err;
 
 	if (!prog)
 		return false;
 
-	xdp.data = va + *rx_headroom;
-	xdp_set_data_meta_invalid(&xdp);
-	xdp.data_end = xdp.data + *len;
-	xdp.data_hard_start = va;
-	if (xsk)
-		xdp.handle = di->xsk.handle;
-	xdp.rxq = &rq->xdp_rxq;
-	xdp.frame_sz = rq->buff.frame0_sz;
-
-	act = bpf_prog_run_xdp(prog, &xdp);
-	if (xsk) {
-		u64 off = xdp.data - xdp.data_hard_start;
-
-		xdp.handle = xsk_umem_adjust_offset(umem, xdp.handle, off);
-	}
+	act = bpf_prog_run_xdp(prog, xdp);
 	switch (act) {
 	case XDP_PASS:
-		*rx_headroom = xdp.data - xdp.data_hard_start;
-		*len = xdp.data_end - xdp.data;
+		*len = xdp->data_end - xdp->data;
 		return false;
 	case XDP_TX:
-		if (unlikely(!mlx5e_xmit_xdp_buff(rq->xdpsq, rq, di, &xdp)))
+		if (unlikely(!mlx5e_xmit_xdp_buff(rq->xdpsq, rq, di, xdp)))
 			goto xdp_abort;
 		__set_bit(MLX5E_RQ_FLAG_XDP_XMIT, rq->flags); /* non-atomic */
 		return true;
 	case XDP_REDIRECT:
 		/* When XDP enabled then page-refcnt==1 here */
-		err = xdp_do_redirect(rq->netdev, &xdp, prog);
+		err = xdp_do_redirect(rq->netdev, xdp, prog);
 		if (unlikely(err))
 			goto xdp_abort;
 		__set_bit(MLX5E_RQ_FLAG_XDP_XMIT, rq->flags);
 		__set_bit(MLX5E_RQ_FLAG_XDP_REDIRECT, rq->flags);
-		if (!xsk)
+		if (xdp->rxq->mem.type != MEM_TYPE_XSK_BUFF_POOL)
 			mlx5e_page_dma_unmap(rq, di);
 		rq->stats->xdp_redirect++;
 		return true;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
index e2e01f064c1e..2e4e117aeb49 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
@@ -63,7 +63,7 @@
 struct mlx5e_xsk_param;
 int mlx5e_xdp_max_mtu(struct mlx5e_params *params, struct mlx5e_xsk_param *xsk);
 bool mlx5e_xdp_handle(struct mlx5e_rq *rq, struct mlx5e_dma_info *di,
-		      void *va, u16 *rx_headroom, u32 *len, bool xsk);
+		      u32 *len, struct xdp_buff *xdp);
 void mlx5e_xdp_mpwqe_complete(struct mlx5e_xdpsq *sq);
 bool mlx5e_poll_xdpsq_cq(struct mlx5e_cq *cq);
 void mlx5e_free_xdpsq_descs(struct mlx5e_xdpsq *sq);
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.c
index 62fc8a128a8d..a33a1f762c70 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.c
@@ -3,71 +3,10 @@
 
 #include "rx.h"
 #include "en/xdp.h"
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 
 /* RX data path */
 
-bool mlx5e_xsk_pages_enough_umem(struct mlx5e_rq *rq, int count)
-{
-	/* Check in advance that we have enough frames, instead of allocating
-	 * one-by-one, failing and moving frames to the Reuse Ring.
-	 */
-	return xsk_umem_has_addrs_rq(rq->umem, count);
-}
-
-int mlx5e_xsk_page_alloc_umem(struct mlx5e_rq *rq,
-			      struct mlx5e_dma_info *dma_info)
-{
-	struct xdp_umem *umem = rq->umem;
-	u64 handle;
-
-	if (!xsk_umem_peek_addr_rq(umem, &handle))
-		return -ENOMEM;
-
-	dma_info->xsk.handle = xsk_umem_adjust_offset(umem, handle,
-						      rq->buff.umem_headroom);
-	dma_info->xsk.data = xdp_umem_get_data(umem, dma_info->xsk.handle);
-
-	/* No need to add headroom to the DMA address. In striding RQ case, we
-	 * just provide pages for UMR, and headroom is counted at the setup
-	 * stage when creating a WQE. In non-striding RQ case, headroom is
-	 * accounted in mlx5e_alloc_rx_wqe.
-	 */
-	dma_info->addr = xdp_umem_get_dma(umem, handle);
-
-	xsk_umem_release_addr_rq(umem);
-
-	dma_sync_single_for_device(rq->pdev, dma_info->addr, PAGE_SIZE,
-				   DMA_BIDIRECTIONAL);
-
-	return 0;
-}
-
-static inline void mlx5e_xsk_recycle_frame(struct mlx5e_rq *rq, u64 handle)
-{
-	xsk_umem_fq_reuse(rq->umem, handle & rq->umem->chunk_mask);
-}
-
-/* XSKRQ uses pages from UMEM, they must not be released. They are returned to
- * the userspace if possible, and if not, this function is called to reuse them
- * in the driver.
- */
-void mlx5e_xsk_page_release(struct mlx5e_rq *rq,
-			    struct mlx5e_dma_info *dma_info)
-{
-	mlx5e_xsk_recycle_frame(rq, dma_info->xsk.handle);
-}
-
-/* Return a frame back to the hardware to fill in again. It is used by XDP when
- * the XDP program returns XDP_TX or XDP_REDIRECT not to an XSKMAP.
- */
-void mlx5e_xsk_zca_free(struct zero_copy_allocator *zca, unsigned long handle)
-{
-	struct mlx5e_rq *rq = container_of(zca, struct mlx5e_rq, zca);
-
-	mlx5e_xsk_recycle_frame(rq, handle);
-}
-
 static struct sk_buff *mlx5e_xsk_construct_skb(struct mlx5e_rq *rq, void *data,
 					       u32 cqe_bcnt)
 {
@@ -90,11 +29,8 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq,
 						    u32 head_offset,
 						    u32 page_idx)
 {
-	struct mlx5e_dma_info *di = &wi->umr.dma_info[page_idx];
-	u16 rx_headroom = rq->buff.headroom - rq->buff.umem_headroom;
+	struct xdp_buff *xdp = wi->umr.dma_info[page_idx].xsk;
 	u32 cqe_bcnt32 = cqe_bcnt;
-	void *va, *data;
-	u32 frag_size;
 	bool consumed;
 
 	/* Check packet size. Note LRO doesn't use linear SKB */
@@ -103,22 +39,20 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq,
 		return NULL;
 	}
 
-	/* head_offset is not used in this function, because di->xsk.data and
-	 * di->addr point directly to the necessary place. Furthermore, in the
-	 * current implementation, UMR pages are mapped to XSK frames, so
+	/* head_offset is not used in this function, because xdp->data and the
+	 * DMA address point directly to the necessary place. Furthermore, in
+	 * the current implementation, UMR pages are mapped to XSK frames, so
 	 * head_offset should always be 0.
 	 */
 	WARN_ON_ONCE(head_offset);
 
-	va             = di->xsk.data;
-	data           = va + rx_headroom;
-	frag_size      = rq->buff.headroom + cqe_bcnt32;
-
-	dma_sync_single_for_cpu(rq->pdev, di->addr, frag_size, DMA_BIDIRECTIONAL);
-	prefetch(data);
+	xdp->data_end = xdp->data + cqe_bcnt32;
+	xdp_set_data_meta_invalid(xdp);
+	xsk_buff_dma_sync_for_cpu(xdp);
+	prefetch(xdp->data);
 
 	rcu_read_lock();
-	consumed = mlx5e_xdp_handle(rq, di, va, &rx_headroom, &cqe_bcnt32, true);
+	consumed = mlx5e_xdp_handle(rq, NULL, &cqe_bcnt32, xdp);
 	rcu_read_unlock();
 
 	/* Possible flows:
@@ -145,7 +79,7 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq,
 	/* XDP_PASS: copy the data from the UMEM to a new SKB and reuse the
 	 * frame. On SKB allocation failure, NULL is returned.
 	 */
-	return mlx5e_xsk_construct_skb(rq, data, cqe_bcnt32);
+	return mlx5e_xsk_construct_skb(rq, xdp->data, cqe_bcnt32);
 }
 
 struct sk_buff *mlx5e_xsk_skb_from_cqe_linear(struct mlx5e_rq *rq,
@@ -153,25 +87,20 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_linear(struct mlx5e_rq *rq,
 					      struct mlx5e_wqe_frag_info *wi,
 					      u32 cqe_bcnt)
 {
-	struct mlx5e_dma_info *di = wi->di;
-	u16 rx_headroom = rq->buff.headroom - rq->buff.umem_headroom;
-	void *va, *data;
+	struct xdp_buff *xdp = wi->di->xsk;
 	bool consumed;
-	u32 frag_size;
 
-	/* wi->offset is not used in this function, because di->xsk.data and
-	 * di->addr point directly to the necessary place. Furthermore, in the
-	 * current implementation, one page = one packet = one frame, so
+	/* wi->offset is not used in this function, because xdp->data and the
+	 * DMA address point directly to the necessary place. Furthermore, the
+	 * XSK allocator allocates frames per packet, instead of pages, so
 	 * wi->offset should always be 0.
 	 */
 	WARN_ON_ONCE(wi->offset);
 
-	va             = di->xsk.data;
-	data           = va + rx_headroom;
-	frag_size      = rq->buff.headroom + cqe_bcnt;
-
-	dma_sync_single_for_cpu(rq->pdev, di->addr, frag_size, DMA_BIDIRECTIONAL);
-	prefetch(data);
+	xdp->data_end = xdp->data + cqe_bcnt;
+	xdp_set_data_meta_invalid(xdp);
+	xsk_buff_dma_sync_for_cpu(xdp);
+	prefetch(xdp->data);
 
 	if (unlikely(get_cqe_opcode(cqe) != MLX5_CQE_RESP_SEND)) {
 		rq->stats->wqe_err++;
@@ -179,7 +108,7 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_linear(struct mlx5e_rq *rq,
 	}
 
 	rcu_read_lock();
-	consumed = mlx5e_xdp_handle(rq, di, va, &rx_headroom, &cqe_bcnt, true);
+	consumed = mlx5e_xdp_handle(rq, NULL, &cqe_bcnt, xdp);
 	rcu_read_unlock();
 
 	if (likely(consumed))
@@ -189,5 +118,5 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_linear(struct mlx5e_rq *rq,
 	 * will be handled by mlx5e_put_rx_frag.
 	 * On SKB allocation failure, NULL is returned.
 	 */
-	return mlx5e_xsk_construct_skb(rq, data, cqe_bcnt);
+	return mlx5e_xsk_construct_skb(rq, xdp->data, cqe_bcnt);
 }
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.h b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.h
index a8e11adbf426..d147b2f13b54 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/rx.h
@@ -9,12 +9,6 @@
 
 /* RX data path */
 
-bool mlx5e_xsk_pages_enough_umem(struct mlx5e_rq *rq, int count);
-int mlx5e_xsk_page_alloc_umem(struct mlx5e_rq *rq,
-			      struct mlx5e_dma_info *dma_info);
-void mlx5e_xsk_page_release(struct mlx5e_rq *rq,
-			    struct mlx5e_dma_info *dma_info);
-void mlx5e_xsk_zca_free(struct zero_copy_allocator *zca, unsigned long handle);
 struct sk_buff *mlx5e_xsk_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq,
 						    struct mlx5e_mpw_info *wi,
 						    u16 cqe_bcnt,
@@ -25,6 +19,23 @@ struct sk_buff *mlx5e_xsk_skb_from_cqe_linear(struct mlx5e_rq *rq,
 					      struct mlx5e_wqe_frag_info *wi,
 					      u32 cqe_bcnt);
 
+static inline int mlx5e_xsk_page_alloc_umem(struct mlx5e_rq *rq,
+					    struct mlx5e_dma_info *dma_info)
+{
+	dma_info->xsk = xsk_buff_alloc(rq->umem);
+	if (!dma_info->xsk)
+		return -ENOMEM;
+
+	/* Store the DMA address without headroom. In striding RQ case, we just
+	 * provide pages for UMR, and headroom is counted at the setup stage
+	 * when creating a WQE. In non-striding RQ case, headroom is accounted
+	 * in mlx5e_alloc_rx_wqe.
+	 */
+	dma_info->addr = xsk_buff_xdp_get_frame_dma(dma_info->xsk);
+
+	return 0;
+}
+
 static inline bool mlx5e_xsk_update_rx_wakeup(struct mlx5e_rq *rq, bool alloc_err)
 {
 	if (!xsk_umem_uses_need_wakeup(rq->umem))
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/tx.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/tx.c
index 3bcdb5b2fc20..83dce9cdb8c2 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/tx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/tx.c
@@ -5,7 +5,7 @@
 #include "umem.h"
 #include "en/xdp.h"
 #include "en/params.h"
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 
 int mlx5e_xsk_wakeup(struct net_device *dev, u32 qid, u32 flags)
 {
@@ -92,12 +92,11 @@ bool mlx5e_xsk_tx(struct mlx5e_xdpsq *sq, unsigned int budget)
 			break;
 		}
 
