Merge branch 'bpf-nfp-jit-adjust-head-support'

Jakub Kicinski says:

====================
This small set adds support for bpf_xdp_adjust_head() to the offload.
Since we have access to unmodified BPF bytecode translating calls is
pretty trivial.  First part of the series adds handling of BPF
capabilities included in the FW in TLV format.  The last two patches
add adjust head support in the nfp verifier and jit, and a small
optimization in case we can guarantee the constant adjustment
will always meet adjustment constaints.
====================
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

drivers/net/ethernet/netronome/nfp/bpf/fw.h

+/*
+ * Copyright (C) 2017 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_FW_H
+#define NFP_BPF_FW_H 1
+
+#include <linux/bitops.h>
+#include <linux/types.h>
+
+enum bpf_cap_tlv_type {
+	NFP_BPF_CAP_TYPE_ADJUST_HEAD	= 2,
+};
+
+struct nfp_bpf_cap_tlv_adjust_head {
+	__le32 flags;
+	__le32 off_min;
+	__le32 off_max;
+	__le32 guaranteed_sub;
+	__le32 guaranteed_add;
+};
+
+#define NFP_BPF_ADJUST_HEAD_NO_META	BIT(0)
+
+#endif

drivers/net/ethernet/netronome/nfp/bpf/jit.c

@@ -33,6 +33,7 @@
 
 #define pr_fmt(fmt)	"NFP net bpf: " fmt
 
+#include <linux/bug.h>
 #include <linux/kernel.h>
 #include <linux/bpf.h>
 #include <linux/filter.h>
@@ -87,6 +88,18 @@ static unsigned int nfp_prog_current_offset(struct nfp_prog *nfp_prog)
 	return nfp_prog->start_off + nfp_prog->prog_len;
 }
 
+static bool
+nfp_prog_confirm_current_offset(struct nfp_prog *nfp_prog, unsigned int off)
+{
+	/* If there is a recorded error we may have dropped instructions;
+	 * that doesn't have to be due to translator bug, and the translation
+	 * will fail anyway, so just return OK.
+	 */
+	if (nfp_prog->error)
+		return true;
+	return !WARN_ON_ONCE(nfp_prog_current_offset(nfp_prog) != off);
+}
+
 static unsigned int
 nfp_prog_offset_to_index(struct nfp_prog *nfp_prog, unsigned int offset)
 {
@@ -1196,6 +1209,86 @@ static void wrp_end32(struct nfp_prog *nfp_prog, swreg reg_in, u8 gpr_out)
 		      SHF_SC_R_ROT, 16);
 }
 
