nfp: add BPF to NFP code translator

Add translator for JITing eBPF to operations which
can be executed on NFP's programmable engines.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

5 files changed, 2333 insertions, 0 deletions

diff --git a/drivers/net/ethernet/netronome/nfp/Makefile b/drivers/net/ethernet/netronome/nfp/Makefile
index 68178819ff12..5f12689bf523 100644
--- a/drivers/net/ethernet/netronome/nfp/Makefile
+++ b/drivers/net/ethernet/netronome/nfp/Makefile
@@ -5,4 +5,10 @@ nfp_netvf-objs := \
 	    nfp_net_ethtool.o \
 	    nfp_netvf_main.o
 
+ifeq ($(CONFIG_BPF_SYSCALL),y)
+nfp_netvf-objs += \
+	    nfp_bpf_verifier.o \
+	    nfp_bpf_jit.o
+endif
+
 nfp_netvf-$(CONFIG_NFP_NET_DEBUG) += nfp_net_debugfs.o
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_asm.h b/drivers/net/ethernet/netronome/nfp/nfp_asm.h
new file mode 100644
index 000000000000..22484b6fd3e8
--- /dev/null
+++ b/drivers/net/ethernet/netronome/nfp/nfp_asm.h
@@ -0,0 +1,233 @@
+/*
+ * Copyright (C) 2016 Netronome Systems, Inc.
+ *
+ * This software is dual licensed under the GNU General License Version 2,
+ * June 1991 as shown in the file COPYING in the top-level directory of this
+ * source tree or the BSD 2-Clause License provided below.  You have the
+ * option to license this software under the complete terms of either license.
+ *
+ * The BSD 2-Clause License:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      1. Redistributions of source code must retain the above
+ *         copyright notice, this list of conditions and the following
+ *         disclaimer.
+ *
+ *      2. Redistributions in binary form must reproduce the above
+ *         copyright notice, this list of conditions and the following
+ *         disclaimer in the documentation and/or other materials
+ *         provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __NFP_ASM_H__
+#define __NFP_ASM_H__ 1
+
+#include "nfp_bpf.h"
+
+#define REG_NONE	0
+
+#define RE_REG_NO_DST	0x020
+#define RE_REG_IMM	0x020
+#define RE_REG_IMM_encode(x)					\
+	(RE_REG_IMM | ((x) & 0x1f) | (((x) & 0x60) << 1))
+#define RE_REG_IMM_MAX	 0x07fULL
+#define RE_REG_XFR	0x080
+
+#define UR_REG_XFR	0x180
+#define UR_REG_NN	0x280
+#define UR_REG_NO_DST	0x300
+#define UR_REG_IMM	UR_REG_NO_DST
+#define UR_REG_IMM_encode(x) (UR_REG_IMM | (x))
+#define UR_REG_IMM_MAX	 0x0ffULL
+
+#define OP_BR_BASE	0x0d800000020ULL
+#define OP_BR_BASE_MASK	0x0f8000c3ce0ULL
+#define OP_BR_MASK	0x0000000001fULL
+#define OP_BR_EV_PIP	0x00000000300ULL
+#define OP_BR_CSS	0x0000003c000ULL
+#define OP_BR_DEFBR	0x00000300000ULL
+#define OP_BR_ADDR_LO	0x007ffc00000ULL
+#define OP_BR_ADDR_HI	0x10000000000ULL
+
+#define nfp_is_br(_insn)				\
+	(((_insn) & OP_BR_BASE_MASK) == OP_BR_BASE)
+
+enum br_mask {
+	BR_BEQ = 0x00,
+	BR_BNE = 0x01,
+	BR_BHS = 0x04,
+	BR_BLO = 0x05,
+	BR_BGE = 0x08,
+	BR_UNC = 0x18,
+};
+
+enum br_ev_pip {
+	BR_EV_PIP_UNCOND = 0,
+	BR_EV_PIP_COND = 1,
+};
+
+enum br_ctx_signal_state {
+	BR_CSS_NONE = 2,
+};
+
+#define OP_BBYTE_BASE	0x0c800000000ULL
+#define OP_BB_A_SRC	0x000000000ffULL
+#define OP_BB_BYTE	0x00000000300ULL
+#define OP_BB_B_SRC	0x0000003fc00ULL
+#define OP_BB_I8	0x00000040000ULL
+#define OP_BB_EQ	0x00000080000ULL
+#define OP_BB_DEFBR	0x00000300000ULL
+#define OP_BB_ADDR_LO	0x007ffc00000ULL
+#define OP_BB_ADDR_HI	0x10000000000ULL
+
+#define OP_BALU_BASE	0x0e800000000ULL
+#define OP_BA_A_SRC	0x000000003ffULL
+#define OP_BA_B_SRC	0x000000ffc00ULL
+#define OP_BA_DEFBR	0x00000300000ULL
+#define OP_BA_ADDR_HI	0x0007fc00000ULL
+
+#define OP_IMMED_A_SRC	0x000000003ffULL
+#define OP_IMMED_B_SRC	0x000000ffc00ULL
+#define OP_IMMED_IMM	0x0000ff00000ULL
+#define OP_IMMED_WIDTH	0x00060000000ULL
+#define OP_IMMED_INV	0x00080000000ULL
+#define OP_IMMED_SHIFT	0x00600000000ULL
+#define OP_IMMED_BASE	0x0f000000000ULL
+#define OP_IMMED_WR_AB	0x20000000000ULL
+
+enum immed_width {
+	IMMED_WIDTH_ALL = 0,
+	IMMED_WIDTH_BYTE = 1,
+	IMMED_WIDTH_WORD = 2,
+};
+
+enum immed_shift {
+	IMMED_SHIFT_0B = 0,
+	IMMED_SHIFT_1B = 1,
+	IMMED_SHIFT_2B = 2,
+};
+
+#define OP_SHF_BASE	0x08000000000ULL
+#define OP_SHF_A_SRC	0x000000000ffULL
+#define OP_SHF_SC	0x00000000300ULL
+#define OP_SHF_B_SRC	0x0000003fc00ULL
+#define OP_SHF_I8	0x00000040000ULL
+#define OP_SHF_SW	0x00000080000ULL
+#define OP_SHF_DST	0x0000ff00000ULL
+#define OP_SHF_SHIFT	0x001f0000000ULL
+#define OP_SHF_OP	0x00e00000000ULL
+#define OP_SHF_DST_AB	0x01000000000ULL
+#define OP_SHF_WR_AB	0x20000000000ULL
+
+enum shf_op {
+	SHF_OP_NONE = 0,
+	SHF_OP_AND = 2,
+	SHF_OP_OR = 5,
+};
+
+enum shf_sc {
+	SHF_SC_R_ROT = 0,
+	SHF_SC_R_SHF = 1,
+	SHF_SC_L_SHF = 2,
+	SHF_SC_R_DSHF = 3,
+};
+
+#define OP_ALU_A_SRC	0x000000003ffULL
+#define OP_ALU_B_SRC	0x000000ffc00ULL
+#define OP_ALU_DST	0x0003ff00000ULL
+#define OP_ALU_SW	0x00040000000ULL
+#define OP_ALU_OP	0x00f80000000ULL
+#define OP_ALU_DST_AB	0x01000000000ULL
+#define OP_ALU_BASE	0x0a000000000ULL
+#define OP_ALU_WR_AB	0x20000000000ULL
+
+enum alu_op {
+	ALU_OP_NONE	= 0x00,
+	ALU_OP_ADD	= 0x01,
+	ALU_OP_NEG	= 0x04,
+	ALU_OP_AND	= 0x08,
+	ALU_OP_SUB_C	= 0x0d,
+	ALU_OP_ADD_C	= 0x11,
+	ALU_OP_OR	= 0x14,
+	ALU_OP_SUB	= 0x15,
+	ALU_OP_XOR	= 0x18,
+};
+
+enum alu_dst_ab {
+	ALU_DST_A = 0,
+	ALU_DST_B = 1,
+};
+
+#define OP_LDF_BASE	0x0c000000000ULL
+#define OP_LDF_A_SRC	0x000000000ffULL
+#define OP_LDF_SC	0x00000000300ULL
+#define OP_LDF_B_SRC	0x0000003fc00ULL
+#define OP_LDF_I8	0x00000040000ULL
+#define OP_LDF_SW	0x00000080000ULL
+#define OP_LDF_ZF	0x00000100000ULL
+#define OP_LDF_BMASK	0x0000f000000ULL
+#define OP_LDF_SHF	0x001f0000000ULL
+#define OP_LDF_WR_AB	0x20000000000ULL
+
+#define OP_CMD_A_SRC	 0x000000000ffULL
+#define OP_CMD_CTX	 0x00000000300ULL
+#define OP_CMD_B_SRC	 0x0000003fc00ULL
+#define OP_CMD_TOKEN	 0x000000c0000ULL
+#define OP_CMD_XFER	 0x00001f00000ULL
+#define OP_CMD_CNT	 0x0000e000000ULL
+#define OP_CMD_SIG	 0x000f0000000ULL
+#define OP_CMD_TGT_CMD	 0x07f00000000ULL
+#define OP_CMD_MODE	0x1c0000000000ULL
+
+struct cmd_tgt_act {
+	u8 token;
+	u8 tgt_cmd;
+};
+
+enum cmd_tgt_map {
+	CMD_TGT_READ8,
+	CMD_TGT_WRITE8,
+	CMD_TGT_READ_LE,
+	CMD_TGT_READ_SWAP_LE,
+	__CMD_TGT_MAP_SIZE,
+};
+
+enum cmd_mode {
+	CMD_MODE_40b_AB	= 0,
+	CMD_MODE_40b_BA	= 1,
+	CMD_MODE_32b	= 4,
+};
+
+enum cmd_ctx_swap {
+	CMD_CTX_SWAP = 0,
+	CMD_CTX_NO_SWAP = 3,
+};
+
+#define OP_LCSR_BASE	0x0fc00000000ULL
+#define OP_LCSR_A_SRC	0x000000003ffULL
+#define OP_LCSR_B_SRC	0x000000ffc00ULL
+#define OP_LCSR_WRITE	0x00000200000ULL
+#define OP_LCSR_ADDR	0x001ffc00000ULL
+
+enum lcsr_wr_src {
+	LCSR_WR_AREG,
+	LCSR_WR_BREG,
+	LCSR_WR_IMM,
+};
+
+#define OP_CARB_BASE	0x0e000000000ULL
+#define OP_CARB_OR	0x00000010000ULL
+
+#endif
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_bpf.h b/drivers/net/ethernet/netronome/nfp/nfp_bpf.h
new file mode 100644
index 000000000000..3726421e353f
--- /dev/null
+++ b/drivers/net/ethernet/netronome/nfp/nfp_bpf.h
@@ -0,0 +1,208 @@
+/*
+ * Copyright (C) 2016 Netronome Systems, Inc.
+ *
+ * This software is dual licensed under the GNU General License Version 2,
+ * June 1991 as shown in the file COPYING in the top-level directory of this
+ * source tree or the BSD 2-Clause License provided below.  You have the
+ * option to license this software under the complete terms of either license.