-		xdptxd.dma_addr = xdp_umem_get_dma(umem, desc.addr);
-		xdptxd.data = xdp_umem_get_data(umem, desc.addr);
+		xdptxd.dma_addr = xsk_buff_raw_get_dma(umem, desc.addr);
+		xdptxd.data = xsk_buff_raw_get_data(umem, desc.addr);
 		xdptxd.len = desc.len;
 
-		dma_sync_single_for_device(sq->pdev, xdptxd.dma_addr,
-					   xdptxd.len, DMA_BIDIRECTIONAL);
+		xsk_buff_raw_dma_sync_for_device(umem, xdptxd.dma_addr, xdptxd.len);
 
 		if (unlikely(!sq->xmit_xdp_frame(sq, &xdptxd, &xdpi, check_result))) {
 			if (sq->mpwqe.wqe)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/umem.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/umem.c
index 5e49fdb564b3..7b17fcd0a56d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/umem.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xsk/umem.c
@@ -10,40 +10,14 @@ static int mlx5e_xsk_map_umem(struct mlx5e_priv *priv,
 			      struct xdp_umem *umem)
 {
 	struct device *dev = priv->mdev->device;
-	u32 i;
 
-	for (i = 0; i < umem->npgs; i++) {
-		dma_addr_t dma = dma_map_page(dev, umem->pgs[i], 0, PAGE_SIZE,
-					      DMA_BIDIRECTIONAL);
-
-		if (unlikely(dma_mapping_error(dev, dma)))
-			goto err_unmap;
-		umem->pages[i].dma = dma;
-	}
-
-	return 0;
-
-err_unmap:
-	while (i--) {
-		dma_unmap_page(dev, umem->pages[i].dma, PAGE_SIZE,
-			       DMA_BIDIRECTIONAL);
-		umem->pages[i].dma = 0;
-	}
-
-	return -ENOMEM;
+	return xsk_buff_dma_map(umem, dev, 0);
 }
 
 static void mlx5e_xsk_unmap_umem(struct mlx5e_priv *priv,
 				 struct xdp_umem *umem)
 {
-	struct device *dev = priv->mdev->device;
-	u32 i;
-
-	for (i = 0; i < umem->npgs; i++) {
-		dma_unmap_page(dev, umem->pages[i].dma, PAGE_SIZE,
-			       DMA_BIDIRECTIONAL);
-		umem->pages[i].dma = 0;
-	}
+	return xsk_buff_dma_unmap(umem, 0);
 }
 
 static int mlx5e_xsk_get_umems(struct mlx5e_xsk *xsk)
@@ -90,13 +64,14 @@ static void mlx5e_xsk_remove_umem(struct mlx5e_xsk *xsk, u16 ix)
 
 static bool mlx5e_xsk_is_umem_sane(struct xdp_umem *umem)
 {
-	return umem->headroom <= 0xffff && umem->chunk_size_nohr <= 0xffff;
+	return xsk_umem_get_headroom(umem) <= 0xffff &&
+		xsk_umem_get_chunk_size(umem) <= 0xffff;
 }
 
 void mlx5e_build_xsk_param(struct xdp_umem *umem, struct mlx5e_xsk_param *xsk)
 {
-	xsk->headroom = umem->headroom;
-	xsk->chunk_size = umem->chunk_size_nohr + umem->headroom;
+	xsk->headroom = xsk_umem_get_headroom(umem);
+	xsk->chunk_size = xsk_umem_get_chunk_size(umem);
 }
 
 static int mlx5e_xsk_enable_locked(struct mlx5e_priv *priv,
@@ -241,18 +216,6 @@ int mlx5e_xsk_setup_umem(struct net_device *dev, struct xdp_umem *umem, u16 qid)
 		      mlx5e_xsk_disable_umem(priv, ix);
 }
 
-int mlx5e_xsk_resize_reuseq(struct xdp_umem *umem, u32 nentries)
-{
-	struct xdp_umem_fq_reuse *reuseq;
-
-	reuseq = xsk_reuseq_prepare(nentries);
-	if (unlikely(!reuseq))
-		return -ENOMEM;
-	xsk_reuseq_free(xsk_reuseq_swap(umem, reuseq));
-
-	return 0;
-}
-
 u16 mlx5e_xsk_first_unused_channel(struct mlx5e_params *params, struct mlx5e_xsk *xsk)
 {
 	u16 res = xsk->refcnt ? params->num_channels : 0;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
index 0e4ca08ddca9..4041132723a3 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
@@ -38,7 +38,7 @@
 #include <linux/bpf.h>
 #include <linux/if_bridge.h>
 #include <net/page_pool.h>
-#include <net/xdp_sock.h>
+#include <net/xdp_sock_drv.h>
 #include "eswitch.h"
 #include "en.h"
 #include "en/txrx.h"
@@ -374,7 +374,6 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 	struct mlx5_core_dev *mdev = c->mdev;
 	void *rqc = rqp->rqc;
 	void *rqc_wq = MLX5_ADDR_OF(rqc, rqc, wq);
-	u32 num_xsk_frames = 0;
 	u32 rq_xdp_ix;
 	u32 pool_size;
 	int wq_sz;
@@ -414,7 +413,6 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 
 	rq->buff.map_dir = rq->xdp_prog ? DMA_BIDIRECTIONAL : DMA_FROM_DEVICE;
 	rq->buff.headroom = mlx5e_get_rq_headroom(mdev, params, xsk);
-	rq->buff.umem_headroom = xsk ? xsk->headroom : 0;
 	pool_size = 1 << params->log_rq_mtu_frames;
 
 	switch (rq->wq_type) {
@@ -428,10 +426,6 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 
 		wq_sz = mlx5_wq_ll_get_size(&rq->mpwqe.wq);
 
-		if (xsk)
-			num_xsk_frames = wq_sz <<
-				mlx5e_mpwqe_get_log_num_strides(mdev, params, xsk);
-
 		pool_size = MLX5_MPWRQ_PAGES_PER_WQE <<
 			mlx5e_mpwqe_get_log_rq_size(params, xsk);
 
@@ -483,9 +477,6 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 
 		wq_sz = mlx5_wq_cyc_get_size(&rq->wqe.wq);
 
-		if (xsk)
-			num_xsk_frames = wq_sz << rq->wqe.info.log_num_frags;
-
 		rq->wqe.info = rqp->frags_info;
 		rq->buff.frame0_sz = rq->wqe.info.arr[0].frag_stride;
 
@@ -526,19 +517,9 @@ static int mlx5e_alloc_rq(struct mlx5e_channel *c,
 	}
 
 	if (xsk) {
-		rq->buff.frame0_sz = xsk_umem_xdp_frame_sz(umem);
-
-		err = mlx5e_xsk_resize_reuseq(umem, num_xsk_frames);
-		if (unlikely(err)) {
-			mlx5_core_err(mdev, "Unable to allocate the Reuse Ring for %u frames\n",
-				      num_xsk_frames);
-			goto err_free;
-		}
-
-		rq->zca.free = mlx5e_xsk_zca_free;
 		err = xdp_rxq_info_reg_mem_model(&rq->xdp_rxq,
-						 MEM_TYPE_ZERO_COPY,
-						 &rq->zca);
+						 MEM_TYPE_XSK_BUFF_POOL, NULL);
+		xsk_buff_set_rxq_info(rq->umem, &rq->xdp_rxq);
 	} else {
 		/* Create a page_pool and register it with rxq */
 		pp_params.order     = 0;
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
index 821f94beda7a..d7b24e8905f1 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/en_rx.c
@@ -300,7 +300,7 @@ static inline void mlx5e_page_release(struct mlx5e_rq *rq,
 		 * put into the Reuse Ring, because there is no way to return
 		 * the page to the userspace when the interface goes down.
 		 */
-		mlx5e_xsk_page_release(rq, dma_info);
+		xsk_buff_free(dma_info->xsk);
 	else
 		mlx5e_page_release_dynamic(rq, dma_info, recycle);
 }
@@ -385,7 +385,11 @@ static int mlx5e_alloc_rx_wqes(struct mlx5e_rq *rq, u16 ix, u8 wqe_bulk)
 	if (rq->umem) {
 		int pages_desired = wqe_bulk << rq->wqe.info.log_num_frags;
 
-		if (unlikely(!mlx5e_xsk_pages_enough_umem(rq, pages_desired)))
+		/* Check in advance that we have enough frames, instead of
+		 * allocating one-by-one, failing and moving frames to the
+		 * Reuse Ring.
+		 */
+		if (unlikely(!xsk_buff_can_alloc(rq->umem, pages_desired)))
 			return -ENOMEM;
 	}
 
@@ -480,8 +484,11 @@ static int mlx5e_alloc_rx_mpwqe(struct mlx5e_rq *rq, u16 ix)
 	int err;
 	int i;
 
+	/* Check in advance that we have enough frames, instead of allocating
+	 * one-by-one, failing and moving frames to the Reuse Ring.
+	 */
 	if (rq->umem &&
-	    unlikely(!mlx5e_xsk_pages_enough_umem(rq, MLX5_MPWRQ_PAGES_PER_WQE))) {
+	    unlikely(!xsk_buff_can_alloc(rq->umem, MLX5_MPWRQ_PAGES_PER_WQE))) {
 		err = -ENOMEM;
 		goto err;
 	}
@@ -1044,12 +1051,24 @@ struct sk_buff *mlx5e_build_linear_skb(struct mlx5e_rq *rq, void *va,
 	return skb;
 }
 
+static void mlx5e_fill_xdp_buff(struct mlx5e_rq *rq, void *va, u16 headroom,
+				u32 len, struct xdp_buff *xdp)
+{
+	xdp->data_hard_start = va;
+	xdp_set_data_meta_invalid(xdp);
+	xdp->data = va + headroom;
+	xdp->data_end = xdp->data + len;
+	xdp->rxq = &rq->xdp_rxq;
+	xdp->frame_sz = rq->buff.frame0_sz;
+}
+
 struct sk_buff *
 mlx5e_skb_from_cqe_linear(struct mlx5e_rq *rq, struct mlx5_cqe64 *cqe,
 			  struct mlx5e_wqe_frag_info *wi, u32 cqe_bcnt)
 {
 	struct mlx5e_dma_info *di = wi->di;
 	u16 rx_headroom = rq->buff.headroom;
+	struct xdp_buff xdp;
 	struct sk_buff *skb;
 	void *va, *data;
 	bool consumed;
@@ -1065,11 +1084,13 @@ mlx5e_skb_from_cqe_linear(struct mlx5e_rq *rq, struct mlx5_cqe64 *cqe,
 	prefetch(data);
 
 	rcu_read_lock();
-	consumed = mlx5e_xdp_handle(rq, di, va, &rx_headroom, &cqe_bcnt, false);
+	mlx5e_fill_xdp_buff(rq, va, rx_headroom, cqe_bcnt, &xdp);
+	consumed = mlx5e_xdp_handle(rq, di, &cqe_bcnt, &xdp);
 	rcu_read_unlock();
 	if (consumed)
 		return NULL; /* page/packet was consumed by XDP */
 
+	rx_headroom = xdp.data - xdp.data_hard_start;
 	frag_size = MLX5_SKB_FRAG_SZ(rx_headroom + cqe_bcnt);
 	skb = mlx5e_build_linear_skb(rq, va, frag_size, rx_headroom, cqe_bcnt);
 	if (unlikely(!skb))
@@ -1343,6 +1364,7 @@ mlx5e_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq, struct mlx5e_mpw_info *wi,
 	struct mlx5e_dma_info *di = &wi->umr.dma_info[page_idx];
 	u16 rx_headroom = rq->buff.headroom;
 	u32 cqe_bcnt32 = cqe_bcnt;
+	struct xdp_buff xdp;
 	struct sk_buff *skb;
 	void *va, *data;
 	u32 frag_size;
@@ -1364,7 +1386,8 @@ mlx5e_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq, struct mlx5e_mpw_info *wi,
 	prefetch(data);
 
 	rcu_read_lock();
-	consumed = mlx5e_xdp_handle(rq, di, va, &rx_headroom, &cqe_bcnt32, false);
+	mlx5e_fill_xdp_buff(rq, va, rx_headroom, cqe_bcnt32, &xdp);
+	consumed = mlx5e_xdp_handle(rq, di, &cqe_bcnt32, &xdp);
 	rcu_read_unlock();
 	if (consumed) {
 		if (__test_and_clear_bit(MLX5E_RQ_FLAG_XDP_XMIT, rq->flags))
@@ -1372,6 +1395,7 @@ mlx5e_skb_from_cqe_mpwrq_linear(struct mlx5e_rq *rq, struct mlx5e_mpw_info *wi,
 		return NULL; /* page/packet was consumed by XDP */
 	}
 