+static int adjust_head(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	swreg tmp = imm_a(nfp_prog), tmp_len = imm_b(nfp_prog);
+	struct nfp_bpf_cap_adjust_head *adjust_head;
+	u32 ret_einval, end;
+
+	adjust_head = &nfp_prog->bpf->adjust_head;
+
+	/* Optimized version - 5 vs 14 cycles */
+	if (nfp_prog->adjust_head_location != UINT_MAX) {
+		if (WARN_ON_ONCE(nfp_prog->adjust_head_location != meta->n))
+			return -EINVAL;
+
+		emit_alu(nfp_prog, pptr_reg(nfp_prog),
+			 reg_a(2 * 2), ALU_OP_ADD, pptr_reg(nfp_prog));
+		emit_alu(nfp_prog, plen_reg(nfp_prog),
+			 plen_reg(nfp_prog), ALU_OP_SUB, reg_a(2 * 2));
+		emit_alu(nfp_prog, pv_len(nfp_prog),
+			 pv_len(nfp_prog), ALU_OP_SUB, reg_a(2 * 2));
+
+		wrp_immed(nfp_prog, reg_both(0), 0);
+		wrp_immed(nfp_prog, reg_both(1), 0);
+
+		/* TODO: when adjust head is guaranteed to succeed we can
+		 * also eliminate the following if (r0 == 0) branch.
+		 */
+
+		return 0;
+	}
+
+	ret_einval = nfp_prog_current_offset(nfp_prog) + 14;
+	end = ret_einval + 2;
+
+	/* We need to use a temp because offset is just a part of the pkt ptr */
+	emit_alu(nfp_prog, tmp,
+		 reg_a(2 * 2), ALU_OP_ADD_2B, pptr_reg(nfp_prog));
+
+	/* Validate result will fit within FW datapath constraints */
+	emit_alu(nfp_prog, reg_none(),
+		 tmp, ALU_OP_SUB, reg_imm(adjust_head->off_min));
+	emit_br(nfp_prog, BR_BLO, ret_einval, 0);
+	emit_alu(nfp_prog, reg_none(),
+		 reg_imm(adjust_head->off_max), ALU_OP_SUB, tmp);
+	emit_br(nfp_prog, BR_BLO, ret_einval, 0);
+
+	/* Validate the length is at least ETH_HLEN */
+	emit_alu(nfp_prog, tmp_len,
+		 plen_reg(nfp_prog), ALU_OP_SUB, reg_a(2 * 2));
+	emit_alu(nfp_prog, reg_none(),
+		 tmp_len, ALU_OP_SUB, reg_imm(ETH_HLEN));
+	emit_br(nfp_prog, BR_BMI, ret_einval, 0);
+
+	/* Load the ret code */
+	wrp_immed(nfp_prog, reg_both(0), 0);
+	wrp_immed(nfp_prog, reg_both(1), 0);
+
+	/* Modify the packet metadata */
+	emit_ld_field(nfp_prog, pptr_reg(nfp_prog), 0x3, tmp, SHF_SC_NONE, 0);
+
+	/* Skip over the -EINVAL ret code (defer 2) */
+	emit_br_def(nfp_prog, end, 2);
+
+	emit_alu(nfp_prog, plen_reg(nfp_prog),
+		 plen_reg(nfp_prog), ALU_OP_SUB, reg_a(2 * 2));
+	emit_alu(nfp_prog, pv_len(nfp_prog),
+		 pv_len(nfp_prog), ALU_OP_SUB, reg_a(2 * 2));
+
+	/* return -EINVAL target */
+	if (!nfp_prog_confirm_current_offset(nfp_prog, ret_einval))
+		return -EINVAL;
+
+	wrp_immed(nfp_prog, reg_both(0), -22);
+	wrp_immed(nfp_prog, reg_both(1), ~0);
+
+	if (!nfp_prog_confirm_current_offset(nfp_prog, end))
+		return -EINVAL;
+
+	return 0;
+}
+
 /* --- Callbacks --- */
 static int mov_reg64(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
 {
@@ -1930,6 +2023,17 @@ 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 call(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta)
+{
+	switch (meta->insn.imm) {
+	case BPF_FUNC_xdp_adjust_head:
+		return adjust_head(nfp_prog, meta);
+	default:
+		WARN_ONCE(1, "verifier allowed unsupported function\n");
+		return -EOPNOTSUPP;
+	}
+}
+
 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);
@@ -2002,6 +2106,7 @@ static const instr_cb_t instr_cb[256] = {
 	[BPF_JMP | BPF_JLE | BPF_X] =	jle_reg,
 	[BPF_JMP | BPF_JSET | BPF_X] =	jset_reg,
 	[BPF_JMP | BPF_JNE | BPF_X] =	jne_reg,
+	[BPF_JMP | BPF_CALL] =		call,
 	[BPF_JMP | BPF_EXIT] =		goto_out,
 };
 
@@ -2026,6 +2131,8 @@ static int nfp_fixup_branches(struct nfp_prog *nfp_prog)
 	list_for_each_entry(meta, &nfp_prog->insns, l) {
 		if (meta->skip)
 			continue;
+		if (meta->insn.code == (BPF_JMP | BPF_CALL))
+			continue;
 		if (BPF_CLASS(meta->insn.code) != BPF_JMP)
 			continue;
 

drivers/net/ethernet/netronome/nfp/bpf/main.c

@@ -34,10 +34,12 @@
 #include <net/pkt_cls.h>
 
 #include "../nfpcore/nfp_cpp.h"
+#include "../nfpcore/nfp_nffw.h"
 #include "../nfp_app.h"
 #include "../nfp_main.h"
 #include "../nfp_net.h"
 #include "../nfp_port.h"
+#include "fw.h"
 #include "main.h"
 
 static bool nfp_net_ebpf_capable(struct nfp_net *nn)
@@ -155,10 +157,123 @@ static bool nfp_bpf_tc_busy(struct nfp_app *app, struct nfp_net *nn)
 	return nn->dp.ctrl & NFP_NET_CFG_CTRL_BPF;
 }
 