+ *
+ * The BSD 2-Clause License:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      1. Redistributions of source code must retain the above
+ *         copyright notice, this list of conditions and the following
+ *         disclaimer.
+ *
+ *      2. Redistributions in binary form must reproduce the above
+ *         copyright notice, this list of conditions and the following
+ *         disclaimer in the documentation and/or other materials
+ *         provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef __NFP_BPF_H__
+#define __NFP_BPF_H__ 1
+
+#include <linux/bitfield.h>
+#include <linux/bpf.h>
+#include <linux/list.h>
+#include <linux/types.h>
+
+#define FIELD_FIT(mask, val)  (!((((u64)val) << __bf_shf(mask)) & ~(mask)))
+
+/* For branch fixup logic use up-most byte of branch instruction as scratch
+ * area.  Remember to clear this before sending instructions to HW!
+ */
+#define OP_BR_SPECIAL	0xff00000000000000ULL
+
+enum br_special {
+	OP_BR_NORMAL = 0,
+	OP_BR_GO_OUT,
+	OP_BR_GO_ABORT,
+};
+
+enum static_regs {
+	STATIC_REG_PKT		= 1,
+#define REG_PKT_BANK	ALU_DST_A
+	STATIC_REG_IMM		= 2, /* Bank AB */
+};
+
+enum nfp_bpf_action_type {
+	NN_ACT_TC_DROP,
+};
+
+/* Software register representation, hardware encoding in asm.h */
+#define NN_REG_TYPE	GENMASK(31, 24)
+#define NN_REG_VAL	GENMASK(7, 0)
+
+enum nfp_bpf_reg_type {
+	NN_REG_GPR_A =	BIT(0),
+	NN_REG_GPR_B =	BIT(1),
+	NN_REG_NNR =	BIT(2),
+	NN_REG_XFER =	BIT(3),
+	NN_REG_IMM =	BIT(4),
+	NN_REG_NONE =	BIT(5),
+};
+
+#define NN_REG_GPR_BOTH	(NN_REG_GPR_A | NN_REG_GPR_B)
+
+#define reg_both(x)	((x) | FIELD_PREP(NN_REG_TYPE, NN_REG_GPR_BOTH))
+#define reg_a(x)	((x) | FIELD_PREP(NN_REG_TYPE, NN_REG_GPR_A))
+#define reg_b(x)	((x) | FIELD_PREP(NN_REG_TYPE, NN_REG_GPR_B))
+#define reg_nnr(x)	((x) | FIELD_PREP(NN_REG_TYPE, NN_REG_NNR))
+#define reg_xfer(x)	((x) | FIELD_PREP(NN_REG_TYPE, NN_REG_XFER))
+#define reg_imm(x)	((x) | FIELD_PREP(NN_REG_TYPE, NN_REG_IMM))
+#define reg_none()	(FIELD_PREP(NN_REG_TYPE, NN_REG_NONE))
+
+#define pkt_reg(np)	reg_a((np)->regs_per_thread - STATIC_REG_PKT)
+#define imm_a(np)	reg_a((np)->regs_per_thread - STATIC_REG_IMM)
+#define imm_b(np)	reg_b((np)->regs_per_thread - STATIC_REG_IMM)
+#define imm_both(np)	reg_both((np)->regs_per_thread - STATIC_REG_IMM)
+
+#define NFP_BPF_ABI_FLAGS	reg_nnr(0)
+#define NFP_BPF_ABI_PKT		reg_nnr(2)
+#define NFP_BPF_ABI_LEN		reg_nnr(3)
+
+struct nfp_prog;
+struct nfp_insn_meta;
+typedef int (*instr_cb_t)(struct nfp_prog *, struct nfp_insn_meta *);
+
+#define nfp_prog_first_meta(nfp_prog)					\
+	list_first_entry(&(nfp_prog)->insns, struct nfp_insn_meta, l)
+#define nfp_prog_last_meta(nfp_prog)					\
+	list_last_entry(&(nfp_prog)->insns, struct nfp_insn_meta, l)
+#define nfp_meta_next(meta)	list_next_entry(meta, l)
+#define nfp_meta_prev(meta)	list_prev_entry(meta, l)
+
+/**
+ * struct nfp_insn_meta - BPF instruction wrapper
+ * @insn: BPF instruction
+ * @off: index of first generated machine instruction (in nfp_prog.prog)
+ * @n: eBPF instruction number
+ * @skip: skip this instruction (optimized out)
+ * @double_cb: callback for second part of the instruction
+ * @l: link on nfp_prog->insns list
+ */
+struct nfp_insn_meta {
+	struct bpf_insn insn;
+	unsigned int off;
+	unsigned short n;
+	bool skip;
+	instr_cb_t double_cb;
+
+	struct list_head l;
+};
+
+#define BPF_SIZE_MASK	0x18
+
+static inline u8 mbpf_class(const struct nfp_insn_meta *meta)
+{
+	return BPF_CLASS(meta->insn.code);
+}
+
+static inline u8 mbpf_src(const struct nfp_insn_meta *meta)
+{
+	return BPF_SRC(meta->insn.code);
+}
+
+static inline u8 mbpf_op(const struct nfp_insn_meta *meta)
+{
+	return BPF_OP(meta->insn.code);
+}
+
+static inline u8 mbpf_mode(const struct nfp_insn_meta *meta)
+{
+	return BPF_MODE(meta->insn.code);
+}
+
+/**
+ * struct nfp_prog - nfp BPF program
+ * @prog: machine code
+ * @prog_len: number of valid instructions in @prog array
+ * @__prog_alloc_len: alloc size of @prog array
+ * @act: BPF program/action type (TC DA, TC with action, XDP etc.)
+ * @num_regs: number of registers used by this program
+ * @regs_per_thread: number of basic registers allocated per thread
+ * @start_off: address of the first instruction in the memory
+ * @tgt_out: jump target for normal exit
+ * @tgt_abort: jump target for abort (e.g. access outside of packet buffer)
+ * @tgt_done: jump target to get the next packet
+ * @n_translated: number of successfully translated instructions (for errors)
+ * @error: error code if something went wrong
+ * @insns: list of BPF instruction wrappers (struct nfp_insn_meta)
+ */
+struct nfp_prog {
+	u64 *prog;
+	unsigned int prog_len;
+	unsigned int __prog_alloc_len;
+
+	enum nfp_bpf_action_type act;
+
+	unsigned int num_regs;
+	unsigned int regs_per_thread;
+
+	unsigned int start_off;
+	unsigned int tgt_out;
+	unsigned int tgt_abort;
+	unsigned int tgt_done;
+
+	unsigned int n_translated;
+	int error;
+
+	struct list_head insns;
+};
+
+struct nfp_bpf_result {
+	unsigned int n_instr;
+	bool dense_mode;
+};
+
+#ifdef CONFIG_BPF_SYSCALL
+int
+nfp_bpf_jit(struct bpf_prog *filter, void *prog, enum nfp_bpf_action_type act,
+	    unsigned int prog_start, unsigned int prog_done,
+	    unsigned int prog_sz, struct nfp_bpf_result *res);
+#else
+int
+nfp_bpf_jit(struct bpf_prog *filter, void *prog, enum nfp_bpf_action_type act,
+	    unsigned int prog_start, unsigned int prog_done,
+	    unsigned int prog_sz, struct nfp_bpf_result *res)
+{
+	return -ENOTSUPP;
+}
+#endif
+
+int nfp_prog_verify(struct nfp_prog *nfp_prog, struct bpf_prog *prog);
+
+#endif
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_bpf_jit.c b/drivers/net/ethernet/netronome/nfp/nfp_bpf_jit.c
new file mode 100644
index 000000000000..cfbf53607fc9
--- /dev/null
+++ b/drivers/net/ethernet/netronome/nfp/nfp_bpf_jit.c
@@ -0,0 +1,1724 @@
+/*
+ * Copyright (C) 2016 Netronome Systems, Inc.
+ *
+ * This software is dual licensed under the GNU General License Version 2,
+ * June 1991 as shown in the file COPYING in the top-level directory of this
+ * source tree or the BSD 2-Clause License provided below.  You have the
+ * option to license this software under the complete terms of either license.
+ *
+ * The BSD 2-Clause License:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      1. Redistributions of source code must retain the above
+ *         copyright notice, this list of conditions and the following
+ *         disclaimer.
+ *
+ *      2. Redistributions in binary form must reproduce the above
+ *         copyright notice, this list of conditions and the following
+ *         disclaimer in the documentation and/or other materials
+ *         provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#define pr_fmt(fmt)	"NFP net bpf: " fmt
+
+#include <linux/kernel.h>
+#include <linux/bpf.h>
+#include <linux/filter.h>
+#include <linux/pkt_cls.h>
+#include <linux/unistd.h>
+
+#include "nfp_asm.h"
+#include "nfp_bpf.h"
+
+/* --- NFP prog --- */
+/* Foreach "multiple" entries macros provide pos and next<n> pointers.
+ * It's safe to modify the next pointers (but not pos).
