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));