+static int
+nfp_bpf_parse_cap_adjust_head(struct nfp_app_bpf *bpf, void __iomem *value,
+			      u32 length)
+{
+	struct nfp_bpf_cap_tlv_adjust_head __iomem *cap = value;
+	struct nfp_cpp *cpp = bpf->app->pf->cpp;
+
+	if (length < sizeof(*cap)) {
+		nfp_err(cpp, "truncated adjust_head TLV: %d\n", length);
+		return -EINVAL;
+	}
+
+	bpf->adjust_head.flags = readl(&cap->flags);
+	bpf->adjust_head.off_min = readl(&cap->off_min);
+	bpf->adjust_head.off_max = readl(&cap->off_max);
+	bpf->adjust_head.guaranteed_sub = readl(&cap->guaranteed_sub);
+	bpf->adjust_head.guaranteed_add = readl(&cap->guaranteed_add);
+
+	if (bpf->adjust_head.off_min > bpf->adjust_head.off_max) {
+		nfp_err(cpp, "invalid adjust_head TLV: min > max\n");
+		return -EINVAL;
+	}
+	if (!FIELD_FIT(UR_REG_IMM_MAX, bpf->adjust_head.off_min) ||
+	    !FIELD_FIT(UR_REG_IMM_MAX, bpf->adjust_head.off_max)) {
+		nfp_warn(cpp, "disabling adjust_head - driver expects min/max to fit in as immediates\n");
+		memset(&bpf->adjust_head, 0, sizeof(bpf->adjust_head));
+		return 0;
+	}
+
+	return 0;
+}
+
+static int nfp_bpf_parse_capabilities(struct nfp_app *app)
+{
+	struct nfp_cpp *cpp = app->pf->cpp;
+	struct nfp_cpp_area *area;
+	u8 __iomem *mem, *start;
+
+	mem = nfp_rtsym_map(app->pf->rtbl, "_abi_bpf_capabilities", "bpf.cap",
+			    8, &area);
+	if (IS_ERR(mem))
+		return PTR_ERR(mem) == -ENOENT ? 0 : PTR_ERR(mem);
+
+	start = mem;
+	while (mem - start + 8 < nfp_cpp_area_size(area)) {
+		u8 __iomem *value;
+		u32 type, length;
+
+		type = readl(mem);
+		length = readl(mem + 4);
+		value = mem + 8;
+
+		mem += 8 + length;
+		if (mem - start > nfp_cpp_area_size(area))
+			goto err_release_free;
+
+		switch (type) {
+		case NFP_BPF_CAP_TYPE_ADJUST_HEAD:
+			if (nfp_bpf_parse_cap_adjust_head(app->priv, value,
+							  length))
+				goto err_release_free;
+			break;
+		default:
+			nfp_dbg(cpp, "unknown BPF capability: %d\n", type);
+			break;
+		}
+	}
+	if (mem - start != nfp_cpp_area_size(area)) {
+		nfp_err(cpp, "BPF capabilities left after parsing, parsed:%lu total length:%lu\n",
+			mem - start, nfp_cpp_area_size(area));
+		goto err_release_free;
+	}
+
+	nfp_cpp_area_release_free(area);
+
+	return 0;
+
+err_release_free:
+	nfp_err(cpp, "invalid BPF capabilities at offset:%ld\n", mem - start);
+	nfp_cpp_area_release_free(area);
+	return -EINVAL;
+}
+
+static int nfp_bpf_init(struct nfp_app *app)
+{
+	struct nfp_app_bpf *bpf;
+	int err;
+
+	bpf = kzalloc(sizeof(*bpf), GFP_KERNEL);
+	if (!bpf)
+		return -ENOMEM;
+	bpf->app = app;
+	app->priv = bpf;
+
+	err = nfp_bpf_parse_capabilities(app);
+	if (err)
+		goto err_free_bpf;
+
+	return 0;
+
+err_free_bpf:
+	kfree(bpf);
+	return err;
+}
+
+static void nfp_bpf_clean(struct nfp_app *app)
+{
+	kfree(app->priv);
+}
+
 const struct nfp_app_type app_bpf = {
 	.id		= NFP_APP_BPF_NIC,
 	.name		= "ebpf",
 
+	.init		= nfp_bpf_init,
+	.clean		= nfp_bpf_clean,
+
 	.extra_cap	= nfp_bpf_extra_cap,
 
 	.vnic_alloc	= nfp_app_nic_vnic_alloc,

drivers/net/ethernet/netronome/nfp/bpf/main.h

@@ -78,6 +78,29 @@ enum pkt_vec {
 #define NFP_BPF_ABI_FLAGS	reg_imm(0)
 #define   NFP_BPF_ABI_FLAG_MARK	1
 