+ */
+#define nfp_for_each_insn_walk2(nfp_prog, pos, next)			\
+	for (pos = list_first_entry(&(nfp_prog)->insns, typeof(*pos), l), \
+	     next = list_next_entry(pos, l);			\
+	     &(nfp_prog)->insns != &pos->l &&			\
+	     &(nfp_prog)->insns != &next->l;			\
+	     pos = nfp_meta_next(pos),				\
+	     next = nfp_meta_next(pos))
+
+#define nfp_for_each_insn_walk3(nfp_prog, pos, next, next2)		\
+	for (pos = list_first_entry(&(nfp_prog)->insns, typeof(*pos), l), \
+	     next = list_next_entry(pos, l),			\
+	     next2 = list_next_entry(next, l);			\
+	     &(nfp_prog)->insns != &pos->l &&			\
+	     &(nfp_prog)->insns != &next->l &&			\
+	     &(nfp_prog)->insns != &next2->l;			\
+	     pos = nfp_meta_next(pos),				\
+	     next = nfp_meta_next(pos),				\
+	     next2 = nfp_meta_next(next))
+
+static bool
+nfp_meta_has_next(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return meta->l.next != &nfp_prog->insns;
+}
+
+static bool
+nfp_meta_has_prev(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return meta->l.prev != &nfp_prog->insns;
+}
+
+static void nfp_prog_free(struct nfp_prog *nfp_prog)
+{
+	struct nfp_insn_meta *meta, *tmp;
+
+	list_for_each_entry_safe(meta, tmp, &nfp_prog->insns, l) {
+		list_del(&meta->l);
+		kfree(meta);
+	}
+	kfree(nfp_prog);
+}
+
+static void nfp_prog_push(struct nfp_prog *nfp_prog, u64 insn)
+{
+	if (nfp_prog->__prog_alloc_len == nfp_prog->prog_len) {
+		nfp_prog->error = -ENOSPC;
+		return;
+	}
+
+	nfp_prog->prog[nfp_prog->prog_len] = insn;
+	nfp_prog->prog_len++;
+}
+
+static unsigned int nfp_prog_current_offset(struct nfp_prog *nfp_prog)
+{
+	return nfp_prog->start_off + nfp_prog->prog_len;
+}
+
+static unsigned int
+nfp_prog_offset_to_index(struct nfp_prog *nfp_prog, unsigned int offset)
+{
+	return offset - nfp_prog->start_off;
+}
+
+/* --- SW reg --- */
+struct nfp_insn_ur_regs {
+	enum alu_dst_ab dst_ab;
+	u16 dst;
+	u16 areg, breg;
+	bool swap;
+	bool wr_both;
+};
+
+struct nfp_insn_re_regs {
+	enum alu_dst_ab dst_ab;
+	u8 dst;
+	u8 areg, breg;
+	bool swap;
+	bool wr_both;
+	bool i8;
+};
+
+static u16 nfp_swreg_to_unreg(u32 swreg, bool is_dst)
+{
+	u16 val = FIELD_GET(NN_REG_VAL, swreg);
+
+	switch (FIELD_GET(NN_REG_TYPE, swreg)) {
+	case NN_REG_GPR_A:
+	case NN_REG_GPR_B:
+	case NN_REG_GPR_BOTH:
+		return val;
+	case NN_REG_NNR:
+		return UR_REG_NN | val;
+	case NN_REG_XFER:
+		return UR_REG_XFR | val;
+	case NN_REG_IMM:
+		if (val & ~0xff) {
+			pr_err("immediate too large\n");
+			return 0;
+		}
+		return UR_REG_IMM_encode(val);
+	case NN_REG_NONE:
+		return is_dst ? UR_REG_NO_DST : REG_NONE;
+	default:
+		pr_err("unrecognized reg encoding %08x\n", swreg);
+		return 0;
+	}
+}
+
+static int
+swreg_to_unrestricted(u32 dst, u32 lreg, u32 rreg, struct nfp_insn_ur_regs *reg)
+{
+	memset(reg, 0, sizeof(*reg));
+
+	/* Decode destination */
+	if (FIELD_GET(NN_REG_TYPE, dst) == NN_REG_IMM)
+		return -EFAULT;
+
+	if (FIELD_GET(NN_REG_TYPE, dst) == NN_REG_GPR_B)
+		reg->dst_ab = ALU_DST_B;
+	if (FIELD_GET(NN_REG_TYPE, dst) == NN_REG_GPR_BOTH)
+		reg->wr_both = true;
+	reg->dst = nfp_swreg_to_unreg(dst, true);
+
+	/* Decode source operands */
+	if (FIELD_GET(NN_REG_TYPE, lreg) == FIELD_GET(NN_REG_TYPE, rreg))
+		return -EFAULT;
+
+	if (FIELD_GET(NN_REG_TYPE, lreg) == NN_REG_GPR_B ||
+	    FIELD_GET(NN_REG_TYPE, rreg) == NN_REG_GPR_A) {
+		reg->areg = nfp_swreg_to_unreg(rreg, false);
+		reg->breg = nfp_swreg_to_unreg(lreg, false);
+		reg->swap = true;
+	} else {
+		reg->areg = nfp_swreg_to_unreg(lreg, false);
+		reg->breg = nfp_swreg_to_unreg(rreg, false);
+	}
+
+	return 0;
+}
+
+static u16 nfp_swreg_to_rereg(u32 swreg, bool is_dst, bool has_imm8, bool *i8)
+{
+	u16 val = FIELD_GET(NN_REG_VAL, swreg);
+
+	switch (FIELD_GET(NN_REG_TYPE, swreg)) {
+	case NN_REG_GPR_A:
+	case NN_REG_GPR_B:
+	case NN_REG_GPR_BOTH:
+		return val;
+	case NN_REG_XFER:
+		return RE_REG_XFR | val;
+	case NN_REG_IMM:
+		if (val & ~(0x7f | has_imm8 << 7)) {
+			pr_err("immediate too large\n");
+			return 0;
+		}
+		*i8 = val & 0x80;
+		return RE_REG_IMM_encode(val & 0x7f);
+	case NN_REG_NONE:
+		return is_dst ? RE_REG_NO_DST : REG_NONE;
+	default:
+		pr_err("unrecognized reg encoding\n");
+		return 0;
+	}
+}
+
+static int
+swreg_to_restricted(u32 dst, u32 lreg, u32 rreg, struct nfp_insn_re_regs *reg,
+		    bool has_imm8)
+{
+	memset(reg, 0, sizeof(*reg));
+
+	/* Decode destination */
+	if (FIELD_GET(NN_REG_TYPE, dst) == NN_REG_IMM)
+		return -EFAULT;
+
+	if (FIELD_GET(NN_REG_TYPE, dst) == NN_REG_GPR_B)
+		reg->dst_ab = ALU_DST_B;
+	if (FIELD_GET(NN_REG_TYPE, dst) == NN_REG_GPR_BOTH)
+		reg->wr_both = true;
+	reg->dst = nfp_swreg_to_rereg(dst, true, false, NULL);
+
+	/* Decode source operands */
+	if (FIELD_GET(NN_REG_TYPE, lreg) == FIELD_GET(NN_REG_TYPE, rreg))
+		return -EFAULT;
+
+	if (FIELD_GET(NN_REG_TYPE, lreg) == NN_REG_GPR_B ||
+	    FIELD_GET(NN_REG_TYPE, rreg) == NN_REG_GPR_A) {
+		reg->areg = nfp_swreg_to_rereg(rreg, false, has_imm8, &reg->i8);
+		reg->breg = nfp_swreg_to_rereg(lreg, false, has_imm8, &reg->i8);
+		reg->swap = true;
+	} else {
+		reg->areg = nfp_swreg_to_rereg(lreg, false, has_imm8, &reg->i8);
+		reg->breg = nfp_swreg_to_rereg(rreg, false, has_imm8, &reg->i8);
+	}
+
+	return 0;
+}
+
+/* --- Emitters --- */
+static const struct cmd_tgt_act cmd_tgt_act[__CMD_TGT_MAP_SIZE] = {
+	[CMD_TGT_WRITE8] =		{ 0x00, 0x42 },
+	[CMD_TGT_READ8] =		{ 0x01, 0x43 },
+	[CMD_TGT_READ_LE] =		{ 0x01, 0x40 },
+	[CMD_TGT_READ_SWAP_LE] =	{ 0x03, 0x40 },
+};
+
+static void
+__emit_cmd(struct nfp_prog *nfp_prog, enum cmd_tgt_map op,
+	   u8 mode, u8 xfer, u8 areg, u8 breg, u8 size, bool sync)
+{
+	enum cmd_ctx_swap ctx;
+	u64 insn;
+
+	if (sync)
+		ctx = CMD_CTX_SWAP;
+	else
+		ctx = CMD_CTX_NO_SWAP;
+
+	insn =	FIELD_PREP(OP_CMD_A_SRC, areg) |
+		FIELD_PREP(OP_CMD_CTX, ctx) |
+		FIELD_PREP(OP_CMD_B_SRC, breg) |
+		FIELD_PREP(OP_CMD_TOKEN, cmd_tgt_act[op].token) |
+		FIELD_PREP(OP_CMD_XFER, xfer) |
+		FIELD_PREP(OP_CMD_CNT, size) |
+		FIELD_PREP(OP_CMD_SIG, sync) |
+		FIELD_PREP(OP_CMD_TGT_CMD, cmd_tgt_act[op].tgt_cmd) |
+		FIELD_PREP(OP_CMD_MODE, mode);
+
+	nfp_prog_push(nfp_prog, insn);
+}
+
+static void
+emit_cmd(struct nfp_prog *nfp_prog, enum cmd_tgt_map op,
+	 u8 mode, u8 xfer, u32 lreg, u32 rreg, u8 size, bool sync)
+{
+	struct nfp_insn_re_regs reg;
+	int err;
+
+	err = swreg_to_restricted(reg_none(), lreg, rreg, &reg, false);
+	if (err) {
+		nfp_prog->error = err;
+		return;
+	}
+	if (reg.swap) {
+		pr_err("cmd can't swap arguments\n");
+		nfp_prog->error = -EFAULT;
+		return;
+	}
+
+	__emit_cmd(nfp_prog, op, mode, xfer, reg.areg, reg.breg, size, sync);
+}
+
+static void
+__emit_br(struct nfp_prog *nfp_prog, enum br_mask mask, enum br_ev_pip ev_pip,
+	  enum br_ctx_signal_state css, u16 addr, u8 defer)
+{
+	u16 addr_lo, addr_hi;
+	u64 insn;
+
+	addr_lo = addr & (OP_BR_ADDR_LO >> __bf_shf(OP_BR_ADDR_LO));
+	addr_hi = addr != addr_lo;
+
+	insn = OP_BR_BASE |
+		FIELD_PREP(OP_BR_MASK, mask) |
+		FIELD_PREP(OP_BR_EV_PIP, ev_pip) |
+		FIELD_PREP(OP_BR_CSS, css) |
+		FIELD_PREP(OP_BR_DEFBR, defer) |
+		FIELD_PREP(OP_BR_ADDR_LO, addr_lo) |
+		FIELD_PREP(OP_BR_ADDR_HI, addr_hi);
+
+	nfp_prog_push(nfp_prog, insn);
+}
+
+static void
+emit_br(struct nfp_prog *nfp_prog, enum br_mask mask, u16 addr, u8 defer)
+{
+	__emit_br(nfp_prog, mask,
+		  mask != BR_UNC ? BR_EV_PIP_COND : BR_EV_PIP_UNCOND,
+		  BR_CSS_NONE, addr, defer);
+}
+
+static void
+__emit_br_byte(struct nfp_prog *nfp_prog, u8 areg, u8 breg, bool imm8,
+	       u8 byte, bool equal, u16 addr, u8 defer)