+	rx_headroom = xdp.data - xdp.data_hard_start;
 	frag_size = MLX5_SKB_FRAG_SZ(rx_headroom + cqe_bcnt32);
 	skb = mlx5e_build_linear_skb(rq, va, frag_size, rx_headroom, cqe_bcnt32);
 	if (unlikely(!skb))
-- 
2.25.1

[PATCH bpf-next v4 12/15] xsk: remove MEM_TYPE_ZERO_COPY and corresponding code

[Björn Töpel]

From: Björn Töpel <bjorn.topel@intel.com>

There are no users of MEM_TYPE_ZERO_COPY. Remove all corresponding
code, including the "handle" member of struct xdp_buff.

rfc->v1: Fixed spelling in commit message. (Björn)

Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
 drivers/net/hyperv/netvsc_bpf.c |   1 -
 include/net/xdp.h               |   9 +-
 include/net/xdp_sock.h          |  45 ----------
 include/net/xdp_sock_drv.h      | 149 --------------------------------
 include/trace/events/xdp.h      |   1 -
 net/core/xdp.c                  |  42 ++-------
 net/xdp/xdp_umem.c              |  56 +-----------
 net/xdp/xsk.c                   |  48 +---------
 net/xdp/xsk_buff_pool.c         |   7 ++
 net/xdp/xsk_queue.c             |  62 -------------
 net/xdp/xsk_queue.h             | 105 ----------------------
 11 files changed, 15 insertions(+), 510 deletions(-)

diff --git a/drivers/net/hyperv/netvsc_bpf.c b/drivers/net/hyperv/netvsc_bpf.c
index 1e0c024b0a93..8e4141552423 100644
--- a/drivers/net/hyperv/netvsc_bpf.c
+++ b/drivers/net/hyperv/netvsc_bpf.c
@@ -50,7 +50,6 @@ u32 netvsc_run_xdp(struct net_device *ndev, struct netvsc_channel *nvchan,
 	xdp->data_end = xdp->data + len;
 	xdp->rxq = &nvchan->xdp_rxq;
 	xdp->frame_sz = PAGE_SIZE;
-	xdp->handle = 0;
 
 	memcpy(xdp->data, data, len);
 
diff --git a/include/net/xdp.h b/include/net/xdp.h
index f432134c7c00..90f11760bd12 100644
--- a/include/net/xdp.h
+++ b/include/net/xdp.h
@@ -39,7 +39,6 @@ enum xdp_mem_type {
 	MEM_TYPE_PAGE_SHARED = 0, /* Split-page refcnt based model */
 	MEM_TYPE_PAGE_ORDER0,     /* Orig XDP full page model */
 	MEM_TYPE_PAGE_POOL,
-	MEM_TYPE_ZERO_COPY,
 	MEM_TYPE_XSK_BUFF_POOL,
 	MEM_TYPE_MAX,
 };
@@ -55,10 +54,6 @@ struct xdp_mem_info {
 
 struct page_pool;
 
-struct zero_copy_allocator {
-	void (*free)(struct zero_copy_allocator *zca, unsigned long handle);
-};
-
 struct xdp_rxq_info {
 	struct net_device *dev;
 	u32 queue_index;
@@ -71,7 +66,6 @@ struct xdp_buff {
 	void *data_end;
 	void *data_meta;
 	void *data_hard_start;
-	unsigned long handle;
 	struct xdp_rxq_info *rxq;
 	u32 frame_sz; /* frame size to deduce data_hard_end/reserved tailroom*/
 };
@@ -120,8 +114,7 @@ struct xdp_frame *convert_to_xdp_frame(struct xdp_buff *xdp)
 	int metasize;
 	int headroom;
 
-	if (xdp->rxq->mem.type == MEM_TYPE_ZERO_COPY ||
-	    xdp->rxq->mem.type == MEM_TYPE_XSK_BUFF_POOL)
+	if (xdp->rxq->mem.type == MEM_TYPE_XSK_BUFF_POOL)
 		return xdp_convert_zc_to_xdp_frame(xdp);
 
 	/* Assure headroom is available for storing info */
diff --git a/include/net/xdp_sock.h b/include/net/xdp_sock.h
index 6e7265f63c04..96bfc5f5f24e 100644
--- a/include/net/xdp_sock.h
+++ b/include/net/xdp_sock.h
@@ -17,26 +17,12 @@ struct net_device;
 struct xsk_queue;
 struct xdp_buff;
 
-struct xdp_umem_page {
-	void *addr;
-	dma_addr_t dma;
-};
-
-struct xdp_umem_fq_reuse {
-	u32 nentries;
-	u32 length;
-	u64 handles[];
-};
-
 struct xdp_umem {
 	struct xsk_queue *fq;
 	struct xsk_queue *cq;
 	struct xsk_buff_pool *pool;
-	struct xdp_umem_page *pages;
-	u64 chunk_mask;
 	u64 size;
 	u32 headroom;
-	u32 chunk_size_nohr;
 	u32 chunk_size;
 	struct user_struct *user;
 	refcount_t users;
@@ -48,7 +34,6 @@ struct xdp_umem {
 	u8 flags;
 	int id;
 	struct net_device *dev;
-	struct xdp_umem_fq_reuse *fq_reuse;
 	bool zc;
 	spinlock_t xsk_tx_list_lock;
 	struct list_head xsk_tx_list;
@@ -109,21 +94,6 @@ static inline struct xdp_sock *__xsk_map_lookup_elem(struct bpf_map *map,
 	return xs;
 }
 
-static inline u64 xsk_umem_extract_addr(u64 addr)
-{
-	return addr & XSK_UNALIGNED_BUF_ADDR_MASK;
-}
-
-static inline u64 xsk_umem_extract_offset(u64 addr)
-{
-	return addr >> XSK_UNALIGNED_BUF_OFFSET_SHIFT;
-}
-
-static inline u64 xsk_umem_add_offset_to_addr(u64 addr)
-{
-	return xsk_umem_extract_addr(addr) + xsk_umem_extract_offset(addr);
-}
-
 #else
 
 static inline int xsk_generic_rcv(struct xdp_sock *xs, struct xdp_buff *xdp)
@@ -146,21 +116,6 @@ static inline struct xdp_sock *__xsk_map_lookup_elem(struct bpf_map *map,
 	return NULL;
 }
 
-static inline u64 xsk_umem_extract_addr(u64 addr)
-{
-	return 0;
-}
-
-static inline u64 xsk_umem_extract_offset(u64 addr)
-{
-	return 0;
-}
-
-static inline u64 xsk_umem_add_offset_to_addr(u64 addr)
-{
-	return 0;
-}
-
 #endif /* CONFIG_XDP_SOCKETS */
 
 #endif /* _LINUX_XDP_SOCK_H */
diff --git a/include/net/xdp_sock_drv.h b/include/net/xdp_sock_drv.h
index 7752c8663d1b..ccf848f7efa4 100644
--- a/include/net/xdp_sock_drv.h
+++ b/include/net/xdp_sock_drv.h
@@ -11,16 +11,9 @@
 
 #ifdef CONFIG_XDP_SOCKETS
 
-bool xsk_umem_has_addrs(struct xdp_umem *umem, u32 cnt);
-bool xsk_umem_peek_addr(struct xdp_umem *umem, u64 *addr);
-void xsk_umem_release_addr(struct xdp_umem *umem);
 void xsk_umem_complete_tx(struct xdp_umem *umem, u32 nb_entries);
 bool xsk_umem_consume_tx(struct xdp_umem *umem, struct xdp_desc *desc);
 void xsk_umem_consume_tx_done(struct xdp_umem *umem);
-struct xdp_umem_fq_reuse *xsk_reuseq_prepare(u32 nentries);
-struct xdp_umem_fq_reuse *xsk_reuseq_swap(struct xdp_umem *umem,
-					  struct xdp_umem_fq_reuse *newq);
-void xsk_reuseq_free(struct xdp_umem_fq_reuse *rq);
 struct xdp_umem *xdp_get_umem_from_qid(struct net_device *dev, u16 queue_id);
 void xsk_set_rx_need_wakeup(struct xdp_umem *umem);
 void xsk_set_tx_need_wakeup(struct xdp_umem *umem);
@@ -28,80 +21,6 @@ void xsk_clear_rx_need_wakeup(struct xdp_umem *umem);
 void xsk_clear_tx_need_wakeup(struct xdp_umem *umem);
 bool xsk_umem_uses_need_wakeup(struct xdp_umem *umem);
 
-static inline char *xdp_umem_get_data(struct xdp_umem *umem, u64 addr)
-{
-	unsigned long page_addr;
-
-	addr = xsk_umem_add_offset_to_addr(addr);
-	page_addr = (unsigned long)umem->pages[addr >> PAGE_SHIFT].addr;
-
-	return (char *)(page_addr & PAGE_MASK) + (addr & ~PAGE_MASK);
-}
-
-static inline dma_addr_t xdp_umem_get_dma(struct xdp_umem *umem, u64 addr)
-{
-	addr = xsk_umem_add_offset_to_addr(addr);
-
-	return umem->pages[addr >> PAGE_SHIFT].dma + (addr & ~PAGE_MASK);
-}
-
-/* Reuse-queue aware version of FILL queue helpers */
-static inline bool xsk_umem_has_addrs_rq(struct xdp_umem *umem, u32 cnt)
-{
-	struct xdp_umem_fq_reuse *rq = umem->fq_reuse;
-
-	if (rq->length >= cnt)
-		return true;
-
-	return xsk_umem_has_addrs(umem, cnt - rq->length);
-}
-
-static inline bool xsk_umem_peek_addr_rq(struct xdp_umem *umem, u64 *addr)
-{
-	struct xdp_umem_fq_reuse *rq = umem->fq_reuse;
-
-	if (!rq->length)
-		return xsk_umem_peek_addr(umem, addr);
-
-	*addr = rq->handles[rq->length - 1];
-	return addr;
-}
-
-static inline void xsk_umem_release_addr_rq(struct xdp_umem *umem)
-{
-	struct xdp_umem_fq_reuse *rq = umem->fq_reuse;
-
-	if (!rq->length)
-		xsk_umem_release_addr(umem);
-	else
-		rq->length--;
-}
-
-static inline void xsk_umem_fq_reuse(struct xdp_umem *umem, u64 addr)
-{
-	struct xdp_umem_fq_reuse *rq = umem->fq_reuse;
-
-	rq->handles[rq->length++] = addr;
-}
-
-/* Handle the offset appropriately depending on aligned or unaligned mode.
- * For unaligned mode, we store the offset in the upper 16-bits of the address.
- * For aligned mode, we simply add the offset to the address.
- */
-static inline u64 xsk_umem_adjust_offset(struct xdp_umem *umem, u64 address,
-					 u64 offset)
-{
-	if (umem->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG)
-		return address + (offset << XSK_UNALIGNED_BUF_OFFSET_SHIFT);
-	else
-		return address + offset;
-}
-
-static inline u32 xsk_umem_xdp_frame_sz(struct xdp_umem *umem)
-{
-	return umem->chunk_size_nohr;
-}
-
 static inline u32 xsk_umem_get_headroom(struct xdp_umem *umem)
 {
 	return XDP_PACKET_HEADROOM + umem->headroom;
@@ -192,20 +111,6 @@ static inline void xsk_buff_raw_dma_sync_for_device(struct xdp_umem *umem,
 
 #else
 
-static inline bool xsk_umem_has_addrs(struct xdp_umem *umem, u32 cnt)
-{
-	return false;
-}
-
-static inline u64 *xsk_umem_peek_addr(struct xdp_umem *umem, u64 *addr)
-{
-	return NULL;
-}
-
-static inline void xsk_umem_release_addr(struct xdp_umem *umem)
-{
-}
-
 static inline void xsk_umem_complete_tx(struct xdp_umem *umem, u32 nb_entries)
 {
 }
@@ -220,55 +125,12 @@ static inline void xsk_umem_consume_tx_done(struct xdp_umem *umem)
 {
 }
 
-static inline struct xdp_umem_fq_reuse *xsk_reuseq_prepare(u32 nentries)
-{
-	return NULL;
-}
-
-static inline struct xdp_umem_fq_reuse *xsk_reuseq_swap(
-	struct xdp_umem *umem, struct xdp_umem_fq_reuse *newq)
-{
-	return NULL;
-}
-
-static inline void xsk_reuseq_free(struct xdp_umem_fq_reuse *rq)
-{
-}
-
 static inline struct xdp_umem *xdp_get_umem_from_qid(struct net_device *dev,
 						     u16 queue_id)
 {
 	return NULL;
 }
 
-static inline char *xdp_umem_get_data(struct xdp_umem *umem, u64 addr)
-{
-	return NULL;
-}
-
-static inline dma_addr_t xdp_umem_get_dma(struct xdp_umem *umem, u64 addr)
-{
-	return 0;
-}
-
-static inline bool xsk_umem_has_addrs_rq(struct xdp_umem *umem, u32 cnt)
-{
-	return false;
-}
-
-static inline u64 *xsk_umem_peek_addr_rq(struct xdp_umem *umem, u64 *addr)
-{
-	return NULL;
-}
-
-static inline void xsk_umem_release_addr_rq(struct xdp_umem *umem)
-{
-}
-
-static inline void xsk_umem_fq_reuse(struct xdp_umem *umem, u64 addr)
-{
-}
-
 static inline void xsk_set_rx_need_wakeup(struct xdp_umem *umem)
 {
 }
@@ -290,17 +152,6 @@ static inline bool xsk_umem_uses_need_wakeup(struct xdp_umem *umem)
 	return false;
 }
 