+/**
+ * struct nfp_app_bpf - bpf app priv structure
+ * @app:		backpointer to the app
+ *
+ * @adjust_head:	adjust head capability
+ * @flags:		extra flags for adjust head
+ * @off_min:		minimal packet offset within buffer required
+ * @off_max:		maximum packet offset within buffer required
+ * @guaranteed_sub:	amount of negative adjustment guaranteed possible
+ * @guaranteed_add:	amount of positive adjustment guaranteed possible
+ */
+struct nfp_app_bpf {
+	struct nfp_app *app;
+
+	struct nfp_bpf_cap_adjust_head {
+		u32 flags;
+		int off_min;
+		int off_max;
+		int guaranteed_sub;
+		int guaranteed_add;
+	} adjust_head;
+};
+
 struct nfp_prog;
 struct nfp_insn_meta;
 typedef int (*instr_cb_t)(struct nfp_prog *, struct nfp_insn_meta *);
@@ -97,6 +120,7 @@ typedef int (*instr_cb_t)(struct nfp_prog *, struct nfp_insn_meta *);
  * @ptr: pointer type for memory operations
  * @ldst_gather_len: memcpy length gathered from load/store sequence
  * @paired_st: the paired store insn at the head of the sequence
+ * @arg2: arg2 for call instructions
  * @ptr_not_const: pointer is not always constant
  * @jmp_dst: destination info for jump instructions
  * @off: index of first generated machine instruction (in nfp_prog.prog)
@@ -116,6 +140,7 @@ struct nfp_insn_meta {
 			bool ptr_not_const;
 		};
 		struct nfp_insn_meta *jmp_dst;
+		struct bpf_reg_state arg2;
 	};
 	unsigned int off;
 	unsigned short n;
@@ -160,6 +185,7 @@ static inline bool is_mbpf_store(const struct nfp_insn_meta *meta)
 
 /**
  * struct nfp_prog - nfp BPF program
+ * @bpf: backpointer to the bpf app priv structure
  * @prog: machine code
  * @prog_len: number of valid instructions in @prog array
  * @__prog_alloc_len: alloc size of @prog array
@@ -173,9 +199,12 @@ static inline bool is_mbpf_store(const struct nfp_insn_meta *meta)
  * @n_translated: number of successfully translated instructions (for errors)
  * @error: error code if something went wrong
  * @stack_depth: max stack depth from the verifier
+ * @adjust_head_location: if program has single adjust head call - the insn no.
  * @insns: list of BPF instruction wrappers (struct nfp_insn_meta)
  */
 struct nfp_prog {
+	struct nfp_app_bpf *bpf;
+
 	u64 *prog;
 	unsigned int prog_len;
 	unsigned int __prog_alloc_len;
@@ -194,6 +223,7 @@ struct nfp_prog {
 	int error;
 
 	unsigned int stack_depth;
+	unsigned int adjust_head_location;
 
 	struct list_head insns;
 };

drivers/net/ethernet/netronome/nfp/bpf/offload.c

@@ -48,6 +48,7 @@
 #include <net/tc_act/tc_mirred.h>
 
 #include "main.h"
+#include "../nfp_app.h"
 #include "../nfp_net_ctrl.h"
 #include "../nfp_net.h"
 
@@ -115,6 +116,7 @@ int nfp_bpf_verifier_prep(struct nfp_app *app, struct nfp_net *nn,
 
 	INIT_LIST_HEAD(&nfp_prog->insns);
 	nfp_prog->type = prog->type;
+	nfp_prog->bpf = app->priv;
 
 	ret = nfp_prog_prepare(nfp_prog, prog->insnsi, prog->len);
 	if (ret)

drivers/net/ethernet/netronome/nfp/bpf/verifier.c

@@ -38,6 +38,7 @@
 #include <linux/kernel.h>
 #include <linux/pkt_cls.h>
 
+#include "fw.h"
 #include "main.h"
 
 struct nfp_insn_meta *
@@ -68,6 +69,73 @@ nfp_bpf_goto_meta(struct nfp_prog *nfp_prog, struct nfp_insn_meta *meta,
 	return meta;
 }
 