+{
+	u16 addr_lo, addr_hi;
+	u64 insn;
+
+	addr_lo = addr & (OP_BB_ADDR_LO >> __bf_shf(OP_BB_ADDR_LO));
+	addr_hi = addr != addr_lo;
+
+	insn = OP_BBYTE_BASE |
+		FIELD_PREP(OP_BB_A_SRC, areg) |
+		FIELD_PREP(OP_BB_BYTE, byte) |
+		FIELD_PREP(OP_BB_B_SRC, breg) |
+		FIELD_PREP(OP_BB_I8, imm8) |
+		FIELD_PREP(OP_BB_EQ, equal) |
+		FIELD_PREP(OP_BB_DEFBR, defer) |
+		FIELD_PREP(OP_BB_ADDR_LO, addr_lo) |
+		FIELD_PREP(OP_BB_ADDR_HI, addr_hi);
+
+	nfp_prog_push(nfp_prog, insn);
+}
+
+static void
+emit_br_byte_neq(struct nfp_prog *nfp_prog,
+		 u32 dst, u8 imm, u8 byte, u16 addr, u8 defer)
+{
+	struct nfp_insn_re_regs reg;
+	int err;
+
+	err = swreg_to_restricted(reg_none(), dst, reg_imm(imm), &reg, true);
+	if (err) {
+		nfp_prog->error = err;
+		return;
+	}
+
+	__emit_br_byte(nfp_prog, reg.areg, reg.breg, reg.i8, byte, false, addr,
+		       defer);
+}
+
+static void
+__emit_immed(struct nfp_prog *nfp_prog, u16 areg, u16 breg, u16 imm_hi,
+	     enum immed_width width, bool invert,
+	     enum immed_shift shift, bool wr_both)
+{
+	u64 insn;
+
+	insn = OP_IMMED_BASE |
+		FIELD_PREP(OP_IMMED_A_SRC, areg) |
+		FIELD_PREP(OP_IMMED_B_SRC, breg) |
+		FIELD_PREP(OP_IMMED_IMM, imm_hi) |
+		FIELD_PREP(OP_IMMED_WIDTH, width) |
+		FIELD_PREP(OP_IMMED_INV, invert) |
+		FIELD_PREP(OP_IMMED_SHIFT, shift) |
+		FIELD_PREP(OP_IMMED_WR_AB, wr_both);
+
+	nfp_prog_push(nfp_prog, insn);
+}
+
+static void
+emit_immed(struct nfp_prog *nfp_prog, u32 dst, u16 imm,
+	   enum immed_width width, bool invert, enum immed_shift shift)
+{
+	struct nfp_insn_ur_regs reg;
+	int err;
+
+	if (FIELD_GET(NN_REG_TYPE, dst) == NN_REG_IMM) {
+		nfp_prog->error = -EFAULT;
+		return;
+	}
+
+	err = swreg_to_unrestricted(dst, dst, reg_imm(imm & 0xff), &reg);
+	if (err) {
+		nfp_prog->error = err;
+		return;
+	}
+
+	__emit_immed(nfp_prog, reg.areg, reg.breg, imm >> 8, width,
+		     invert, shift, reg.wr_both);
+}
+
+static void
+__emit_shf(struct nfp_prog *nfp_prog, u16 dst, enum alu_dst_ab dst_ab,
+	   enum shf_sc sc, u8 shift,
+	   u16 areg, enum shf_op op, u16 breg, bool i8, bool sw, bool wr_both)
+{
+	u64 insn;
+
+	if (!FIELD_FIT(OP_SHF_SHIFT, shift)) {
+		nfp_prog->error = -EFAULT;
+		return;
+	}
+
+	if (sc == SHF_SC_L_SHF)
+		shift = 32 - shift;
+
+	insn = OP_SHF_BASE |
+		FIELD_PREP(OP_SHF_A_SRC, areg) |
+		FIELD_PREP(OP_SHF_SC, sc) |
+		FIELD_PREP(OP_SHF_B_SRC, breg) |
+		FIELD_PREP(OP_SHF_I8, i8) |
+		FIELD_PREP(OP_SHF_SW, sw) |
+		FIELD_PREP(OP_SHF_DST, dst) |
+		FIELD_PREP(OP_SHF_SHIFT, shift) |
+		FIELD_PREP(OP_SHF_OP, op) |
+		FIELD_PREP(OP_SHF_DST_AB, dst_ab) |
+		FIELD_PREP(OP_SHF_WR_AB, wr_both);
+
+	nfp_prog_push(nfp_prog, insn);
+}
+
+static void
+emit_shf(struct nfp_prog *nfp_prog, u32 dst, u32 lreg, enum shf_op op, u32 rreg,
+	 enum shf_sc sc, u8 shift)
+{
+	struct nfp_insn_re_regs reg;
+	int err;
+
+	err = swreg_to_restricted(dst, lreg, rreg, &reg, true);
+	if (err) {
+		nfp_prog->error = err;
+		return;
+	}
+
+	__emit_shf(nfp_prog, reg.dst, reg.dst_ab, sc, shift,
+		   reg.areg, op, reg.breg, reg.i8, reg.swap, reg.wr_both);
+}
+
+static void
+__emit_alu(struct nfp_prog *nfp_prog, u16 dst, enum alu_dst_ab dst_ab,
+	   u16 areg, enum alu_op op, u16 breg, bool swap, bool wr_both)
+{
+	u64 insn;
+
+	insn = OP_ALU_BASE |
+		FIELD_PREP(OP_ALU_A_SRC, areg) |
+		FIELD_PREP(OP_ALU_B_SRC, breg) |
+		FIELD_PREP(OP_ALU_DST, dst) |
+		FIELD_PREP(OP_ALU_SW, swap) |
+		FIELD_PREP(OP_ALU_OP, op) |
+		FIELD_PREP(OP_ALU_DST_AB, dst_ab) |
+		FIELD_PREP(OP_ALU_WR_AB, wr_both);
+
+	nfp_prog_push(nfp_prog, insn);
+}
+
+static void
+emit_alu(struct nfp_prog *nfp_prog, u32 dst, u32 lreg, enum alu_op op, u32 rreg)
+{
+	struct nfp_insn_ur_regs reg;
+	int err;
+
+	err = swreg_to_unrestricted(dst, lreg, rreg, &reg);
+	if (err) {
+		nfp_prog->error = err;
+		return;
+	}
+
+	__emit_alu(nfp_prog, reg.dst, reg.dst_ab,
+		   reg.areg, op, reg.breg, reg.swap, reg.wr_both);
+}
+
+static void
+__emit_ld_field(struct nfp_prog *nfp_prog, enum shf_sc sc,
+		u8 areg, u8 bmask, u8 breg, u8 shift, bool imm8,
+		bool zero, bool swap, bool wr_both)
+{
+	u64 insn;
+
+	insn = OP_LDF_BASE |
+		FIELD_PREP(OP_LDF_A_SRC, areg) |
+		FIELD_PREP(OP_LDF_SC, sc) |
+		FIELD_PREP(OP_LDF_B_SRC, breg) |
+		FIELD_PREP(OP_LDF_I8, imm8) |
+		FIELD_PREP(OP_LDF_SW, swap) |
+		FIELD_PREP(OP_LDF_ZF, zero) |
+		FIELD_PREP(OP_LDF_BMASK, bmask) |
+		FIELD_PREP(OP_LDF_SHF, shift) |
+		FIELD_PREP(OP_LDF_WR_AB, wr_both);
+
+	nfp_prog_push(nfp_prog, insn);
+}
+
+static void
+emit_ld_field_any(struct nfp_prog *nfp_prog, enum shf_sc sc, u8 shift,
+		  u32 dst, u8 bmask, u32 src, bool zero)
+{
+	struct nfp_insn_re_regs reg;
+	int err;
+
+	err = swreg_to_restricted(reg_none(), dst, src, &reg, true);
+	if (err) {
+		nfp_prog->error = err;
+		return;
+	}
+
+	__emit_ld_field(nfp_prog, sc, reg.areg, bmask, reg.breg, shift,
+			reg.i8, zero, reg.swap, reg.wr_both);
+}
+
+static void
+emit_ld_field(struct nfp_prog *nfp_prog, u32 dst, u8 bmask, u32 src,
+	      enum shf_sc sc, u8 shift)
+{
+	emit_ld_field_any(nfp_prog, sc, shift, dst, bmask, src, false);
+}
+
+/* --- Wrappers --- */
+static bool pack_immed(u32 imm, u16 *val, enum immed_shift *shift)
+{
+	if (!(imm & 0xffff0000)) {
+		*val = imm;
+		*shift = IMMED_SHIFT_0B;
+	} else if (!(imm & 0xff0000ff)) {
+		*val = imm >> 8;
+		*shift = IMMED_SHIFT_1B;
+	} else if (!(imm & 0x0000ffff)) {
+		*val = imm >> 16;
+		*shift = IMMED_SHIFT_2B;
+	} else {
+		return false;
+	}
+
+	return true;
+}
+
+static void wrp_immed(struct nfp_prog *nfp_prog, u32 dst, u32 imm)
+{
+	enum immed_shift shift;
+	u16 val;
+
+	if (pack_immed(imm, &val, &shift)) {
+		emit_immed(nfp_prog, dst, val, IMMED_WIDTH_ALL, false, shift);
+	} else if (pack_immed(~imm, &val, &shift)) {
+		emit_immed(nfp_prog, dst, val, IMMED_WIDTH_ALL, true, shift);
+	} else {
+		emit_immed(nfp_prog, dst, imm & 0xffff, IMMED_WIDTH_ALL,
+			   false, IMMED_SHIFT_0B);
+		emit_immed(nfp_prog, dst, imm >> 16, IMMED_WIDTH_WORD,
+			   false, IMMED_SHIFT_2B);
+	}
+}
+
+/* ur_load_imm_any() - encode immediate or use tmp register (unrestricted)
+ * If the @imm is small enough encode it directly in operand and return
+ * otherwise load @imm to a spare register and return its encoding.
+ */
+static u32 ur_load_imm_any(struct nfp_prog *nfp_prog, u32 imm, u32 tmp_reg)
+{
+	if (FIELD_FIT(UR_REG_IMM_MAX, imm))
+		return reg_imm(imm);
+
+	wrp_immed(nfp_prog, tmp_reg, imm);
+	return tmp_reg;
+}
+
+/* re_load_imm_any() - encode immediate or use tmp register (restricted)
+ * If the @imm is small enough encode it directly in operand and return
+ * otherwise load @imm to a spare register and return its encoding.
+ */
+static u32 re_load_imm_any(struct nfp_prog *nfp_prog, u32 imm, u32 tmp_reg)
+{
+	if (FIELD_FIT(RE_REG_IMM_MAX, imm))
+		return reg_imm(imm);
+
+	wrp_immed(nfp_prog, tmp_reg, imm);
+	return tmp_reg;
+}
+
+static void
+wrp_br_special(struct nfp_prog *nfp_prog, enum br_mask mask,
+	       enum br_special special)
+{
+	emit_br(nfp_prog, mask, 0, 0);
+
+	nfp_prog->prog[nfp_prog->prog_len - 1] |=
+		FIELD_PREP(OP_BR_SPECIAL, special);
+}
+
+static void wrp_reg_mov(struct nfp_prog *nfp_prog, u16 dst, u16 src)
+{
+	emit_alu(nfp_prog, reg_both(dst), reg_none(), ALU_OP_NONE, reg_b(src));
+}
+
+static int
+construct_data_ind_ld(struct nfp_prog *nfp_prog, u16 offset,
+		      u16 src, bool src_valid, u8 size)
+{
+	unsigned int i;
+	u16 shift, sz;
+	u32 tmp_reg;
+
+	/* We load the value from the address indicated in @offset and then
+	 * shift out the data we don't need.  Note: this is big endian!