-static inline u64 xsk_umem_adjust_offset(struct xdp_umem *umem, u64 handle,
-					 u64 offset)
-{
-	return 0;
-}
-
-static inline u32 xsk_umem_xdp_frame_sz(struct xdp_umem *umem)
-{
-	return 0;
-}
-
 static inline u32 xsk_umem_get_headroom(struct xdp_umem *umem)
 {
 	return 0;
diff --git a/include/trace/events/xdp.h b/include/trace/events/xdp.h
index 48547a12fa27..b73d3e141323 100644
--- a/include/trace/events/xdp.h
+++ b/include/trace/events/xdp.h
@@ -287,7 +287,6 @@ TRACE_EVENT(xdp_devmap_xmit,
 	FN(PAGE_SHARED)		\
 	FN(PAGE_ORDER0)		\
 	FN(PAGE_POOL)		\
-	FN(ZERO_COPY)		\
 	FN(XSK_BUFF_POOL)
 
 #define __MEM_TYPE_TP_FN(x)	\
diff --git a/net/core/xdp.c b/net/core/xdp.c
index f0ce8b195193..a8c2f243367d 100644
--- a/net/core/xdp.c
+++ b/net/core/xdp.c
@@ -110,27 +110,6 @@ static void mem_allocator_disconnect(void *allocator)
 	mutex_unlock(&mem_id_lock);
 }
 
-static void mem_id_disconnect(int id)
-{
-	struct xdp_mem_allocator *xa;
-
-	mutex_lock(&mem_id_lock);
-
-	xa = rhashtable_lookup_fast(mem_id_ht, &id, mem_id_rht_params);
-	if (!xa) {
-		mutex_unlock(&mem_id_lock);
-		WARN(1, "Request remove non-existing id(%d), driver bug?", id);
-		return;
-	}
-
-	trace_mem_disconnect(xa);
-
-	if (!rhashtable_remove_fast(mem_id_ht, &xa->node, mem_id_rht_params))
-		call_rcu(&xa->rcu, __xdp_mem_allocator_rcu_free);
-
-	mutex_unlock(&mem_id_lock);
-}
-
 void xdp_rxq_info_unreg_mem_model(struct xdp_rxq_info *xdp_rxq)
 {
 	struct xdp_mem_allocator *xa;
@@ -144,9 +123,6 @@ void xdp_rxq_info_unreg_mem_model(struct xdp_rxq_info *xdp_rxq)
 	if (id == 0)
 		return;
 
-	if (xdp_rxq->mem.type == MEM_TYPE_ZERO_COPY)
-		return mem_id_disconnect(id);
-
 	if (xdp_rxq->mem.type == MEM_TYPE_PAGE_POOL) {
 		rcu_read_lock();
 		xa = rhashtable_lookup(mem_id_ht, &id, mem_id_rht_params);
@@ -302,7 +278,7 @@ int xdp_rxq_info_reg_mem_model(struct xdp_rxq_info *xdp_rxq,
 	xdp_rxq->mem.type = type;
 
 	if (!allocator) {
-		if (type == MEM_TYPE_PAGE_POOL || type == MEM_TYPE_ZERO_COPY)
+		if (type == MEM_TYPE_PAGE_POOL)
 			return -EINVAL; /* Setup time check page_pool req */
 		return 0;
 	}
@@ -362,7 +338,7 @@ EXPORT_SYMBOL_GPL(xdp_rxq_info_reg_mem_model);
  * of xdp_frames/pages in those cases.
  */
 static void __xdp_return(void *data, struct xdp_mem_info *mem, bool napi_direct,
-			 unsigned long handle, struct xdp_buff *xdp)
+			 struct xdp_buff *xdp)
 {
 	struct xdp_mem_allocator *xa;
 	struct page *page;
@@ -384,14 +360,6 @@ static void __xdp_return(void *data, struct xdp_mem_info *mem, bool napi_direct,
 		page = virt_to_page(data); /* Assumes order0 page*/
 		put_page(page);
 		break;
-	case MEM_TYPE_ZERO_COPY:
-		/* NB! Only valid from an xdp_buff! */
-		rcu_read_lock();
-		/* mem->id is valid, checked in xdp_rxq_info_reg_mem_model() */
-		xa = rhashtable_lookup(mem_id_ht, &mem->id, mem_id_rht_params);
-		xa->zc_alloc->free(xa->zc_alloc, handle);
-		rcu_read_unlock();
-		break;
 	case MEM_TYPE_XSK_BUFF_POOL:
 		/* NB! Only valid from an xdp_buff! */
 		xsk_buff_free(xdp);
@@ -404,19 +372,19 @@ static void __xdp_return(void *data, struct xdp_mem_info *mem, bool napi_direct,
 
 void xdp_return_frame(struct xdp_frame *xdpf)
 {
-	__xdp_return(xdpf->data, &xdpf->mem, false, 0, NULL);
+	__xdp_return(xdpf->data, &xdpf->mem, false, NULL);
 }
 EXPORT_SYMBOL_GPL(xdp_return_frame);
 
 void xdp_return_frame_rx_napi(struct xdp_frame *xdpf)
 {
-	__xdp_return(xdpf->data, &xdpf->mem, true, 0, NULL);
+	__xdp_return(xdpf->data, &xdpf->mem, true, NULL);
 }
 EXPORT_SYMBOL_GPL(xdp_return_frame_rx_napi);
 
 void xdp_return_buff(struct xdp_buff *xdp)
 {
-	__xdp_return(xdp->data, &xdp->rxq->mem, true, xdp->handle, xdp);
+	__xdp_return(xdp->data, &xdp->rxq->mem, true, xdp);
 }
 EXPORT_SYMBOL_GPL(xdp_return_buff);
 
diff --git a/net/xdp/xdp_umem.c b/net/xdp/xdp_umem.c
index 7f04688045d5..19e59d1a5e9f 100644
--- a/net/xdp/xdp_umem.c
+++ b/net/xdp/xdp_umem.c
@@ -179,37 +179,6 @@ void xdp_umem_clear_dev(struct xdp_umem *umem)
 	umem->zc = false;
 }
 
-static void xdp_umem_unmap_pages(struct xdp_umem *umem)
-{
-	unsigned int i;
-
-	for (i = 0; i < umem->npgs; i++)
-		if (PageHighMem(umem->pgs[i]))
-			vunmap(umem->pages[i].addr);
-}
-
-static int xdp_umem_map_pages(struct xdp_umem *umem)
-{
-	unsigned int i;
-	void *addr;
-
-	for (i = 0; i < umem->npgs; i++) {
-		if (PageHighMem(umem->pgs[i]))
-			addr = vmap(&umem->pgs[i], 1, VM_MAP, PAGE_KERNEL);
-		else
-			addr = page_address(umem->pgs[i]);
-
-		if (!addr) {
-			xdp_umem_unmap_pages(umem);
-			return -ENOMEM;
-		}
-
-		umem->pages[i].addr = addr;
-	}
-
-	return 0;
-}
-
 static void xdp_umem_unpin_pages(struct xdp_umem *umem)
 {
 	unpin_user_pages_dirty_lock(umem->pgs, umem->npgs, true);
@@ -244,14 +213,9 @@ static void xdp_umem_release(struct xdp_umem *umem)
 		umem->cq = NULL;
 	}
 
-	xsk_reuseq_destroy(umem);
 	xp_destroy(umem->pool);
-	xdp_umem_unmap_pages(umem);
 	xdp_umem_unpin_pages(umem);
 
-	kvfree(umem->pages);
-	umem->pages = NULL;
-
 	xdp_umem_unaccount_pages(umem);
 	kfree(umem);
 }
@@ -385,11 +349,8 @@ static int xdp_umem_reg(struct xdp_umem *umem, struct xdp_umem_reg *mr)
 	if (headroom >= chunk_size - XDP_PACKET_HEADROOM)
 		return -EINVAL;
 
-	umem->chunk_mask = unaligned_chunks ? XSK_UNALIGNED_BUF_ADDR_MASK
-					    : ~((u64)chunk_size - 1);
 	umem->size = size;
 	umem->headroom = headroom;
-	umem->chunk_size_nohr = chunk_size - headroom;
 	umem->chunk_size = chunk_size;
 	umem->npgs = size / PAGE_SIZE;
 	umem->pgs = NULL;
@@ -408,29 +369,14 @@ static int xdp_umem_reg(struct xdp_umem *umem, struct xdp_umem_reg *mr)
 	if (err)
 		goto out_account;
 
-	umem->pages = kvcalloc(umem->npgs, sizeof(*umem->pages),
-			       GFP_KERNEL_ACCOUNT);
-	if (!umem->pages) {
-		err = -ENOMEM;
-		goto out_pin;
-	}
-
-	err = xdp_umem_map_pages(umem);
-	if (err)
-		goto out_pages;
-
 	umem->pool = xp_create(umem->pgs, umem->npgs, chunks, chunk_size,
 			       headroom, size, unaligned_chunks);
 	if (!umem->pool) {
 		err = -ENOMEM;
-		goto out_unmap;
+		goto out_pin;
 	}
 	return 0;
 
-out_unmap:
-	xdp_umem_unmap_pages(umem);
-out_pages:
-	kvfree(umem->pages);
 out_pin:
 	xdp_umem_unpin_pages(umem);
 out_account:
diff --git a/net/xdp/xsk.c b/net/xdp/xsk.c
index 6933f0d494ba..3f2ab732ab8b 100644
--- a/net/xdp/xsk.c
+++ b/net/xdp/xsk.c
@@ -39,24 +39,6 @@ bool xsk_is_setup_for_bpf_map(struct xdp_sock *xs)
 		READ_ONCE(xs->umem->fq);
 }
 
-bool xsk_umem_has_addrs(struct xdp_umem *umem, u32 cnt)
-{
-	return xskq_cons_has_entries(umem->fq, cnt);
-}
-EXPORT_SYMBOL(xsk_umem_has_addrs);
-
-bool xsk_umem_peek_addr(struct xdp_umem *umem, u64 *addr)
-{
-	return xskq_cons_peek_addr(umem->fq, addr, umem);
-}
-EXPORT_SYMBOL(xsk_umem_peek_addr);
-
-void xsk_umem_release_addr(struct xdp_umem *umem)
-{
-	xskq_cons_release(umem->fq);
-}
-EXPORT_SYMBOL(xsk_umem_release_addr);
-
 void xsk_set_rx_need_wakeup(struct xdp_umem *umem)
 {
 	if (umem->need_wakeup & XDP_WAKEUP_RX)
@@ -203,8 +185,7 @@ static int xsk_rcv(struct xdp_sock *xs, struct xdp_buff *xdp,
 
 	len = xdp->data_end - xdp->data;
 
-	return xdp->rxq->mem.type == MEM_TYPE_ZERO_COPY ||
-		xdp->rxq->mem.type == MEM_TYPE_XSK_BUFF_POOL ?
+	return xdp->rxq->mem.type == MEM_TYPE_XSK_BUFF_POOL ?
 		__xsk_rcv_zc(xs, xdp, len) :
 		__xsk_rcv(xs, xdp, len, explicit_free);
 }
@@ -588,24 +569,6 @@ static struct socket *xsk_lookup_xsk_from_fd(int fd)
 	return sock;
 }
 
-/* Check if umem pages are contiguous.
- * If zero-copy mode, use the DMA address to do the page contiguity check
- * For all other modes we use addr (kernel virtual address)
- * Store the result in the low bits of addr.
- */
-static void xsk_check_page_contiguity(struct xdp_umem *umem, u32 flags)
-{
-	struct xdp_umem_page *pgs = umem->pages;
-	int i, is_contig;
-
-	for (i = 0; i < umem->npgs - 1; i++) {
-		is_contig = (flags & XDP_ZEROCOPY) ?
-			(pgs[i].dma + PAGE_SIZE == pgs[i + 1].dma) :
-			(pgs[i].addr + PAGE_SIZE == pgs[i + 1].addr);
-		pgs[i].addr += is_contig << XSK_NEXT_PG_CONTIG_SHIFT;
-	}
-}
-
 static int xsk_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
 {
 	struct sockaddr_xdp *sxdp = (struct sockaddr_xdp *)addr;
@@ -688,23 +651,14 @@ static int xsk_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
 		goto out_unlock;
 	} else {
 		/* This xsk has its own umem. */
-		xskq_set_umem(xs->umem->fq, xs->umem->size,
-			      xs->umem->chunk_mask);
-		xskq_set_umem(xs->umem->cq, xs->umem->size,
-			      xs->umem->chunk_mask);
-
 		err = xdp_umem_assign_dev(xs->umem, dev, qid, flags);
 		if (err)
 			goto out_unlock;
-
-		xsk_check_page_contiguity(xs->umem, flags);
 	}
 
 	xs->dev = dev;
 	xs->zc = xs->umem->zc;
 	xs->queue_id = qid;
-	xskq_set_umem(xs->rx, xs->umem->size, xs->umem->chunk_mask);
-	xskq_set_umem(xs->tx, xs->umem->size, xs->umem->chunk_mask);
 	xdp_add_sk_umem(xs->umem, xs);
 
 out_unlock:
diff --git a/net/xdp/xsk_buff_pool.c b/net/xdp/xsk_buff_pool.c
index e214a5795a62..89dae78865e7 100644
--- a/net/xdp/xsk_buff_pool.c
+++ b/net/xdp/xsk_buff_pool.c
@@ -8,6 +8,13 @@
 