+static void
+nfp_record_adjust_head(struct nfp_app_bpf *bpf, struct nfp_prog *nfp_prog,
+		       struct nfp_insn_meta *meta,
+		       const struct bpf_reg_state *reg2)
+{
+	unsigned int location =	UINT_MAX;
+	int imm;
+
+	/* Datapath usually can give us guarantees on how much adjust head
+	 * can be done without the need for any checks.  Optimize the simple
+	 * case where there is only one adjust head by a constant.
+	 */
+	if (reg2->type != SCALAR_VALUE || !tnum_is_const(reg2->var_off))
+		goto exit_set_location;
+	imm = reg2->var_off.value;
+	/* Translator will skip all checks, we need to guarantee min pkt len */
+	if (imm > ETH_ZLEN - ETH_HLEN)
+		goto exit_set_location;
+	if (imm > (int)bpf->adjust_head.guaranteed_add ||
+	    imm < -bpf->adjust_head.guaranteed_sub)
+		goto exit_set_location;
+
+	if (nfp_prog->adjust_head_location) {
+		/* Only one call per program allowed */
+		if (nfp_prog->adjust_head_location != meta->n)
+			goto exit_set_location;
+
+		if (meta->arg2.var_off.value != imm)
+			goto exit_set_location;
+	}
+
+	location = meta->n;
+exit_set_location:
+	nfp_prog->adjust_head_location = location;
+}
+
+static int
+nfp_bpf_check_call(struct nfp_prog *nfp_prog, struct bpf_verifier_env *env,
+		   struct nfp_insn_meta *meta)
+{
+	const struct bpf_reg_state *reg2 = cur_regs(env) + BPF_REG_2;
+	struct nfp_app_bpf *bpf = nfp_prog->bpf;
+	u32 func_id = meta->insn.imm;
+
+	switch (func_id) {
+	case BPF_FUNC_xdp_adjust_head:
+		if (!bpf->adjust_head.off_max) {
+			pr_warn("adjust_head not supported by FW\n");
+			return -EOPNOTSUPP;
+		}
+		if (!(bpf->adjust_head.flags & NFP_BPF_ADJUST_HEAD_NO_META)) {
+			pr_warn("adjust_head: FW requires shifting metadata, not supported by the driver\n");
+			return -EOPNOTSUPP;
+		}
+
+		nfp_record_adjust_head(bpf, nfp_prog, meta, reg2);
+		break;
+	default:
+		pr_warn("unsupported function id: %d\n", func_id);
+		return -EOPNOTSUPP;
+	}
+
+	meta->arg2 = *reg2;
+
+	return 0;
+}
+
 static int
 nfp_bpf_check_exit(struct nfp_prog *nfp_prog,
 		   struct bpf_verifier_env *env)
@@ -177,6 +245,8 @@ nfp_verify_insn(struct bpf_verifier_env *env, int insn_idx, int prev_insn_idx)
 		return -EINVAL;
 	}
 
+	if (meta->insn.code == (BPF_JMP | BPF_CALL))
+		return nfp_bpf_check_call(nfp_prog, env, meta);
 	if (meta->insn.code == (BPF_JMP | BPF_EXIT))
 		return nfp_bpf_check_exit(nfp_prog, env);
 

drivers/net/ethernet/netronome/nfp/nfp_asm.h

@@ -77,6 +77,7 @@
 enum br_mask {
 	BR_BEQ = 0x00,
 	BR_BNE = 0x01,
+	BR_BMI = 0x02,
 	BR_BHS = 0x04,
 	BR_BLO = 0x05,
 	BR_BGE = 0x08,
@@ -175,6 +176,7 @@ enum alu_op {
 	ALU_OP_NONE	= 0x00,
 	ALU_OP_ADD	= 0x01,
 	ALU_OP_NOT	= 0x04,
+	ALU_OP_ADD_2B	= 0x05,
 	ALU_OP_AND	= 0x08,
 	ALU_OP_SUB_C	= 0x0d,
 	ALU_OP_ADD_C	= 0x11,

drivers/net/ethernet/netronome/nfp/nfpcore/nfp_cpp.h

@@ -242,6 +242,7 @@ int nfp_cpp_area_read(struct nfp_cpp_area *area, unsigned long offset,
 		      void *buffer, size_t length);
 int nfp_cpp_area_write(struct nfp_cpp_area *area, unsigned long offset,
 		       const void *buffer, size_t length);
+size_t nfp_cpp_area_size(struct nfp_cpp_area *area);
 const char *nfp_cpp_area_name(struct nfp_cpp_area *cpp_area);
 void *nfp_cpp_area_priv(struct nfp_cpp_area *cpp_area);
 struct nfp_cpp *nfp_cpp_area_cpp(struct nfp_cpp_area *cpp_area);

drivers/net/ethernet/netronome/nfp/nfpcore/nfp_cppcore.c

@@ -567,6 +567,17 @@ int nfp_cpp_area_write(struct nfp_cpp_area *area,
 	return area->cpp->op->area_write(area, kernel_vaddr, offset, length);
 }
 
+/**
+ * nfp_cpp_area_size() - return size of a CPP area
+ * @cpp_area:	CPP area handle
+ *
+ * Return: Size of the area
+ */
+size_t nfp_cpp_area_size(struct nfp_cpp_area *cpp_area)
+{
+	return cpp_area->size;
+}
+
 /**
  * nfp_cpp_area_name() - return name of a CPP area
  * @cpp_area:	CPP area handle