+	 */
+	sz = size < 4 ? 4 : size;
+	shift = size < 4 ? 4 - size : 0;
+
+	if (src_valid) {
+		/* Calculate the true offset (src_reg + imm) */
+		tmp_reg = ur_load_imm_any(nfp_prog, offset, imm_b(nfp_prog));
+		emit_alu(nfp_prog, imm_both(nfp_prog),
+			 reg_a(src), ALU_OP_ADD, tmp_reg);
+		/* Check packet length (size guaranteed to fit b/c it's u8) */
+		emit_alu(nfp_prog, imm_a(nfp_prog),
+			 imm_a(nfp_prog), ALU_OP_ADD, reg_imm(size));
+		emit_alu(nfp_prog, reg_none(),
+			 NFP_BPF_ABI_LEN, ALU_OP_SUB, imm_a(nfp_prog));
+		wrp_br_special(nfp_prog, BR_BLO, OP_BR_GO_ABORT);
+		/* Load data */
+		emit_cmd(nfp_prog, CMD_TGT_READ8, CMD_MODE_32b, 0,
+			 pkt_reg(nfp_prog), imm_b(nfp_prog), sz - 1, true);
+	} else {
+		/* Check packet length */
+		tmp_reg = ur_load_imm_any(nfp_prog, offset + size,
+					  imm_a(nfp_prog));
+		emit_alu(nfp_prog, reg_none(),
+			 NFP_BPF_ABI_LEN, ALU_OP_SUB, tmp_reg);
+		wrp_br_special(nfp_prog, BR_BLO, OP_BR_GO_ABORT);
+		/* Load data */
+		tmp_reg = re_load_imm_any(nfp_prog, offset, imm_b(nfp_prog));
+		emit_cmd(nfp_prog, CMD_TGT_READ8, CMD_MODE_32b, 0,
+			 pkt_reg(nfp_prog), tmp_reg, sz - 1, true);
+	}
+
+	i = 0;
+	if (shift)
+		emit_shf(nfp_prog, reg_both(0), reg_none(), SHF_OP_NONE,
+			 reg_xfer(0), SHF_SC_R_SHF, shift * 8);
+	else
+		for (; i * 4 < size; i++)
+			emit_alu(nfp_prog, reg_both(i),
+				 reg_none(), ALU_OP_NONE, reg_xfer(i));
+
+	if (i < 2)
+		wrp_immed(nfp_prog, reg_both(1), 0);
+
+	return 0;
+}
+
+static int construct_data_ld(struct nfp_prog *nfp_prog, u16 offset, u8 size)
+{
+	return construct_data_ind_ld(nfp_prog, offset, 0, false, size);
+}
+
+static void
+wrp_alu_imm(struct nfp_prog *nfp_prog, u8 dst, enum alu_op alu_op, u32 imm)
+{
+	u32 tmp_reg;
+
+	if (alu_op == ALU_OP_AND) {
+		if (!imm)
+			wrp_immed(nfp_prog, reg_both(dst), 0);
+		if (!imm || !~imm)
+			return;
+	}
+	if (alu_op == ALU_OP_OR) {
+		if (!~imm)
+			wrp_immed(nfp_prog, reg_both(dst), ~0U);
+		if (!imm || !~imm)
+			return;
+	}
+	if (alu_op == ALU_OP_XOR) {
+		if (!~imm)
+			emit_alu(nfp_prog, reg_both(dst), reg_none(),
+				 ALU_OP_NEG, reg_b(dst));
+		if (!imm || !~imm)
+			return;
+	}
+
+	tmp_reg = ur_load_imm_any(nfp_prog, imm, imm_b(nfp_prog));
+	emit_alu(nfp_prog, reg_both(dst), reg_a(dst), alu_op, tmp_reg);
+}
+
+static int
+wrp_alu64_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+	      enum alu_op alu_op, bool skip)
+{
+	const struct bpf_insn *insn = &meta->insn;
+	u64 imm = insn->imm; /* sign extend */
+
+	if (skip) {
+		meta->skip = true;
+		return 0;
+	}
+
+	wrp_alu_imm(nfp_prog, insn->dst_reg * 2, alu_op, imm & ~0U);
+	wrp_alu_imm(nfp_prog, insn->dst_reg * 2 + 1, alu_op, imm >> 32);
+
+	return 0;
+}
+
+static int
+wrp_alu64_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+	      enum alu_op alu_op)
+{
+	u8 dst = meta->insn.dst_reg * 2, src = meta->insn.src_reg * 2;
+
+	emit_alu(nfp_prog, reg_both(dst), reg_a(dst), alu_op, reg_b(src));
+	emit_alu(nfp_prog, reg_both(dst + 1),
+		 reg_a(dst + 1), alu_op, reg_b(src + 1));
+
+	return 0;
+}
+
+static int
+wrp_alu32_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+	      enum alu_op alu_op, bool skip)
+{
+	const struct bpf_insn *insn = &meta->insn;
+
+	if (skip) {
+		meta->skip = true;
+		return 0;
+	}
+
+	wrp_alu_imm(nfp_prog, insn->dst_reg * 2, alu_op, insn->imm);
+	wrp_immed(nfp_prog, reg_both(insn->dst_reg * 2 + 1), 0);
+
+	return 0;
+}
+
+static int
+wrp_alu32_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+	      enum alu_op alu_op)
+{
+	u8 dst = meta->insn.dst_reg * 2, src = meta->insn.src_reg * 2;
+
+	emit_alu(nfp_prog, reg_both(dst), reg_a(dst), alu_op, reg_b(src));
+	wrp_immed(nfp_prog, reg_both(meta->insn.dst_reg * 2 + 1), 0);
+
+	return 0;
+}
+
+static void
+wrp_test_reg_one(struct nfp_prog *nfp_prog, u8 dst, enum alu_op alu_op, u8 src,
+		 enum br_mask br_mask, u16 off)
+{
+	emit_alu(nfp_prog, reg_none(), reg_a(dst), alu_op, reg_b(src));
+	emit_br(nfp_prog, br_mask, off, 0);
+}
+
+static int
+wrp_test_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+	     enum alu_op alu_op, enum br_mask br_mask)
+{
+	const struct bpf_insn *insn = &meta->insn;
+
+	if (insn->off < 0) /* TODO */
+		return -ENOTSUPP;
+
+	wrp_test_reg_one(nfp_prog, insn->dst_reg * 2, alu_op,
+			 insn->src_reg * 2, br_mask, insn->off);
+	wrp_test_reg_one(nfp_prog, insn->dst_reg * 2 + 1, alu_op,
+			 insn->src_reg * 2 + 1, br_mask, insn->off);
+
+	return 0;
+}
+
+static int
+wrp_cmp_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+	    enum br_mask br_mask, bool swap)
+{
+	const struct bpf_insn *insn = &meta->insn;
+	u64 imm = insn->imm; /* sign extend */
+	u8 reg = insn->dst_reg * 2;
+	u32 tmp_reg;
+
+	if (insn->off < 0) /* TODO */
+		return -ENOTSUPP;
+
+	tmp_reg = ur_load_imm_any(nfp_prog, imm & ~0U, imm_b(nfp_prog));
+	if (!swap)
+		emit_alu(nfp_prog, reg_none(), reg_a(reg), ALU_OP_SUB, tmp_reg);
+	else
+		emit_alu(nfp_prog, reg_none(), tmp_reg, ALU_OP_SUB, reg_a(reg));
+
+	tmp_reg = ur_load_imm_any(nfp_prog, imm >> 32, imm_b(nfp_prog));
+	if (!swap)
+		emit_alu(nfp_prog, reg_none(),
+			 reg_a(reg + 1), ALU_OP_SUB_C, tmp_reg);
+	else
+		emit_alu(nfp_prog, reg_none(),
+			 tmp_reg, ALU_OP_SUB_C, reg_a(reg + 1));
+
+	emit_br(nfp_prog, br_mask, insn->off, 0);
+
+	return 0;
+}
+
+static int
+wrp_cmp_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+	    enum br_mask br_mask, bool swap)
+{
+	const struct bpf_insn *insn = &meta->insn;
+	u8 areg = insn->src_reg * 2, breg = insn->dst_reg * 2;
+
+	if (insn->off < 0) /* TODO */
+		return -ENOTSUPP;
+
+	if (swap) {
+		areg ^= breg;
+		breg ^= areg;
+		areg ^= breg;
+	}
+
+	emit_alu(nfp_prog, reg_none(), reg_a(areg), ALU_OP_SUB, reg_b(breg));
+	emit_alu(nfp_prog, reg_none(),
+		 reg_a(areg + 1), ALU_OP_SUB_C, reg_b(breg + 1));
+	emit_br(nfp_prog, br_mask, insn->off, 0);
+
+	return 0;
+}
+
+/* --- Callbacks --- */
+static int mov_reg64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+
+	wrp_reg_mov(nfp_prog, insn->dst_reg * 2, insn->src_reg * 2);
+	wrp_reg_mov(nfp_prog, insn->dst_reg * 2 + 1, insn->src_reg * 2 + 1);
+
+	return 0;
+}
+
+static int mov_imm64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	u64 imm = meta->insn.imm; /* sign extend */
+
+	wrp_immed(nfp_prog, reg_both(meta->insn.dst_reg * 2), imm & ~0U);
+	wrp_immed(nfp_prog, reg_both(meta->insn.dst_reg * 2 + 1), imm >> 32);
+
+	return 0;
+}
+
+static int xor_reg64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu64_reg(nfp_prog, meta, ALU_OP_XOR);
+}
+
+static int xor_imm64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu64_imm(nfp_prog, meta, ALU_OP_XOR, !meta->insn.imm);
+}
+
+static int and_reg64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu64_reg(nfp_prog, meta, ALU_OP_AND);
+}
+
+static int and_imm64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu64_imm(nfp_prog, meta, ALU_OP_AND, !~meta->insn.imm);
+}
+
+static int or_reg64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu64_reg(nfp_prog, meta, ALU_OP_OR);
+}
+
+static int or_imm64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu64_imm(nfp_prog, meta, ALU_OP_OR, !meta->insn.imm);
+}
+
+static int add_reg64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+
+	emit_alu(nfp_prog, reg_both(insn->dst_reg * 2),
+		 reg_a(insn->dst_reg * 2), ALU_OP_ADD,
+		 reg_b(insn->src_reg * 2));
+	emit_alu(nfp_prog, reg_both(insn->dst_reg * 2 + 1),
+		 reg_a(insn->dst_reg * 2 + 1), ALU_OP_ADD_C,
+		 reg_b(insn->src_reg * 2 + 1));
+
+	return 0;
+}
+
+static int add_imm64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+	u64 imm = insn->imm; /* sign extend */
+
+	wrp_alu_imm(nfp_prog, insn->dst_reg * 2, ALU_OP_ADD, imm & ~0U);
+	wrp_alu_imm(nfp_prog, insn->dst_reg * 2 + 1, ALU_OP_ADD_C, imm >> 32);
+
+	return 0;
+}
+
+static int sub_reg64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+
+	emit_alu(nfp_prog, reg_both(insn->dst_reg * 2),
+		 reg_a(insn->dst_reg * 2), ALU_OP_SUB,
+		 reg_b(insn->src_reg * 2));
+	emit_alu(nfp_prog, reg_both(insn->dst_reg * 2 + 1),
+		 reg_a(insn->dst_reg * 2 + 1), ALU_OP_SUB_C,
+		 reg_b(insn->src_reg * 2 + 1));
+
+	return 0;
+}
+
+static int sub_imm64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+	u64 imm = insn->imm; /* sign extend */