 #include "xsk_queue.h"
 
+/* Masks for xdp_umem_page flags.
+ * The low 12-bits of the addr will be 0 since this is the page address, so we
+ * can use them for flags.
+ */
+#define XSK_NEXT_PG_CONTIG_SHIFT 0
+#define XSK_NEXT_PG_CONTIG_MASK BIT_ULL(XSK_NEXT_PG_CONTIG_SHIFT)
+
 struct xsk_buff_pool {
 	struct xsk_queue *fq;
 	struct list_head free_list;
diff --git a/net/xdp/xsk_queue.c b/net/xdp/xsk_queue.c
index 57fb81bd593c..c33b5d985b9b 100644
--- a/net/xdp/xsk_queue.c
+++ b/net/xdp/xsk_queue.c
@@ -9,15 +9,6 @@
 
 #include "xsk_queue.h"
 
-void xskq_set_umem(struct xsk_queue *q, u64 umem_size, u64 chunk_mask)
-{
-	if (!q)
-		return;
-
-	q->umem_size = umem_size;
-	q->chunk_mask = chunk_mask;
-}
-
 static size_t xskq_get_ring_size(struct xsk_queue *q, bool umem_queue)
 {
 	struct xdp_umem_ring *umem_ring;
@@ -63,56 +54,3 @@ void xskq_destroy(struct xsk_queue *q)
 	page_frag_free(q->ring);
 	kfree(q);
 }
-
-struct xdp_umem_fq_reuse *xsk_reuseq_prepare(u32 nentries)
-{
-	struct xdp_umem_fq_reuse *newq;
-
-	/* Check for overflow */
-	if (nentries > (u32)roundup_pow_of_two(nentries))
-		return NULL;
-	nentries = roundup_pow_of_two(nentries);
-
-	newq = kvmalloc(struct_size(newq, handles, nentries), GFP_KERNEL);
-	if (!newq)
-		return NULL;
-	memset(newq, 0, offsetof(typeof(*newq), handles));
-
-	newq->nentries = nentries;
-	return newq;
-}
-EXPORT_SYMBOL_GPL(xsk_reuseq_prepare);
-
-struct xdp_umem_fq_reuse *xsk_reuseq_swap(struct xdp_umem *umem,
-					  struct xdp_umem_fq_reuse *newq)
-{
-	struct xdp_umem_fq_reuse *oldq = umem->fq_reuse;
-
-	if (!oldq) {
-		umem->fq_reuse = newq;
-		return NULL;
-	}
-
-	if (newq->nentries < oldq->length)
-		return newq;
-
-	memcpy(newq->handles, oldq->handles,
-	       array_size(oldq->length, sizeof(u64)));
-	newq->length = oldq->length;
-
-	umem->fq_reuse = newq;
-	return oldq;
-}
-EXPORT_SYMBOL_GPL(xsk_reuseq_swap);
-
-void xsk_reuseq_free(struct xdp_umem_fq_reuse *rq)
-{
-	kvfree(rq);
-}
-EXPORT_SYMBOL_GPL(xsk_reuseq_free);
-
-void xsk_reuseq_destroy(struct xdp_umem *umem)
-{
-	xsk_reuseq_free(umem->fq_reuse);
-	umem->fq_reuse = NULL;
-}
diff --git a/net/xdp/xsk_queue.h b/net/xdp/xsk_queue.h
index 9151aef7dbca..16bf15864788 100644
--- a/net/xdp/xsk_queue.h
+++ b/net/xdp/xsk_queue.h
@@ -32,8 +32,6 @@ struct xdp_umem_ring {
 };
 
 struct xsk_queue {
-	u64 chunk_mask;
-	u64 umem_size;
 	u32 ring_mask;
 	u32 nentries;
 	u32 cached_prod;
@@ -106,90 +104,6 @@ struct xsk_queue {
 
 /* Functions that read and validate content from consumer rings. */
 
-static inline bool xskq_cons_crosses_non_contig_pg(struct xdp_umem *umem,
-						   u64 addr,
-						   u64 length)
-{
-	bool cross_pg = (addr & (PAGE_SIZE - 1)) + length > PAGE_SIZE;
-	bool next_pg_contig =
-		(unsigned long)umem->pages[(addr >> PAGE_SHIFT)].addr &
-			XSK_NEXT_PG_CONTIG_MASK;
-
-	return cross_pg && !next_pg_contig;
-}
-
-static inline bool xskq_cons_is_valid_unaligned(struct xsk_queue *q,
-						u64 addr,
-						u64 length,
-						struct xdp_umem *umem)
-{
-	u64 base_addr = xsk_umem_extract_addr(addr);
-
-	addr = xsk_umem_add_offset_to_addr(addr);
-	if (base_addr >= q->umem_size || addr >= q->umem_size ||
-	    xskq_cons_crosses_non_contig_pg(umem, addr, length)) {
-		q->invalid_descs++;
-		return false;
-	}
-
-	return true;
-}
-
-static inline bool xskq_cons_is_valid_addr(struct xsk_queue *q, u64 addr)
-{
-	if (addr >= q->umem_size) {
-		q->invalid_descs++;
-		return false;
-	}
-
-	return true;
-}
-
-static inline bool xskq_cons_read_addr(struct xsk_queue *q, u64 *addr,
-				       struct xdp_umem *umem)
-{
-	struct xdp_umem_ring *ring = (struct xdp_umem_ring *)q->ring;
-
-	while (q->cached_cons != q->cached_prod) {
-		u32 idx = q->cached_cons & q->ring_mask;
-
-		*addr = ring->desc[idx] & q->chunk_mask;
-
-		if (umem->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG) {
-			if (xskq_cons_is_valid_unaligned(q, *addr,
-							 umem->chunk_size_nohr,
-							 umem))
-				return true;
-			goto out;
-		}
-
-		if (xskq_cons_is_valid_addr(q, *addr))
-			return true;
-
-out:
-		q->cached_cons++;
-	}
-
-	return false;
-}
-
-static inline bool xskq_cons_read_addr_aligned(struct xsk_queue *q, u64 *addr)
-{
-	struct xdp_umem_ring *ring = (struct xdp_umem_ring *)q->ring;
-
-	while (q->cached_cons != q->cached_prod) {
-		u32 idx = q->cached_cons & q->ring_mask;
-
-		*addr = ring->desc[idx];
-		if (xskq_cons_is_valid_addr(q, *addr))
-			return true;
-
-		q->cached_cons++;
-	}
-
-	return false;
-}
-
 static inline bool xskq_cons_read_addr_unchecked(struct xsk_queue *q, u64 *addr)
 {
 	struct xdp_umem_ring *ring = (struct xdp_umem_ring *)q->ring;
@@ -267,21 +181,6 @@ static inline bool xskq_cons_has_entries(struct xsk_queue *q, u32 cnt)
 	return entries >= cnt;
 }
 
-static inline bool xskq_cons_peek_addr(struct xsk_queue *q, u64 *addr,
-				       struct xdp_umem *umem)
-{
-	if (q->cached_prod == q->cached_cons)
-		xskq_cons_get_entries(q);
-	return xskq_cons_read_addr(q, addr, umem);
-}
-
-static inline bool xskq_cons_peek_addr_aligned(struct xsk_queue *q, u64 *addr)
-{
-	if (q->cached_prod == q->cached_cons)
-		xskq_cons_get_entries(q);
-	return xskq_cons_read_addr_aligned(q, addr);
-}
-
 static inline bool xskq_cons_peek_addr_unchecked(struct xsk_queue *q, u64 *addr)
 {
 	if (q->cached_prod == q->cached_cons)
@@ -410,11 +309,7 @@ static inline u64 xskq_nb_invalid_descs(struct xsk_queue *q)
 	return q ? q->invalid_descs : 0;
 }
 
-void xskq_set_umem(struct xsk_queue *q, u64 umem_size, u64 chunk_mask);
 struct xsk_queue *xskq_create(u32 nentries, bool umem_queue);
 void xskq_destroy(struct xsk_queue *q_ops);
 
-/* Executed by the core when the entire UMEM gets freed */
-void xsk_reuseq_destroy(struct xdp_umem *umem);
-
 #endif /* _LINUX_XSK_QUEUE_H */
-- 
2.25.1

[PATCH bpf-next v4 13/15] xdp: simplify xdp_return_{frame,frame_rx_napi,buff}

[Björn Töpel]

From: Björn Töpel <bjorn.topel@intel.com>

The xdp_return_{frame,frame_rx_napi,buff} function are never used,
except in xdp_convert_zc_to_xdp_frame(), by the MEM_TYPE_XSK_BUFF_POOL
memory type.

To simplify and reduce code, change so that
xdp_convert_zc_to_xdp_frame() calls xsk_buff_free() directly since the
type is know, and remove MEM_TYPE_XSK_BUFF_POOL from the switch
statement in __xdp_return() function.

Suggested-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
 net/core/xdp.c | 21 +++++++++------------
 1 file changed, 9 insertions(+), 12 deletions(-)

diff --git a/net/core/xdp.c b/net/core/xdp.c
index a8c2f243367d..90f44f382115 100644
--- a/net/core/xdp.c
+++ b/net/core/xdp.c
@@ -335,10 +335,11 @@ EXPORT_SYMBOL_GPL(xdp_rxq_info_reg_mem_model);
  * scenarios (e.g. queue full), it is possible to return the xdp_frame
  * while still leveraging this protection.  The @napi_direct boolean
  * is used for those calls sites.  Thus, allowing for faster recycling
- * of xdp_frames/pages in those cases.
+ * of xdp_frames/pages in those cases. This path is never used by the
+ * MEM_TYPE_XSK_BUFF_POOL memory type, so it's explicitly not part of
+ * the switch-statement.
  */
-static void __xdp_return(void *data, struct xdp_mem_info *mem, bool napi_direct,
-			 struct xdp_buff *xdp)
+static void __xdp_return(void *data, struct xdp_mem_info *mem, bool napi_direct)
 {
 	struct xdp_mem_allocator *xa;
 	struct page *page;
@@ -360,33 +361,29 @@ static void __xdp_return(void *data, struct xdp_mem_info *mem, bool napi_direct,
 		page = virt_to_page(data); /* Assumes order0 page*/
 		put_page(page);
 		break;
-	case MEM_TYPE_XSK_BUFF_POOL:
-		/* NB! Only valid from an xdp_buff! */
-		xsk_buff_free(xdp);
-		break;
 	default:
 		/* Not possible, checked in xdp_rxq_info_reg_mem_model() */
+		WARN(1, "Incorrect XDP memory type (%d) usage", mem->type);
 		break;
 	}
 }
 
 void xdp_return_frame(struct xdp_frame *xdpf)
 {
-	__xdp_return(xdpf->data, &xdpf->mem, false, NULL);
+	__xdp_return(xdpf->data, &xdpf->mem, false);
 }
 EXPORT_SYMBOL_GPL(xdp_return_frame);
 
 void xdp_return_frame_rx_napi(struct xdp_frame *xdpf)
 {
-	__xdp_return(xdpf->data, &xdpf->mem, true, NULL);
+	__xdp_return(xdpf->data, &xdpf->mem, true);
 }
 EXPORT_SYMBOL_GPL(xdp_return_frame_rx_napi);
 
 void xdp_return_buff(struct xdp_buff *xdp)
 {
-	__xdp_return(xdp->data, &xdp->rxq->mem, true, xdp);
+	__xdp_return(xdp->data, &xdp->rxq->mem, true);
 }
-EXPORT_SYMBOL_GPL(xdp_return_buff);
 
 /* Only called for MEM_TYPE_PAGE_POOL see xdp.h */
 void __xdp_release_frame(void *data, struct xdp_mem_info *mem)
@@ -467,7 +464,7 @@ struct xdp_frame *xdp_convert_zc_to_xdp_frame(struct xdp_buff *xdp)
 	xdpf->metasize = metasize;
 	xdpf->mem.type = MEM_TYPE_PAGE_ORDER0;
 
-	xdp_return_buff(xdp);
+	xsk_buff_free(xdp);
 	return xdpf;
 }
 EXPORT_SYMBOL_GPL(xdp_convert_zc_to_xdp_frame);
-- 
2.25.1

[PATCH bpf-next v4 14/15] xsk: explicitly inline functions and move definitions

[Björn Töpel]

From: Björn Töpel <bjorn.topel@intel.com>

In order to reduce the number of function calls, the struct
xsk_buff_pool definition is moved to xsk_buff_pool.h. The functions
xp_get_dma(), xp_dma_sync_for_cpu(), xp_dma_sync_for_device(),
xp_validate_desc() and various helper functions are explicitly
inlined.

Further, move xp_get_handle() and xp_release() to xsk.c, to allow for
the compiler to perform inlining.

rfc->v1: Make sure xp_validate_desc() is inlined for Tx perf. (Maxim)

Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
 include/net/xsk_buff_pool.h |  98 ++++++++++++++++++++++--
 net/xdp/xsk.c               |  15 ++++
 net/xdp/xsk_buff_pool.c     | 148 ++----------------------------------
 net/xdp/xsk_queue.h         |  45 +++++++++++
 4 files changed, 156 insertions(+), 150 deletions(-)

diff --git a/include/net/xsk_buff_pool.h b/include/net/xsk_buff_pool.h
index 9f221b36e405..a4ff226505c9 100644
--- a/include/net/xsk_buff_pool.h
+++ b/include/net/xsk_buff_pool.h
@@ -4,6 +4,7 @@
 #ifndef XSK_BUFF_POOL_H_
 #define XSK_BUFF_POOL_H_
 
+#include <linux/if_xdp.h>
 #include <linux/types.h>
 #include <linux/dma-mapping.h>
 #include <net/xdp.h>
@@ -25,6 +26,27 @@ struct xdp_buff_xsk {
 	struct list_head free_list_node;
 };
 
+struct xsk_buff_pool {
+	struct xsk_queue *fq;
+	struct list_head free_list;
+	dma_addr_t *dma_pages;
+	struct xdp_buff_xsk *heads;
+	u64 chunk_mask;
+	u64 addrs_cnt;
+	u32 free_list_cnt;
+	u32 dma_pages_cnt;
+	u32 heads_cnt;
+	u32 free_heads_cnt;
+	u32 headroom;
+	u32 chunk_size;
+	u32 frame_len;
+	bool cheap_dma;
+	bool unaligned;
+	void *addrs;
+	struct device *dev;
+	struct xdp_buff_xsk *free_heads[];
+};
+
 /* AF_XDP core. */
 struct xsk_buff_pool *xp_create(struct page **pages, u32 nr_pages, u32 chunks,
 				u32 chunk_size, u32 headroom, u64 size,
@@ -32,8 +54,6 @@ struct xsk_buff_pool *xp_create(struct page **pages, u32 nr_pages, u32 chunks,
 void xp_set_fq(struct xsk_buff_pool *pool, struct xsk_queue *fq);
 void xp_destroy(struct xsk_buff_pool *pool);
 void xp_release(struct xdp_buff_xsk *xskb);
-u64 xp_get_handle(struct xdp_buff_xsk *xskb);
-bool xp_validate_desc(struct xsk_buff_pool *pool, struct xdp_desc *desc);
 
 /* AF_XDP, and XDP core. */
 void xp_free(struct xdp_buff_xsk *xskb);
@@ -47,10 +67,74 @@ struct xdp_buff *xp_alloc(struct xsk_buff_pool *pool);
 bool xp_can_alloc(struct xsk_buff_pool *pool, u32 count);
 void *xp_raw_get_data(struct xsk_buff_pool *pool, u64 addr);
 dma_addr_t xp_raw_get_dma(struct xsk_buff_pool *pool, u64 addr);
-dma_addr_t xp_get_dma(struct xdp_buff_xsk *xskb);
-dma_addr_t xp_get_frame_dma(struct xdp_buff_xsk *xskb);
-void xp_dma_sync_for_cpu(struct xdp_buff_xsk *xskb);
-void xp_dma_sync_for_device(struct xsk_buff_pool *pool, dma_addr_t dma,
-			    size_t size);
+static inline dma_addr_t xp_get_dma(struct xdp_buff_xsk *xskb)
+{
+	return xskb->dma;
+}
+
+static inline dma_addr_t xp_get_frame_dma(struct xdp_buff_xsk *xskb)
+{
+	return xskb->frame_dma;
+}
+
+void xp_dma_sync_for_cpu_slow(struct xdp_buff_xsk *xskb);
+static inline void xp_dma_sync_for_cpu(struct xdp_buff_xsk *xskb)
+{
+	if (xskb->pool->cheap_dma)
+		return;
+
+	xp_dma_sync_for_cpu_slow(xskb);
+}
+
+void xp_dma_sync_for_device_slow(struct xsk_buff_pool *pool, dma_addr_t dma,
+				 size_t size);
+static inline void xp_dma_sync_for_device(struct xsk_buff_pool *pool,
+					  dma_addr_t dma, size_t size)
+{
+	if (pool->cheap_dma)
+		return;
+
+	xp_dma_sync_for_device_slow(pool, dma, size);
+}
+
+/* Masks for xdp_umem_page flags.
+ * The low 12-bits of the addr will be 0 since this is the page address, so we
+ * can use them for flags.
+ */
+#define XSK_NEXT_PG_CONTIG_SHIFT 0
+#define XSK_NEXT_PG_CONTIG_MASK BIT_ULL(XSK_NEXT_PG_CONTIG_SHIFT)
+
+static inline bool xp_desc_crosses_non_contig_pg(struct xsk_buff_pool *pool,
+						 u64 addr, u32 len)
+{
+	bool cross_pg = (addr & (PAGE_SIZE - 1)) + len > PAGE_SIZE;
+
+	if (pool->dma_pages_cnt && cross_pg) {
+		return !(pool->dma_pages[addr >> PAGE_SHIFT] &
+			 XSK_NEXT_PG_CONTIG_MASK);
+	}
+	return false;
+}
+
+static inline u64 xp_aligned_extract_addr(struct xsk_buff_pool *pool, u64 addr)
+{
+	return addr & pool->chunk_mask;
+}
+
+static inline u64 xp_unaligned_extract_addr(u64 addr)
+{
+	return addr & XSK_UNALIGNED_BUF_ADDR_MASK;
+}
+
+static inline u64 xp_unaligned_extract_offset(u64 addr)
+{
+	return addr >> XSK_UNALIGNED_BUF_OFFSET_SHIFT;
+}
+
+static inline u64 xp_unaligned_add_offset_to_addr(u64 addr)
+{
+	return xp_unaligned_extract_addr(addr) +
+		xp_unaligned_extract_offset(addr);
+}
 
 #endif /* XSK_BUFF_POOL_H_ */
diff --git a/net/xdp/xsk.c b/net/xdp/xsk.c
index 3f2ab732ab8b..b6c0f08bd80d 100644
--- a/net/xdp/xsk.c
+++ b/net/xdp/xsk.c
@@ -99,6 +99,21 @@ bool xsk_umem_uses_need_wakeup(struct xdp_umem *umem)
 }
 EXPORT_SYMBOL(xsk_umem_uses_need_wakeup);
 
+void xp_release(struct xdp_buff_xsk *xskb)
+{
+	xskb->pool->free_heads[xskb->pool->free_heads_cnt++] = xskb;
+}
+
+static u64 xp_get_handle(struct xdp_buff_xsk *xskb)
+{
+	u64 offset = xskb->xdp.data - xskb->xdp.data_hard_start;
+
+	offset += xskb->pool->headroom;
+	if (!xskb->pool->unaligned)
+		return xskb->orig_addr + offset;
+	return xskb->orig_addr + (offset << XSK_UNALIGNED_BUF_OFFSET_SHIFT);
+}
+
 static int __xsk_rcv_zc(struct xdp_sock *xs, struct xdp_buff *xdp, u32 len)
 {
 	struct xdp_buff_xsk *xskb = container_of(xdp, struct xdp_buff_xsk, xdp);
diff --git a/net/xdp/xsk_buff_pool.c b/net/xdp/xsk_buff_pool.c
index 89dae78865e7..540ed75e4482 100644
--- a/net/xdp/xsk_buff_pool.c
+++ b/net/xdp/xsk_buff_pool.c
@@ -8,34 +8,6 @@
 
 #include "xsk_queue.h"
 
-/* Masks for xdp_umem_page flags.
- * The low 12-bits of the addr will be 0 since this is the page address, so we
- * can use them for flags.
- */
-#define XSK_NEXT_PG_CONTIG_SHIFT 0
-#define XSK_NEXT_PG_CONTIG_MASK BIT_ULL(XSK_NEXT_PG_CONTIG_SHIFT)
-
-struct xsk_buff_pool {
-	struct xsk_queue *fq;
-	struct list_head free_list;
-	dma_addr_t *dma_pages;
-	struct xdp_buff_xsk *heads;
-	u64 chunk_mask;
-	u64 addrs_cnt;
-	u32 free_list_cnt;
-	u32 dma_pages_cnt;
-	u32 heads_cnt;
-	u32 free_heads_cnt;
-	u32 headroom;
-	u32 chunk_size;
-	u32 frame_len;
-	bool cheap_dma;
-	bool unaligned;
-	void *addrs;
-	struct device *dev;
-	struct xdp_buff_xsk *free_heads[];
-};
-
 static void xp_addr_unmap(struct xsk_buff_pool *pool)
 {
 	vunmap(pool->addrs);
@@ -228,50 +200,12 @@ int xp_dma_map(struct xsk_buff_pool *pool, struct device *dev,
 }
 EXPORT_SYMBOL(xp_dma_map);
 
-static bool xp_desc_crosses_non_contig_pg(struct xsk_buff_pool *pool,
-					  u64 addr, u32 len)
-{
-	bool cross_pg = (addr & (PAGE_SIZE - 1)) + len > PAGE_SIZE;
-
-	if (pool->dma_pages_cnt && cross_pg) {
-		return !(pool->dma_pages[addr >> PAGE_SHIFT] &
-			 XSK_NEXT_PG_CONTIG_MASK);
-	}
-	return false;
-}
-
 static bool xp_addr_crosses_non_contig_pg(struct xsk_buff_pool *pool,
 					  u64 addr)
 {
 	return xp_desc_crosses_non_contig_pg(pool, addr, pool->chunk_size);
 }
 
-void xp_release(struct xdp_buff_xsk *xskb)
-{
-	xskb->pool->free_heads[xskb->pool->free_heads_cnt++] = xskb;
-}
-
-static u64 xp_aligned_extract_addr(struct xsk_buff_pool *pool, u64 addr)
-{
-	return addr & pool->chunk_mask;
-}
-
-static u64 xp_unaligned_extract_addr(u64 addr)
-{
-	return addr & XSK_UNALIGNED_BUF_ADDR_MASK;
-}
-
-static u64 xp_unaligned_extract_offset(u64 addr)
-{
-	return addr >> XSK_UNALIGNED_BUF_OFFSET_SHIFT;
-}
-
-static u64 xp_unaligned_add_offset_to_addr(u64 addr)
-{
-	return xp_unaligned_extract_addr(addr) +
-		xp_unaligned_extract_offset(addr);
-}
-
 static bool xp_check_unaligned(struct xsk_buff_pool *pool, u64 *addr)
 {
 	*addr = xp_unaligned_extract_addr(*addr);
@@ -370,60 +304,6 @@ void xp_free(struct xdp_buff_xsk *xskb)
 }
 EXPORT_SYMBOL(xp_free);
 
-static bool xp_aligned_validate_desc(struct xsk_buff_pool *pool,
-				     struct xdp_desc *desc)
-{
-	u64 chunk, chunk_end;
-
-	chunk = xp_aligned_extract_addr(pool, desc->addr);
-	chunk_end = xp_aligned_extract_addr(pool, desc->addr + desc->len);
-	if (chunk != chunk_end)
-		return false;
-
-	if (chunk >= pool->addrs_cnt)
-		return false;
-
-	if (desc->options)
-		return false;
-	return true;
-}
-
-static bool xp_unaligned_validate_desc(struct xsk_buff_pool *pool,
-				       struct xdp_desc *desc)
-{
-	u64 addr, base_addr;
-
-	base_addr = xp_unaligned_extract_addr(desc->addr);
-	addr = xp_unaligned_add_offset_to_addr(desc->addr);
-
-	if (desc->len > pool->chunk_size)
-		return false;
-
-	if (base_addr >= pool->addrs_cnt || addr >= pool->addrs_cnt ||
-	    xp_desc_crosses_non_contig_pg(pool, addr, desc->len))
-		return false;
-
-	if (desc->options)
-		return false;
-	return true;
-}
-
-bool xp_validate_desc(struct xsk_buff_pool *pool, struct xdp_desc *desc)
-{
-	return pool->unaligned ? xp_unaligned_validate_desc(pool, desc) :
-		xp_aligned_validate_desc(pool, desc);
-}
-
-u64 xp_get_handle(struct xdp_buff_xsk *xskb)
-{
-	u64 offset = xskb->xdp.data - xskb->xdp.data_hard_start;
-
-	offset += xskb->pool->headroom;
-	if (!xskb->pool->unaligned)
-		return xskb->orig_addr + offset;
-	return xskb->orig_addr + (offset << XSK_UNALIGNED_BUF_OFFSET_SHIFT);
-}
-
 void *xp_raw_get_data(struct xsk_buff_pool *pool, u64 addr)
 {
 	addr = pool->unaligned ? xp_unaligned_add_offset_to_addr(addr) : addr;
@@ -440,35 +320,17 @@ dma_addr_t xp_raw_get_dma(struct xsk_buff_pool *pool, u64 addr)
 }
 EXPORT_SYMBOL(xp_raw_get_dma);
 