+
+	wrp_alu_imm(nfp_prog, insn->dst_reg * 2, ALU_OP_SUB, imm & ~0U);
+	wrp_alu_imm(nfp_prog, insn->dst_reg * 2 + 1, ALU_OP_SUB_C, imm >> 32);
+
+	return 0;
+}
+
+static int shl_imm64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+
+	if (insn->imm != 32)
+		return 1; /* TODO */
+
+	wrp_reg_mov(nfp_prog, insn->dst_reg * 2 + 1, insn->dst_reg * 2);
+	wrp_immed(nfp_prog, reg_both(insn->dst_reg * 2), 0);
+
+	return 0;
+}
+
+static int shr_imm64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+
+	if (insn->imm != 32)
+		return 1; /* TODO */
+
+	wrp_reg_mov(nfp_prog, insn->dst_reg * 2, insn->dst_reg * 2 + 1);
+	wrp_immed(nfp_prog, reg_both(insn->dst_reg * 2 + 1), 0);
+
+	return 0;
+}
+
+static int mov_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+
+	wrp_reg_mov(nfp_prog, insn->dst_reg * 2,  insn->src_reg * 2);
+	wrp_immed(nfp_prog, reg_both(insn->dst_reg * 2 + 1), 0);
+
+	return 0;
+}
+
+static int mov_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+
+	wrp_immed(nfp_prog, reg_both(insn->dst_reg * 2), insn->imm);
+	wrp_immed(nfp_prog, reg_both(insn->dst_reg * 2 + 1), 0);
+
+	return 0;
+}
+
+static int xor_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu32_reg(nfp_prog, meta, ALU_OP_XOR);
+}
+
+static int xor_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu32_imm(nfp_prog, meta, ALU_OP_XOR, !~meta->insn.imm);
+}
+
+static int and_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu32_reg(nfp_prog, meta, ALU_OP_AND);
+}
+
+static int and_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu32_imm(nfp_prog, meta, ALU_OP_AND, !~meta->insn.imm);
+}
+
+static int or_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu32_reg(nfp_prog, meta, ALU_OP_OR);
+}
+
+static int or_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu32_imm(nfp_prog, meta, ALU_OP_OR, !meta->insn.imm);
+}
+
+static int add_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu32_reg(nfp_prog, meta, ALU_OP_ADD);
+}
+
+static int add_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu32_imm(nfp_prog, meta, ALU_OP_ADD, !meta->insn.imm);
+}
+
+static int sub_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu32_reg(nfp_prog, meta, ALU_OP_SUB);
+}
+
+static int sub_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_alu32_imm(nfp_prog, meta, ALU_OP_SUB, !meta->insn.imm);
+}
+
+static int shl_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+
+	if (!insn->imm)
+		return 1; /* TODO: zero shift means indirect */
+
+	emit_shf(nfp_prog, reg_both(insn->dst_reg * 2),
+		 reg_none(), SHF_OP_NONE, reg_b(insn->dst_reg * 2),
+		 SHF_SC_L_SHF, insn->imm);
+	wrp_immed(nfp_prog, reg_both(insn->dst_reg * 2 + 1), 0);
+
+	return 0;
+}
+
+static int imm_ld8_part2(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	wrp_immed(nfp_prog, reg_both(nfp_meta_prev(meta)->insn.dst_reg * 2 + 1),
+		  meta->insn.imm);
+
+	return 0;
+}
+
+static int imm_ld8(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+
+	meta->double_cb = imm_ld8_part2;
+	wrp_immed(nfp_prog, reg_both(insn->dst_reg * 2), insn->imm);
+
+	return 0;
+}
+
+static int data_ld1(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return construct_data_ld(nfp_prog, meta->insn.imm, 1);
+}
+
+static int data_ld2(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return construct_data_ld(nfp_prog, meta->insn.imm, 2);
+}
+
+static int data_ld4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return construct_data_ld(nfp_prog, meta->insn.imm, 4);
+}
+
+static int data_ind_ld1(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return construct_data_ind_ld(nfp_prog, meta->insn.imm,
+				     meta->insn.src_reg * 2, true, 1);
+}
+
+static int data_ind_ld2(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return construct_data_ind_ld(nfp_prog, meta->insn.imm,
+				     meta->insn.src_reg * 2, true, 2);
+}
+
+static int data_ind_ld4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return construct_data_ind_ld(nfp_prog, meta->insn.imm,
+				     meta->insn.src_reg * 2, true, 4);
+}
+
+static int mem_ldx4(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	if (meta->insn.off == offsetof(struct sk_buff, len))
+		emit_alu(nfp_prog, reg_both(meta->insn.dst_reg * 2),
+			 reg_none(), ALU_OP_NONE, NFP_BPF_ABI_LEN);
+	else
+		return -ENOTSUPP;
+
+	return 0;
+}
+
+static int jump(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	if (meta->insn.off < 0) /* TODO */
+		return -ENOTSUPP;
+	emit_br(nfp_prog, BR_UNC, meta->insn.off, 0);
+
+	return 0;
+}
+
+static int jeq_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+	u64 imm = insn->imm; /* sign extend */
+	u32 or1 = reg_a(insn->dst_reg * 2), or2 = reg_b(insn->dst_reg * 2 + 1);
+	u32 tmp_reg;
+
+	if (insn->off < 0) /* TODO */
+		return -ENOTSUPP;
+
+	if (imm & ~0U) {
+		tmp_reg = ur_load_imm_any(nfp_prog, imm & ~0U, imm_b(nfp_prog));
+		emit_alu(nfp_prog, imm_a(nfp_prog),
+			 reg_a(insn->dst_reg * 2), ALU_OP_XOR, tmp_reg);
+		or1 = imm_a(nfp_prog);
+	}
+
+	if (imm >> 32) {
+		tmp_reg = ur_load_imm_any(nfp_prog, imm >> 32, imm_b(nfp_prog));
+		emit_alu(nfp_prog, imm_b(nfp_prog),
+			 reg_a(insn->dst_reg * 2 + 1), ALU_OP_XOR, tmp_reg);
+		or2 = imm_b(nfp_prog);
+	}
+
+	emit_alu(nfp_prog, reg_none(), or1, ALU_OP_OR, or2);
+	emit_br(nfp_prog, BR_BEQ, insn->off, 0);
+
+	return 0;
+}
+
+static int jgt_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_cmp_imm(nfp_prog, meta, BR_BLO, false);
+}
+
+static int jge_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_cmp_imm(nfp_prog, meta, BR_BHS, true);
+}
+
+static int jset_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+	u64 imm = insn->imm; /* sign extend */
+	u32 tmp_reg;
+
+	if (insn->off < 0) /* TODO */
+		return -ENOTSUPP;
+
+	if (!imm) {
+		meta->skip = true;
+		return 0;
+	}
+
+	if (imm & ~0U) {
+		tmp_reg = ur_load_imm_any(nfp_prog, imm & ~0U, imm_b(nfp_prog));
+		emit_alu(nfp_prog, reg_none(),
+			 reg_a(insn->dst_reg * 2), ALU_OP_AND, tmp_reg);
+		emit_br(nfp_prog, BR_BNE, insn->off, 0);
+	}
+
+	if (imm >> 32) {
+		tmp_reg = ur_load_imm_any(nfp_prog, imm >> 32, imm_b(nfp_prog));
+		emit_alu(nfp_prog, reg_none(),
+			 reg_a(insn->dst_reg * 2 + 1), ALU_OP_AND, tmp_reg);
+		emit_br(nfp_prog, BR_BNE, insn->off, 0);
+	}
+
+	return 0;
+}
+
+static int jne_imm(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+	u64 imm = insn->imm; /* sign extend */
+	u32 tmp_reg;
+
+	if (insn->off < 0) /* TODO */
+		return -ENOTSUPP;
+
+	if (!imm) {
+		emit_alu(nfp_prog, reg_none(), reg_a(insn->dst_reg * 2),
+			 ALU_OP_OR, reg_b(insn->dst_reg * 2 + 1));
+		emit_br(nfp_prog, BR_BNE, insn->off, 0);
+	}
+
+	tmp_reg = ur_load_imm_any(nfp_prog, imm & ~0U, imm_b(nfp_prog));
+	emit_alu(nfp_prog, reg_none(),
+		 reg_a(insn->dst_reg * 2), ALU_OP_XOR, tmp_reg);
+	emit_br(nfp_prog, BR_BNE, insn->off, 0);
+
+	tmp_reg = ur_load_imm_any(nfp_prog, imm >> 32, imm_b(nfp_prog));
+	emit_alu(nfp_prog, reg_none(),
+		 reg_a(insn->dst_reg * 2 + 1), ALU_OP_XOR, tmp_reg);
+	emit_br(nfp_prog, BR_BNE, insn->off, 0);
+
+	return 0;
+}
+
+static int jeq_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	const struct bpf_insn *insn = &meta->insn;
+
+	if (insn->off < 0) /* TODO */
+		return -ENOTSUPP;
+
+	emit_alu(nfp_prog, imm_a(nfp_prog), reg_a(insn->dst_reg * 2),
+		 ALU_OP_XOR, reg_b(insn->src_reg * 2));
+	emit_alu(nfp_prog, imm_b(nfp_prog), reg_a(insn->dst_reg * 2 + 1),
+		 ALU_OP_XOR, reg_b(insn->src_reg * 2 + 1));
+	emit_alu(nfp_prog, reg_none(),
+		 imm_a(nfp_prog), ALU_OP_OR, imm_b(nfp_prog));
+	emit_br(nfp_prog, BR_BEQ, insn->off, 0);
+
+	return 0;
+}
+
+static int jgt_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_cmp_reg(nfp_prog, meta, BR_BLO, false);
+}
+
+static int jge_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_cmp_reg(nfp_prog, meta, BR_BHS, true);
+}
+
+static int jset_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_test_reg(nfp_prog, meta, ALU_OP_AND, BR_BNE);
+}
+
+static int jne_reg(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	return wrp_test_reg(nfp_prog, meta, ALU_OP_XOR, BR_BNE);
+}
+
+static int goto_out(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	wrp_br_special(nfp_prog, BR_UNC, OP_BR_GO_OUT);