-dma_addr_t xp_get_dma(struct xdp_buff_xsk *xskb)
-{
-	return xskb->dma;
-}
-EXPORT_SYMBOL(xp_get_dma);
-
-dma_addr_t xp_get_frame_dma(struct xdp_buff_xsk *xskb)
+void xp_dma_sync_for_cpu_slow(struct xdp_buff_xsk *xskb)
 {
-	return xskb->frame_dma;
-}
-EXPORT_SYMBOL(xp_get_frame_dma);
-
-void xp_dma_sync_for_cpu(struct xdp_buff_xsk *xskb)
-{
-	if (xskb->pool->cheap_dma)
-		return;
-
 	dma_sync_single_range_for_cpu(xskb->pool->dev, xskb->dma, 0,
 				      xskb->pool->frame_len, DMA_BIDIRECTIONAL);
 }
-EXPORT_SYMBOL(xp_dma_sync_for_cpu);
+EXPORT_SYMBOL(xp_dma_sync_for_cpu_slow);
 
-void xp_dma_sync_for_device(struct xsk_buff_pool *pool, dma_addr_t dma,
-			    size_t size)
+void xp_dma_sync_for_device_slow(struct xsk_buff_pool *pool, dma_addr_t dma,
+				 size_t size)
 {
-	if (pool->cheap_dma)
-		return;
-
 	dma_sync_single_range_for_device(pool->dev, dma, 0,
 					 size, DMA_BIDIRECTIONAL);
 }
-EXPORT_SYMBOL(xp_dma_sync_for_device);
+EXPORT_SYMBOL(xp_dma_sync_for_device_slow);
diff --git a/net/xdp/xsk_queue.h b/net/xdp/xsk_queue.h
index 16bf15864788..5b5d24d2dd37 100644
--- a/net/xdp/xsk_queue.h
+++ b/net/xdp/xsk_queue.h
@@ -118,6 +118,51 @@ static inline bool xskq_cons_read_addr_unchecked(struct xsk_queue *q, u64 *addr)
 	return false;
 }
 
+static inline bool xp_aligned_validate_desc(struct xsk_buff_pool *pool,
+					    struct xdp_desc *desc)
+{
+	u64 chunk, chunk_end;
+
+	chunk = xp_aligned_extract_addr(pool, desc->addr);
+	chunk_end = xp_aligned_extract_addr(pool, desc->addr + desc->len);
+	if (chunk != chunk_end)
+		return false;
+
+	if (chunk >= pool->addrs_cnt)
+		return false;
+
+	if (desc->options)
+		return false;
+	return true;
+}
+
+static inline bool xp_unaligned_validate_desc(struct xsk_buff_pool *pool,
+					      struct xdp_desc *desc)
+{
+	u64 addr, base_addr;
+
+	base_addr = xp_unaligned_extract_addr(desc->addr);
+	addr = xp_unaligned_add_offset_to_addr(desc->addr);
+
+	if (desc->len > pool->chunk_size)
+		return false;
+
+	if (base_addr >= pool->addrs_cnt || addr >= pool->addrs_cnt ||
+	    xp_desc_crosses_non_contig_pg(pool, addr, desc->len))
+		return false;
+
+	if (desc->options)
+		return false;
+	return true;
+}
+
+static inline bool xp_validate_desc(struct xsk_buff_pool *pool,
+				    struct xdp_desc *desc)
+{
+	return pool->unaligned ? xp_unaligned_validate_desc(pool, desc) :
+		xp_aligned_validate_desc(pool, desc);
+}
+
 static inline bool xskq_cons_is_valid_desc(struct xsk_queue *q,
 					   struct xdp_desc *d,
 					   struct xdp_umem *umem)
-- 
2.25.1

[PATCH bpf-next v4 15/15] MAINTAINERS, xsk: update AF_XDP section after moves/adds

[Björn Töpel]

From: Björn Töpel <bjorn.topel@intel.com>

Update MAINTAINERS to correctly mirror the current AF_XDP socket file
layout. Also, add the AF_XDP files of libbpf.

rfc->v1: Sorted file entries. (Joe)

Cc: Joe Perches <joe@perches.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
---
 MAINTAINERS | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/MAINTAINERS b/MAINTAINERS
index b7844f6cfa4a..087e68b21f9f 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -18443,8 +18443,12 @@ R:	Jonathan Lemon <jonathan.lemon@gmail.com>
 L:	netdev@vger.kernel.org
 L:	bpf@vger.kernel.org
 S:	Maintained
-F:	kernel/bpf/xskmap.c
+F:	include/net/xdp_sock*
+F:	include/net/xsk_buffer_pool.h
+F:	include/uapi/linux/if_xdp.h
 F:	net/xdp/
+F:	samples/bpf/xdpsock*
+F:	tools/lib/bpf/xsk*
 
 XEN BLOCK SUBSYSTEM
 M:	Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
-- 
2.25.1

Re: [PATCH bpf-next v4 01/15] xsk: fix xsk_umem_xdp_frame_sz()

[Jesper Dangaard Brouer]

On Wed, 20 May 2020 11:47:28 +0200
Björn Töpel <bjorn.topel@gmail.com> wrote:

> From: Björn Töpel <bjorn.topel@intel.com>
> 
> Calculating the "data_hard_end" for an XDP buffer coming from AF_XDP
> zero-copy mode, the return value of xsk_umem_xdp_frame_sz() is added
> to "data_hard_start".
> 
> Currently, the chunk size of the UMEM is returned by
> xsk_umem_xdp_frame_sz(). This is not correct, if the fixed UMEM
> headroom is non-zero. Fix this by returning the chunk_size without the
> UMEM headroom.
> 
> Fixes: 2a637c5b1aaf ("xdp: For Intel AF_XDP drivers add XDP frame_sz")
> Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
> ---
>  include/net/xdp_sock.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/include/net/xdp_sock.h b/include/net/xdp_sock.h
> index abd72de25fa4..6b1137ce1692 100644
> --- a/include/net/xdp_sock.h
> +++ b/include/net/xdp_sock.h
> @@ -239,7 +239,7 @@ static inline u64 xsk_umem_adjust_offset(struct xdp_umem *umem, u64 address,
>  
>  static inline u32 xsk_umem_xdp_frame_sz(struct xdp_umem *umem)
>  {
> -	return umem->chunk_size_nohr + umem->headroom;
> +	return umem->chunk_size_nohr;

Hmm, is this correct?

As you write "xdp_data_hard_end" is calculated as an offset from
xdp->data_hard_start pointer based on the frame_sz.  Will your
xdp->data_hard_start + frame_sz point to packet end?

#define xdp_data_hard_end(xdp)                          \
        ((xdp)->data_hard_start + (xdp)->frame_sz -     \
         SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))

Note the macro reserves the last 320 bytes (for skb_shared_info), but
for AF_XDP zero-copy mode, it will never create an SKB that use this
area.   Thus, in principle we can allow XDP-progs to extend/grow tail
into this area, but I don't think there is any use-case for this, as
it's much easier to access packet-data in userspace application.
(Thus, it might not be worth the complexity to give AF_XDP
bpf_xdp_adjust_tail access to this area, by e.g. "lying" via adding 320
bytes to frame_sz).

-- 
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer

Re: [PATCH bpf-next v4 01/15] xsk: fix xsk_umem_xdp_frame_sz()

[Björn Töpel]

On 2020-05-20 15:18, Jesper Dangaard Brouer wrote:
> On Wed, 20 May 2020 11:47:28 +0200
> Björn Töpel <bjorn.topel@gmail.com> wrote:
> 
>> From: Björn Töpel <bjorn.topel@intel.com>
>>
>> Calculating the "data_hard_end" for an XDP buffer coming from AF_XDP
>> zero-copy mode, the return value of xsk_umem_xdp_frame_sz() is added
>> to "data_hard_start".
>>
>> Currently, the chunk size of the UMEM is returned by
>> xsk_umem_xdp_frame_sz(). This is not correct, if the fixed UMEM
>> headroom is non-zero. Fix this by returning the chunk_size without the
>> UMEM headroom.
>>
>> Fixes: 2a637c5b1aaf ("xdp: For Intel AF_XDP drivers add XDP frame_sz")
>> Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
>> ---
>>   include/net/xdp_sock.h | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/include/net/xdp_sock.h b/include/net/xdp_sock.h
>> index abd72de25fa4..6b1137ce1692 100644
>> --- a/include/net/xdp_sock.h
>> +++ b/include/net/xdp_sock.h
>> @@ -239,7 +239,7 @@ static inline u64 xsk_umem_adjust_offset(struct xdp_umem *umem, u64 address,
>>   
>>   static inline u32 xsk_umem_xdp_frame_sz(struct xdp_umem *umem)
>>   {
>> -	return umem->chunk_size_nohr + umem->headroom;
>> +	return umem->chunk_size_nohr;
> 
> Hmm, is this correct?
> 
> As you write "xdp_data_hard_end" is calculated as an offset from
> xdp->data_hard_start pointer based on the frame_sz.  Will your
> xdp->data_hard_start + frame_sz point to packet end?
>

Yes, I believe this is correct.

Say that a user uses a chunk size of 2k, and a umem headroom of, say,
64. This means that the kernel should (at least) leave 64B which the
kernel shouldn't touch.

umem->headroom | XDP_PACKET_HEADROOM | packet |          |
                ^                     ^        ^      ^   ^
                a                     b        c      d   e

a: data_hard_start
b: data
c: data_end
d: data_hard_end, (e - 320)
e: hardlimit of chunk, a + umem->chunk_size_nohr

Prior this fix the umem->headroom was *included* in frame_sz.

> #define xdp_data_hard_end(xdp)                          \
>          ((xdp)->data_hard_start + (xdp)->frame_sz -     \
>           SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
> 
> Note the macro reserves the last 320 bytes (for skb_shared_info), but
> for AF_XDP zero-copy mode, it will never create an SKB that use this
> area.   Thus, in principle we can allow XDP-progs to extend/grow tail
> into this area, but I don't think there is any use-case for this, as
> it's much easier to access packet-data in userspace application.
> (Thus, it might not be worth the complexity to give AF_XDP
> bpf_xdp_adjust_tail access to this area, by e.g. "lying" via adding 320
> bytes to frame_sz).
> 

I agree, and in the picture (well...) above that would be "d". IOW
data_hard_end is 320 "off" the real end.


Björn

Re: [PATCH bpf-next v4 03/15] xsk: move driver interface to xdp_sock_drv.h

[Jakub Kicinski]

On Wed, 20 May 2020 11:47:30 +0200 Björn Töpel wrote:
> From: Magnus Karlsson <magnus.karlsson@intel.com>
> 
> Move the AF_XDP zero-copy driver interface to its own include file
> called xdp_sock_drv.h. This, hopefully, will make it more clear for
> NIC driver implementors to know what functions to use for zero-copy
> support.
> 
> Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>

With W=1:

net/xdp/xsk_queue.c:67:26: warning: symbol 'xsk_reuseq_prepare' was not declared. Should it be static?
net/xdp/xsk_queue.c:86:26: warning: symbol 'xsk_reuseq_swap' was not declared. Should it be static?
net/xdp/xsk_queue.c:108:6: warning: symbol 'xsk_reuseq_free' was not declared. Should it be static?
net/xdp/xsk_queue.c:67:27: warning: no previous prototype for xsk_reuseq_prepare [-Wmissing-prototypes]
  67 | struct xdp_umem_fq_reuse *xsk_reuseq_prepare(u32 nentries)
     |                           ^~~~~~~~~~~~~~~~~~
net/xdp/xsk_queue.c:86:27: warning: no previous prototype for xsk_reuseq_swap [-Wmissing-prototypes]
  86 | struct xdp_umem_fq_reuse *xsk_reuseq_swap(struct xdp_umem *umem,
     |                           ^~~~~~~~~~~~~~~
net/xdp/xsk_queue.c:108:6: warning: no previous prototype for xsk_reuseq_free [-Wmissing-prototypes]
 108 | void xsk_reuseq_free(struct xdp_umem_fq_reuse *rq)
     |      ^~~~~~~~~~~~~~~

Re: [PATCH bpf-next v4 07/15] i40e: separate kernel allocated rx_bi rings from AF_XDP rings

[Jakub Kicinski]

On Wed, 20 May 2020 11:47:34 +0200 Björn Töpel wrote:
> From: Björn Töpel <bjorn.topel@intel.com>
> 
> Continuing the path to support MEM_TYPE_XSK_BUFF_POOL, the AF_XDP
> zero-copy/sk_buff rx_bi rings are now separate. Functions to properly
> allocate the different rings are added as well.
> 
> v3->v4: Made i40e_fd_handle_status() static. (kbuild test robot)
> 
> Cc: intel-wired-lan@lists.osuosl.org
> Signed-off-by: Björn Töpel <bjorn.topel@intel.com>

There is a new warning here, at least one. But i40e has so many
existing warnings with W=1, I can't figure out which one is new :(

You most likely forgot to adjust kdoc somewhere after adding or
removing a function parameter.