+
+	return 0;
+}
+
+static const instr_cb_t instr_cb[256] = {
+	[BPF_ALU64 | BPF_MOV | BPF_X] =	mov_reg64,
+	[BPF_ALU64 | BPF_MOV | BPF_K] =	mov_imm64,
+	[BPF_ALU64 | BPF_XOR | BPF_X] =	xor_reg64,
+	[BPF_ALU64 | BPF_XOR | BPF_K] =	xor_imm64,
+	[BPF_ALU64 | BPF_AND | BPF_X] =	and_reg64,
+	[BPF_ALU64 | BPF_AND | BPF_K] =	and_imm64,
+	[BPF_ALU64 | BPF_OR | BPF_X] =	or_reg64,
+	[BPF_ALU64 | BPF_OR | BPF_K] =	or_imm64,
+	[BPF_ALU64 | BPF_ADD | BPF_X] =	add_reg64,
+	[BPF_ALU64 | BPF_ADD | BPF_K] =	add_imm64,
+	[BPF_ALU64 | BPF_SUB | BPF_X] =	sub_reg64,
+	[BPF_ALU64 | BPF_SUB | BPF_K] =	sub_imm64,
+	[BPF_ALU64 | BPF_LSH | BPF_K] =	shl_imm64,
+	[BPF_ALU64 | BPF_RSH | BPF_K] =	shr_imm64,
+	[BPF_ALU | BPF_MOV | BPF_X] =	mov_reg,
+	[BPF_ALU | BPF_MOV | BPF_K] =	mov_imm,
+	[BPF_ALU | BPF_XOR | BPF_X] =	xor_reg,
+	[BPF_ALU | BPF_XOR | BPF_K] =	xor_imm,
+	[BPF_ALU | BPF_AND | BPF_X] =	and_reg,
+	[BPF_ALU | BPF_AND | BPF_K] =	and_imm,
+	[BPF_ALU | BPF_OR | BPF_X] =	or_reg,
+	[BPF_ALU | BPF_OR | BPF_K] =	or_imm,
+	[BPF_ALU | BPF_ADD | BPF_X] =	add_reg,
+	[BPF_ALU | BPF_ADD | BPF_K] =	add_imm,
+	[BPF_ALU | BPF_SUB | BPF_X] =	sub_reg,
+	[BPF_ALU | BPF_SUB | BPF_K] =	sub_imm,
+	[BPF_ALU | BPF_LSH | BPF_K] =	shl_imm,
+	[BPF_LD | BPF_IMM | BPF_DW] =	imm_ld8,
+	[BPF_LD | BPF_ABS | BPF_B] =	data_ld1,
+	[BPF_LD | BPF_ABS | BPF_H] =	data_ld2,
+	[BPF_LD | BPF_ABS | BPF_W] =	data_ld4,
+	[BPF_LD | BPF_IND | BPF_B] =	data_ind_ld1,
+	[BPF_LD | BPF_IND | BPF_H] =	data_ind_ld2,
+	[BPF_LD | BPF_IND | BPF_W] =	data_ind_ld4,
+	[BPF_LDX | BPF_MEM | BPF_W] =	mem_ldx4,
+	[BPF_JMP | BPF_JA | BPF_K] =	jump,
+	[BPF_JMP | BPF_JEQ | BPF_K] =	jeq_imm,
+	[BPF_JMP | BPF_JGT | BPF_K] =	jgt_imm,
+	[BPF_JMP | BPF_JGE | BPF_K] =	jge_imm,
+	[BPF_JMP | BPF_JSET | BPF_K] =	jset_imm,
+	[BPF_JMP | BPF_JNE | BPF_K] =	jne_imm,
+	[BPF_JMP | BPF_JEQ | BPF_X] =	jeq_reg,
+	[BPF_JMP | BPF_JGT | BPF_X] =	jgt_reg,
+	[BPF_JMP | BPF_JGE | BPF_X] =	jge_reg,
+	[BPF_JMP | BPF_JSET | BPF_X] =	jset_reg,
+	[BPF_JMP | BPF_JNE | BPF_X] =	jne_reg,
+	[BPF_JMP | BPF_EXIT] =		goto_out,
+};
+
+/* --- Misc code --- */
+static void br_set_offset(u64 *instr, u16 offset)
+{
+	u16 addr_lo, addr_hi;
+
+	addr_lo = offset & (OP_BR_ADDR_LO >> __bf_shf(OP_BR_ADDR_LO));
+	addr_hi = offset != addr_lo;
+	*instr &= ~(OP_BR_ADDR_HI | OP_BR_ADDR_LO);
+	*instr |= FIELD_PREP(OP_BR_ADDR_HI, addr_hi);
+	*instr |= FIELD_PREP(OP_BR_ADDR_LO, addr_lo);
+}
+
+/* --- Assembler logic --- */
+static int nfp_fixup_branches(struct nfp_prog *nfp_prog)
+{
+	struct nfp_insn_meta *meta, *next;
+	u32 off, br_idx;
+	u32 idx;
+
+	nfp_for_each_insn_walk2(nfp_prog, meta, next) {
+		if (meta->skip)
+			continue;
+		if (BPF_CLASS(meta->insn.code) != BPF_JMP)
+			continue;
+
+		br_idx = nfp_prog_offset_to_index(nfp_prog, next->off) - 1;
+		if (!nfp_is_br(nfp_prog->prog[br_idx])) {
+			pr_err("Fixup found block not ending in branch %d %02x %016llx!!\n",
+			       br_idx, meta->insn.code, nfp_prog->prog[br_idx]);
+			return -ELOOP;
+		}
+		/* Leave special branches for later */
+		if (FIELD_GET(OP_BR_SPECIAL, nfp_prog->prog[br_idx]))
+			continue;
+
+		/* Find the target offset in assembler realm */
+		off = meta->insn.off;
+		if (!off) {
+			pr_err("Fixup found zero offset!!\n");
+			return -ELOOP;
+		}
+
+		while (off && nfp_meta_has_next(nfp_prog, next)) {
+			next = nfp_meta_next(next);
+			off--;
+		}
+		if (off) {
+			pr_err("Fixup found too large jump!! %d\n", off);
+			return -ELOOP;
+		}
+
+		if (next->skip) {
+			pr_err("Branch landing on removed instruction!!\n");
+			return -ELOOP;
+		}
+
+		for (idx = nfp_prog_offset_to_index(nfp_prog, meta->off);
+		     idx <= br_idx; idx++) {
+			if (!nfp_is_br(nfp_prog->prog[idx]))
+				continue;
+			br_set_offset(&nfp_prog->prog[idx], next->off);
+		}
+	}
+
+	/* Fixup 'goto out's separately, they can be scattered around */
+	for (br_idx = 0; br_idx < nfp_prog->prog_len; br_idx++) {
+		enum br_special special;
+
+		if ((nfp_prog->prog[br_idx] & OP_BR_BASE_MASK) != OP_BR_BASE)
+			continue;
+
+		special = FIELD_GET(OP_BR_SPECIAL, nfp_prog->prog[br_idx]);
+		switch (special) {
+		case OP_BR_NORMAL:
+			break;
+		case OP_BR_GO_OUT:
+			br_set_offset(&nfp_prog->prog[br_idx],
+				      nfp_prog->tgt_out);
+			break;
+		case OP_BR_GO_ABORT:
+			br_set_offset(&nfp_prog->prog[br_idx],
+				      nfp_prog->tgt_abort);
+			break;
+		}
+
+		nfp_prog->prog[br_idx] &= ~OP_BR_SPECIAL;
+	}
+
+	return 0;
+}
+
+static void nfp_intro(struct nfp_prog *nfp_prog)
+{
+	emit_alu(nfp_prog, pkt_reg(nfp_prog),
+		 reg_none(), ALU_OP_NONE, NFP_BPF_ABI_PKT);
+}
+
+static void nfp_outro_tc_legacy(struct nfp_prog *nfp_prog)
+{
+	const u8 act2code[] = {
+		[NN_ACT_TC_DROP]  = 0x22,
+	};
+	/* Target for aborts */
+	nfp_prog->tgt_abort = nfp_prog_current_offset(nfp_prog);
+	wrp_immed(nfp_prog, reg_both(0), 0);
+
+	/* Target for normal exits */
+	nfp_prog->tgt_out = nfp_prog_current_offset(nfp_prog);
+	/* Legacy TC mode:
+	 *   0        0x11 -> pass,  count as stat0
+	 *  -1  drop  0x22 -> drop,  count as stat1
+	 *     redir  0x24 -> redir, count as stat1
+	 *  ife mark  0x21 -> pass,  count as stat1
+	 *  ife + tx  0x24 -> redir, count as stat1
+	 */
+	emit_br_byte_neq(nfp_prog, reg_b(0), 0xff, 0, nfp_prog->tgt_done, 2);
+	emit_alu(nfp_prog, reg_a(0),
+		 reg_none(), ALU_OP_NONE, NFP_BPF_ABI_FLAGS);
+	emit_ld_field(nfp_prog, reg_a(0), 0xc, reg_imm(0x11), SHF_SC_L_SHF, 16);
+
+	emit_br(nfp_prog, BR_UNC, nfp_prog->tgt_done, 1);
+	emit_ld_field(nfp_prog, reg_a(0), 0xc, reg_imm(act2code[nfp_prog->act]),
+		      SHF_SC_L_SHF, 16);
+}
+
+static void nfp_outro(struct nfp_prog *nfp_prog)
+{
+	switch (nfp_prog->act) {
+	case NN_ACT_TC_DROP:
+		nfp_outro_tc_legacy(nfp_prog);
+		break;
+	}
+}
+
+static int nfp_translate(struct nfp_prog *nfp_prog)
+{
+	struct nfp_insn_meta *meta;
+	int err;
+
+	nfp_intro(nfp_prog);
+	if (nfp_prog->error)
+		return nfp_prog->error;
+
+	list_for_each_entry(meta, &nfp_prog->insns, l) {
+		instr_cb_t cb = instr_cb[meta->insn.code];
+
+		meta->off = nfp_prog_current_offset(nfp_prog);
+
+		if (meta->skip) {
+			nfp_prog->n_translated++;
+			continue;
+		}
+
+		if (nfp_meta_has_prev(nfp_prog, meta) &&
+		    nfp_meta_prev(meta)->double_cb)
+			cb = nfp_meta_prev(meta)->double_cb;
+		if (!cb)
+			return -ENOENT;
+		err = cb(nfp_prog, meta);
+		if (err)
+			return err;
+
+		nfp_prog->n_translated++;
+	}
+
+	nfp_outro(nfp_prog);
+	if (nfp_prog->error)
+		return nfp_prog->error;
+
+	return nfp_fixup_branches(nfp_prog);
+}
+
+static int
+nfp_prog_prepare(struct nfp_prog *nfp_prog, const struct bpf_insn *prog,
+		 unsigned int cnt)
+{
+	unsigned int i;
+
+	for (i = 0; i < cnt; i++) {
+		struct nfp_insn_meta *meta;
+
+		meta = kzalloc(sizeof(*meta), GFP_KERNEL);
+		if (!meta)
+			return -ENOMEM;
+
+		meta->insn = prog[i];
+		meta->n = i;
+
+		list_add_tail(&meta->l, &nfp_prog->insns);
+	}
+
+	return 0;
+}
+
+/* --- Optimizations --- */
+static void nfp_bpf_opt_reg_init(struct nfp_prog *nfp_prog)
+{
+	struct nfp_insn_meta *meta;
+
+	list_for_each_entry(meta, &nfp_prog->insns, l) {
+		struct bpf_insn insn = meta->insn;
+
+		/* Programs converted from cBPF start with register xoring */
+		if (insn.code == (BPF_ALU64 | BPF_XOR | BPF_X) &&
+		    insn.src_reg == insn.dst_reg)
+			continue;
+
+		/* Programs start with R6 = R1 but we ignore the skb pointer */
+		if (insn.code == (BPF_ALU64 | BPF_MOV | BPF_X) &&
+		    insn.src_reg == 1 && insn.dst_reg == 6)
+			meta->skip = true;
+
+		/* Return as soon as something doesn't match */
+		if (!meta->skip)
+			return;
+	}
+}
+
+/* Try to rename registers so that program uses only low ones */
+static int nfp_bpf_opt_reg_rename(struct nfp_prog *nfp_prog)
+{
+	bool reg_used[MAX_BPF_REG] = {};
+	u8 tgt_reg[MAX_BPF_REG] = {};
+	struct nfp_insn_meta *meta;
+	unsigned int i, j;
+
+	list_for_each_entry(meta, &nfp_prog->insns, l) {
+		if (meta->skip)
+			continue;
+
+		reg_used[meta->insn.src_reg] = true;
+		reg_used[meta->insn.dst_reg] = true;
+	}
+
+	for (i = 0, j = 0; i < ARRAY_SIZE(tgt_reg); i++) {
+		if (!reg_used[i])
+			continue;
+
+		tgt_reg[i] = j++;
+	}
+	nfp_prog->num_regs = j;
+
+	list_for_each_entry(meta, &nfp_prog->insns, l) {
+		meta->insn.src_reg = tgt_reg[meta->insn.src_reg];
+		meta->insn.dst_reg = tgt_reg[meta->insn.dst_reg];
+	}
+
+	return 0;
+}
+
+/* Remove masking after load since our load guarantees this is not needed */
+static void nfp_bpf_opt_ld_mask(struct nfp_prog *nfp_prog)
+{
+	struct nfp_insn_meta *meta1, *meta2;
+	const s32 exp_mask[] = {
+		[BPF_B] = 0x000000ffU,
+		[BPF_H] = 0x0000ffffU,
+		[BPF_W] = 0xffffffffU,
+	};
+
+	nfp_for_each_insn_walk2(nfp_prog, meta1, meta2) {
+		struct bpf_insn insn, next;
+
+		insn = meta1->insn;
+		next = meta2->insn;
+
+		if (BPF_CLASS(insn.code) != BPF_LD)
+			continue;
+		if (BPF_MODE(insn.code) != BPF_ABS &&
+		    BPF_MODE(insn.code) != BPF_IND)
+			continue;
+
+		if (next.code != (BPF_ALU64 | BPF_AND | BPF_K))
+			continue;
+
+		if (!exp_mask[BPF_SIZE(insn.code)])
+			continue;
+		if (exp_mask[BPF_SIZE(insn.code)] != next.imm)
+			continue;
+
+		if (next.src_reg || next.dst_reg)
+			continue;
+
+		meta2->skip = true;
+	}
+}
+
+static void nfp_bpf_opt_ld_shift(struct nfp_prog *nfp_prog)
+{
+	struct nfp_insn_meta *meta1, *meta2, *meta3;
+
+	nfp_for_each_insn_walk3(nfp_prog, meta1, meta2, meta3) {
+		struct bpf_insn insn, next1, next2;
+
+		insn = meta1->insn;
+		next1 = meta2->insn;
+		next2 = meta3->insn;
+
+		if (BPF_CLASS(insn.code) != BPF_LD)
+			continue;
+		if (BPF_MODE(insn.code) != BPF_ABS &&
+		    BPF_MODE(insn.code) != BPF_IND)
+			continue;
+		if (BPF_SIZE(insn.code) != BPF_W)
+			continue;
+
+		if (!(next1.code == (BPF_LSH | BPF_K | BPF_ALU64) &&
+		      next2.code == (BPF_RSH | BPF_K | BPF_ALU64)) &&
+		    !(next1.code == (BPF_RSH | BPF_K | BPF_ALU64) &&
+		      next2.code == (BPF_LSH | BPF_K | BPF_ALU64)))
+			continue;
+
+		if (next1.src_reg || next1.dst_reg ||
+		    next2.src_reg || next2.dst_reg)
+			continue;
+
+		if (next1.imm != 0x20 || next2.imm != 0x20)
+			continue;
+
+		meta2->skip = true;
+		meta3->skip = true;
+	}
+}
+
+static int nfp_bpf_optimize(struct nfp_prog *nfp_prog)
+{
+	int ret;
+
+	nfp_bpf_opt_reg_init(nfp_prog);
+
+	ret = nfp_bpf_opt_reg_rename(nfp_prog);
+	if (ret)
+		return ret;
+
+	nfp_bpf_opt_ld_mask(nfp_prog);
+	nfp_bpf_opt_ld_shift(nfp_prog);
+
+	return 0;
+}
+
+/**
+ * nfp_bpf_jit() - translate BPF code into NFP assembly
+ * @filter:	kernel BPF filter struct
+ * @prog_mem:	memory to store assembler instructions
+ * @act:	action attached to this eBPF program
+ * @prog_start:	offset of the first instruction when loaded
+ * @prog_done:	where to jump on exit
+ * @prog_sz:	size of @prog_mem in instructions
+ * @res:	achieved parameters of translation results
+ */
+int
+nfp_bpf_jit(struct bpf_prog *filter, void *prog_mem,
+	    enum nfp_bpf_action_type act,
+	    unsigned int prog_start, unsigned int prog_done,
+	    unsigned int prog_sz, struct nfp_bpf_result *res)
+{
+	struct nfp_prog *nfp_prog;
+	int ret;
+
+	nfp_prog = kzalloc(sizeof(*nfp_prog), GFP_KERNEL);
+	if (!nfp_prog)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&nfp_prog->insns);
+	nfp_prog->act = act;
+	nfp_prog->start_off = prog_start;
+	nfp_prog->tgt_done = prog_done;
+
+	ret = nfp_prog_prepare(nfp_prog, filter->insnsi, filter->len);
+	if (ret)
+		goto out;
+
+	ret = nfp_prog_verify(nfp_prog, filter);
+	if (ret)
+		goto out;
+
+	ret = nfp_bpf_optimize(nfp_prog);
+	if (ret)
+		goto out;
+
+	if (nfp_prog->num_regs <= 7)
+		nfp_prog->regs_per_thread = 16;
+	else
+		nfp_prog->regs_per_thread = 32;
+
+	nfp_prog->prog = prog_mem;
+	nfp_prog->__prog_alloc_len = prog_sz;
+
+	ret = nfp_translate(nfp_prog);
+	if (ret) {
+		pr_err("Translation failed with error %d (translated: %u)\n",
+		       ret, nfp_prog->n_translated);
+		ret = -EINVAL;
+	}
+
+	res->n_instr = nfp_prog->prog_len;
+	res->dense_mode = nfp_prog->num_regs <= 7;
+out:
+	nfp_prog_free(nfp_prog);
+
+	return ret;
+}
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_bpf_verifier.c b/drivers/net/ethernet/netronome/nfp/nfp_bpf_verifier.c
new file mode 100644
index 000000000000..ef6775b54168
--- /dev/null
+++ b/drivers/net/ethernet/netronome/nfp/nfp_bpf_verifier.c
@@ -0,0 +1,162 @@
+/*
+ * Copyright (C) 2016 Netronome Systems, Inc.
+ *
+ * This software is dual licensed under the GNU General License Version 2,
+ * June 1991 as shown in the file COPYING in the top-level directory of this
+ * source tree or the BSD 2-Clause License provided below.  You have the
+ * option to license this software under the complete terms of either license.
+ *
+ * The BSD 2-Clause License:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      1. Redistributions of source code must retain the above
+ *         copyright notice, this list of conditions and the following
+ *         disclaimer.
+ *
+ *      2. Redistributions in binary form must reproduce the above
+ *         copyright notice, this list of conditions and the following
+ *         disclaimer in the documentation and/or other materials
+ *         provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#define pr_fmt(fmt)	"NFP net bpf: " fmt
+
+#include <linux/bpf.h>
+#include <linux/bpf_verifier.h>
+#include <linux/kernel.h>
+#include <linux/pkt_cls.h>
+
+#include "nfp_bpf.h"
+
+/* Analyzer/verifier definitions */
+struct nfp_bpf_analyzer_priv {
+	struct nfp_prog *prog;
+	struct nfp_insn_meta *meta;
+};
+
+static struct nfp_insn_meta *
+nfp_bpf_goto_meta(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
+		  unsigned int insn_idx, unsigned int n_insns)
+{
+	unsigned int forward, backward, i;
+
+	backward = meta->n - insn_idx;
+	forward = insn_idx - meta->n;
+
+	if (min(forward, backward) > n_insns - insn_idx - 1) {
+		backward = n_insns - insn_idx - 1;
+		meta = nfp_prog_last_meta(nfp_prog);
+	}
+	if (min(forward, backward) > insn_idx && backward > insn_idx) {
+		forward = insn_idx;
+		meta = nfp_prog_first_meta(nfp_prog);
+	}
+
+	if (forward < backward)
+		for (i = 0; i < forward; i++)
+			meta = nfp_meta_next(meta);
+	else
+		for (i = 0; i < backward; i++)
+			meta = nfp_meta_prev(meta);
+
+	return meta;
+}
+
+static int
+nfp_bpf_check_exit(struct nfp_prog *nfp_prog,
+		   const struct bpf_verifier_env *env)
+{
+	const struct bpf_reg_state *reg0 = &env->cur_state.regs[0];
+
+	if (reg0->type != CONST_IMM) {
+		pr_info("unsupported exit state: %d, imm: %llx\n",
+			reg0->type, reg0->imm);
+		return -EINVAL;
+	}
+
+	if (reg0->imm != 0 && (reg0->imm & ~0U) != ~0U) {
+		pr_info("unsupported exit state: %d, imm: %llx\n",
+			reg0->type, reg0->imm);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int
+nfp_bpf_check_ctx_ptr(struct nfp_prog *nfp_prog,
+		      const struct bpf_verifier_env *env, u8 reg)
+{
+	if (env->cur_state.regs[reg].type != PTR_TO_CTX)
+		return -EINVAL;
+
+	return 0;
+}
+
+static int
+nfp_verify_insn(struct bpf_verifier_env *env, int insn_idx, int prev_insn_idx)
+{
+	struct nfp_bpf_analyzer_priv *priv = env->analyzer_priv;
+	struct nfp_insn_meta *meta = priv->meta;
+
+	meta = nfp_bpf_goto_meta(priv->prog, meta, insn_idx, env->prog->len);
+	priv->meta = meta;
+
+	if (meta->insn.src_reg == BPF_REG_10 ||
+	    meta->insn.dst_reg == BPF_REG_10) {
+		pr_err("stack not yet supported\n");
+		return -EINVAL;
+	}
+	if (meta->insn.src_reg >= MAX_BPF_REG ||
+	    meta->insn.dst_reg >= MAX_BPF_REG) {
+		pr_err("program uses extended registers - jit hardening?\n");
+		return -EINVAL;
+	}
+
+	if (meta->insn.code == (BPF_JMP | BPF_EXIT))
+		return nfp_bpf_check_exit(priv->prog, env);
+
+	if ((meta->insn.code & ~BPF_SIZE_MASK) == (BPF_LDX | BPF_MEM))
+		return nfp_bpf_check_ctx_ptr(priv->prog, env,
+					     meta->insn.src_reg);
+	if ((meta->insn.code & ~BPF_SIZE_MASK) == (BPF_STX | BPF_MEM))
+		return nfp_bpf_check_ctx_ptr(priv->prog, env,
+					     meta->insn.dst_reg);
+
+	return 0;
+}
+
+static const struct bpf_ext_analyzer_ops nfp_bpf_analyzer_ops = {
+	.insn_hook = nfp_verify_insn,
+};
+
+int nfp_prog_verify(struct nfp_prog *nfp_prog, struct bpf_prog *prog)
+{
+	struct nfp_bpf_analyzer_priv *priv;
+	int ret;
+
+	priv = kzalloc(sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		return -ENOMEM;
+
+	priv->prog = nfp_prog;
+	priv->meta = nfp_prog_first_meta(nfp_prog);
+
+	ret = bpf_analyzer(prog, &nfp_bpf_analyzer_ops, priv);
+
+	kfree(priv);
+
+	return ret;
+}