Re: [PATCH bpf-next v4 09/15] ice, xsk: migrate to new MEM_TYPE_XSK_BUFF_POOL

[Jakub Kicinski]

On Wed, 20 May 2020 11:47:36 +0200 Björn Töpel wrote:
> From: Björn Töpel <bjorn.topel@intel.com>
> 
> Remove MEM_TYPE_ZERO_COPY in favor of the new MEM_TYPE_XSK_BUFF_POOL
> APIs.
> 
> Cc: intel-wired-lan@lists.osuosl.org
> Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
> Signed-off-by: Björn Töpel <bjorn.topel@intel.com>

patch 8 also has a warning I can't figure out.

But here (patch 9) it's quite clear:

drivers/net/ethernet/intel/ice/ice_xsk.c:414: warning: Excess function parameter 'alloc' description in 'ice_alloc_rx_bufs_zc'
drivers/net/ethernet/intel/ice/ice_xsk.c:480: warning: Excess function parameter 'xdp' description in 'ice_construct_skb_zc'

Re: [PATCH bpf-next v4 03/15] xsk: move driver interface to xdp_sock_drv.h

[Björn Töpel]

On Wed, 20 May 2020 at 18:57, Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Wed, 20 May 2020 11:47:30 +0200 Björn Töpel wrote:
> > From: Magnus Karlsson <magnus.karlsson@intel.com>
> >
> > Move the AF_XDP zero-copy driver interface to its own include file
> > called xdp_sock_drv.h. This, hopefully, will make it more clear for
> > NIC driver implementors to know what functions to use for zero-copy
> > support.
> >
> > Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
>
> With W=1:
>
> net/xdp/xsk_queue.c:67:26: warning: symbol 'xsk_reuseq_prepare' was not declared. Should it be static?
> net/xdp/xsk_queue.c:86:26: warning: symbol 'xsk_reuseq_swap' was not declared. Should it be static?
> net/xdp/xsk_queue.c:108:6: warning: symbol 'xsk_reuseq_free' was not declared. Should it be static?
> net/xdp/xsk_queue.c:67:27: warning: no previous prototype for xsk_reuseq_prepare [-Wmissing-prototypes]
>   67 | struct xdp_umem_fq_reuse *xsk_reuseq_prepare(u32 nentries)
>      |                           ^~~~~~~~~~~~~~~~~~
> net/xdp/xsk_queue.c:86:27: warning: no previous prototype for xsk_reuseq_swap [-Wmissing-prototypes]
>   86 | struct xdp_umem_fq_reuse *xsk_reuseq_swap(struct xdp_umem *umem,
>      |                           ^~~~~~~~~~~~~~~
> net/xdp/xsk_queue.c:108:6: warning: no previous prototype for xsk_reuseq_free [-Wmissing-prototypes]
>  108 | void xsk_reuseq_free(struct xdp_umem_fq_reuse *rq)
>      |      ^~~~~~~~~~~~~~~

Yes, missing include there. After the series these functions are
removed, but still good to get rid of the (bisect) noise.

I'll fix this in a v5. Thanks, Jakub!


Björn

Re: [PATCH bpf-next v4 07/15] i40e: separate kernel allocated rx_bi rings from AF_XDP rings

[Björn Töpel]

On Wed, 20 May 2020 at 19:02, Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Wed, 20 May 2020 11:47:34 +0200 Björn Töpel wrote:
> > From: Björn Töpel <bjorn.topel@intel.com>
> >
> > Continuing the path to support MEM_TYPE_XSK_BUFF_POOL, the AF_XDP
> > zero-copy/sk_buff rx_bi rings are now separate. Functions to properly
> > allocate the different rings are added as well.
> >
> > v3->v4: Made i40e_fd_handle_status() static. (kbuild test robot)
> >
> > Cc: intel-wired-lan@lists.osuosl.org
> > Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
>
> There is a new warning here, at least one. But i40e has so many
> existing warnings with W=1, I can't figure out which one is new :(
>
> You most likely forgot to adjust kdoc somewhere after adding or
> removing a function parameter.

Hmm, yes. A lot of warnings there. I'll see if I can find it. Thanks!

Re: [PATCH bpf-next v4 09/15] ice, xsk: migrate to new MEM_TYPE_XSK_BUFF_POOL

[Björn Töpel]

On Wed, 20 May 2020 at 19:03, Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Wed, 20 May 2020 11:47:36 +0200 Björn Töpel wrote:
> > From: Björn Töpel <bjorn.topel@intel.com>
> >
> > Remove MEM_TYPE_ZERO_COPY in favor of the new MEM_TYPE_XSK_BUFF_POOL
> > APIs.
> >
> > Cc: intel-wired-lan@lists.osuosl.org
> > Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
> > Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
>
> patch 8 also has a warning I can't figure out.
>
> But here (patch 9) it's quite clear:
>
> drivers/net/ethernet/intel/ice/ice_xsk.c:414: warning: Excess function parameter 'alloc' description in 'ice_alloc_rx_bufs_zc'
> drivers/net/ethernet/intel/ice/ice_xsk.c:480: warning: Excess function parameter 'xdp' description in 'ice_construct_skb_zc'

Thanks! I'll clean that up.

Björn

Re: [PATCH bpf-next v4 09/15] ice, xsk: migrate to new MEM_TYPE_XSK_BUFF_POOL

[Björn Töpel]

On Wed, 20 May 2020 at 19:03, Jakub Kicinski <kuba@kernel.org> wrote:
>
[...]
>
> patch 8 also has a warning I can't figure out.
>

Found and fixed. Thanks Jakub!


Björn

Re: [PATCH bpf-next v4 01/15] xsk: fix xsk_umem_xdp_frame_sz()

[Jesper Dangaard Brouer]

On Wed, 20 May 2020 16:34:05 +0200
Björn Töpel <bjorn.topel@intel.com> wrote:

> On 2020-05-20 15:18, Jesper Dangaard Brouer wrote:
> > On Wed, 20 May 2020 11:47:28 +0200
> > Björn Töpel <bjorn.topel@gmail.com> wrote:
> >   
> >> From: Björn Töpel <bjorn.topel@intel.com>
> >>
> >> Calculating the "data_hard_end" for an XDP buffer coming from AF_XDP
> >> zero-copy mode, the return value of xsk_umem_xdp_frame_sz() is added
> >> to "data_hard_start".
> >>
> >> Currently, the chunk size of the UMEM is returned by
> >> xsk_umem_xdp_frame_sz(). This is not correct, if the fixed UMEM
> >> headroom is non-zero. Fix this by returning the chunk_size without the
> >> UMEM headroom.
> >>
> >> Fixes: 2a637c5b1aaf ("xdp: For Intel AF_XDP drivers add XDP frame_sz")
> >> Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
> >> ---
> >>   include/net/xdp_sock.h | 2 +-
> >>   1 file changed, 1 insertion(+), 1 deletion(-)
> >>
> >> diff --git a/include/net/xdp_sock.h b/include/net/xdp_sock.h
> >> index abd72de25fa4..6b1137ce1692 100644
> >> --- a/include/net/xdp_sock.h
> >> +++ b/include/net/xdp_sock.h
> >> @@ -239,7 +239,7 @@ static inline u64 xsk_umem_adjust_offset(struct xdp_umem *umem, u64 address,
> >>   
> >>   static inline u32 xsk_umem_xdp_frame_sz(struct xdp_umem *umem)
> >>   {
> >> -	return umem->chunk_size_nohr + umem->headroom;
> >> +	return umem->chunk_size_nohr;  
> > 
> > Hmm, is this correct?
> > 
> > As you write "xdp_data_hard_end" is calculated as an offset from
> > xdp->data_hard_start pointer based on the frame_sz.  Will your
> > xdp->data_hard_start + frame_sz point to packet end?
> >  
> 
> Yes, I believe this is correct.
> 
> Say that a user uses a chunk size of 2k, and a umem headroom of, say,
> 64. This means that the kernel should (at least) leave 64B which the
> kernel shouldn't touch.
> 
> umem->headroom | XDP_PACKET_HEADROOM | packet |          |
>                 ^                     ^        ^      ^   ^
>                 a                     b        c      d   e
> 
> a: data_hard_start
> b: data
> c: data_end
> d: data_hard_end, (e - 320)
> e: hardlimit of chunk, a + umem->chunk_size_nohr
> 
> Prior this fix the umem->headroom was *included* in frame_sz.

Thanks for the nice ascii art description. I can now see that you are
right.   We should add this kind of documentation, perhaps as a comment
in the code?


> > #define xdp_data_hard_end(xdp)                          \
> >          ((xdp)->data_hard_start + (xdp)->frame_sz -     \
> >           SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
> > 
> > Note the macro reserves the last 320 bytes (for skb_shared_info), but
> > for AF_XDP zero-copy mode, it will never create an SKB that use this
> > area.   Thus, in principle we can allow XDP-progs to extend/grow tail
> > into this area, but I don't think there is any use-case for this, as
> > it's much easier to access packet-data in userspace application.
> > (Thus, it might not be worth the complexity to give AF_XDP
> > bpf_xdp_adjust_tail access to this area, by e.g. "lying" via adding 320
> > bytes to frame_sz).
> >   
> 
> I agree, and in the picture (well...) above that would be "d". IOW
> data_hard_end is 320 "off" the real end.

Yes, we agree.

-- 
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer

Re: [PATCH bpf-next v4 01/15] xsk: fix xsk_umem_xdp_frame_sz()

[Björn Töpel]

On Thu, 21 May 2020 at 06:30, Jesper Dangaard Brouer <brouer@redhat.com> wrote:
>
> On Wed, 20 May 2020 16:34:05 +0200
> Björn Töpel <bjorn.topel@intel.com> wrote:
>
> > On 2020-05-20 15:18, Jesper Dangaard Brouer wrote:
> > > On Wed, 20 May 2020 11:47:28 +0200
> > > Björn Töpel <bjorn.topel@gmail.com> wrote:
> > >
> > >> From: Björn Töpel <bjorn.topel@intel.com>
> > >>
> > >> Calculating the "data_hard_end" for an XDP buffer coming from AF_XDP
> > >> zero-copy mode, the return value of xsk_umem_xdp_frame_sz() is added
> > >> to "data_hard_start".
> > >>
> > >> Currently, the chunk size of the UMEM is returned by
> > >> xsk_umem_xdp_frame_sz(). This is not correct, if the fixed UMEM
> > >> headroom is non-zero. Fix this by returning the chunk_size without the
> > >> UMEM headroom.
> > >>
> > >> Fixes: 2a637c5b1aaf ("xdp: For Intel AF_XDP drivers add XDP frame_sz")
> > >> Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
> > >> ---
> > >>   include/net/xdp_sock.h | 2 +-
> > >>   1 file changed, 1 insertion(+), 1 deletion(-)
> > >>
> > >> diff --git a/include/net/xdp_sock.h b/include/net/xdp_sock.h
> > >> index abd72de25fa4..6b1137ce1692 100644
> > >> --- a/include/net/xdp_sock.h
> > >> +++ b/include/net/xdp_sock.h
> > >> @@ -239,7 +239,7 @@ static inline u64 xsk_umem_adjust_offset(struct xdp_umem *umem, u64 address,
> > >>
> > >>   static inline u32 xsk_umem_xdp_frame_sz(struct xdp_umem *umem)
> > >>   {
> > >> -  return umem->chunk_size_nohr + umem->headroom;
> > >> +  return umem->chunk_size_nohr;
> > >
> > > Hmm, is this correct?
> > >
> > > As you write "xdp_data_hard_end" is calculated as an offset from
> > > xdp->data_hard_start pointer based on the frame_sz.  Will your
> > > xdp->data_hard_start + frame_sz point to packet end?
> > >
> >
> > Yes, I believe this is correct.
> >
> > Say that a user uses a chunk size of 2k, and a umem headroom of, say,
> > 64. This means that the kernel should (at least) leave 64B which the
> > kernel shouldn't touch.
> >
> > umem->headroom | XDP_PACKET_HEADROOM | packet |          |
> >                 ^                     ^        ^      ^   ^
> >                 a                     b        c      d   e
> >
> > a: data_hard_start
> > b: data
> > c: data_end
> > d: data_hard_end, (e - 320)
> > e: hardlimit of chunk, a + umem->chunk_size_nohr
> >
> > Prior this fix the umem->headroom was *included* in frame_sz.
>
> Thanks for the nice ascii art description. I can now see that you are
> right.   We should add this kind of documentation, perhaps as a comment
> in the code?
>

Definitely! I'd say both in code, and af_xdp.rst! I'll make a patch.
Thanks for the suggestion!

Cheers,
Björn