linux-arm-kernel.lists.infradead.org archive mirror
 help / color / mirror / Atom feed
From: Xu Kuohai <xukuohai@huaweicloud.com>
To: bpf@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
	linux-kernel@vger.kernel.org
Cc: Daniel Borkmann <daniel@iogearbox.net>,
	Jean-Philippe Brucker <jean-philippe@linaro.org>,
	Alexei Starovoitov <ast@kernel.org>,
	Andrii Nakryiko <andrii@kernel.org>,
	Martin KaFai Lau <martin.lau@linux.dev>,
	Song Liu <song@kernel.org>, Yonghong Song <yhs@fb.com>,
	John Fastabend <john.fastabend@gmail.com>,
	KP Singh <kpsingh@kernel.org>,
	Stanislav Fomichev <sdf@google.com>, Hao Luo <haoluo@google.com>,
	Jiri Olsa <jolsa@kernel.org>, Zi Shen Lim <zlim.lnx@gmail.com>,
	Catalin Marinas <catalin.marinas@arm.com>,
	Will Deacon <will@kernel.org>
Subject: [PATCH bpf-next 1/2] bpf, arm64: Jit BPF_CALL to direct call when possible
Date: Fri,  2 Sep 2022 11:20:41 -0400	[thread overview]
Message-ID: <20220902152043.721806-2-xukuohai@huaweicloud.com> (raw)
In-Reply-To: <20220902152043.721806-1-xukuohai@huaweicloud.com>

From: Xu Kuohai <xukuohai@huawei.com>

Currently BPF_CALL is always jited to indirect call, but when target is
in the range of direct call, a BPF_CALL can be jited to direct call.

For example, the following BPF_CALL

    call __htab_map_lookup_elem

is always jited to an indirect call:

    mov     x10, #0xffffffffffff18f4
    movk    x10, #0x821, lsl #16
    movk    x10, #0x8000, lsl #32
    blr     x10

When the target is in the range of a direct call, it can be jited to:

    bl      0xfffffffffd33bc98

This patch does such jit when possible.

1. Before allocating jit image memory, jit all BPF_CALL to indirect call,
   so we could get the maximum image size at the end.

2. Allocate jit image memory with the maximum image size.

3. Since we have now allocated jit image memory, every jited instruction
   address is determined, so the distance to call target is determined,
   so whether a BPF_CALL can be jited to direct call is determined.

4. Since step 3 may adjust the position of jited instructions, update the
   offset of jump instructions whose target is within the jit image.

Tested with test_bpf.ko and some arm64 working selftests, nothing failed.

Signed-off-by: Xu Kuohai <xukuohai@huawei.com>
---
 arch/arm64/net/bpf_jit_comp.c | 71 ++++++++++++++++++++++++++++-------
 1 file changed, 58 insertions(+), 13 deletions(-)

diff --git a/arch/arm64/net/bpf_jit_comp.c b/arch/arm64/net/bpf_jit_comp.c
index 389623ae5a91..69eb29f397e5 100644
--- a/arch/arm64/net/bpf_jit_comp.c
+++ b/arch/arm64/net/bpf_jit_comp.c
@@ -72,6 +72,7 @@ static const int bpf2a64[] = {
 struct jit_ctx {
 	const struct bpf_prog *prog;
 	int idx;
+	bool write;
 	int epilogue_offset;
 	int *offset;
 	int exentry_idx;
@@ -91,7 +92,7 @@ struct bpf_plt {
 
 static inline void emit(const u32 insn, struct jit_ctx *ctx)
 {
-	if (ctx->image != NULL)
+	if (ctx->image != NULL && ctx->write)
 		ctx->image[ctx->idx] = cpu_to_le32(insn);
 
 	ctx->idx++;
@@ -178,10 +179,29 @@ static inline void emit_addr_mov_i64(const int reg, const u64 val,
 
 static inline void emit_call(u64 target, struct jit_ctx *ctx)
 {
-	u8 tmp = bpf2a64[TMP_REG_1];
+	u8 tmp;
+	long offset;
+	unsigned long pc;
+	u32 insn = AARCH64_BREAK_FAULT;
+
+	/* if ctx->image == NULL or target == 0, the jump distance is unknown,
+	 * emit indirect call.
+	 */
+	if (ctx->image && target) {
+		pc = (unsigned long)&ctx->image[ctx->idx];
+		offset = (long)target - (long)pc;
+		if (offset >= -SZ_128M && offset < SZ_128M)
+			insn = aarch64_insn_gen_branch_imm(pc, target,
+					AARCH64_INSN_BRANCH_LINK);
+	}
 
-	emit_addr_mov_i64(tmp, target, ctx);
-	emit(A64_BLR(tmp), ctx);
+	if (insn == AARCH64_BREAK_FAULT) {
+		tmp = bpf2a64[TMP_REG_1];
+		emit_addr_mov_i64(tmp, target, ctx);
+		emit(A64_BLR(tmp), ctx);
+	} else {
+		emit(insn, ctx);
+	}
 }
 
 static inline int bpf2a64_offset(int bpf_insn, int off,
@@ -1392,13 +1412,11 @@ static int build_body(struct jit_ctx *ctx, bool extra_pass)
 		const struct bpf_insn *insn = &prog->insnsi[i];
 		int ret;
 
-		if (ctx->image == NULL)
-			ctx->offset[i] = ctx->idx;
+		ctx->offset[i] = ctx->idx;
 		ret = build_insn(insn, ctx, extra_pass);
 		if (ret > 0) {
 			i++;
-			if (ctx->image == NULL)
-				ctx->offset[i] = ctx->idx;
+			ctx->offset[i] = ctx->idx;
 			continue;
 		}
 		if (ret)
@@ -1409,8 +1427,7 @@ static int build_body(struct jit_ctx *ctx, bool extra_pass)
 	 * the last element with the offset after the last
 	 * instruction (end of program)
 	 */
-	if (ctx->image == NULL)
-		ctx->offset[i] = ctx->idx;
+	ctx->offset[i] = ctx->idx;
 
 	return 0;
 }
@@ -1461,6 +1478,8 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog)
 	bool extra_pass = false;
 	struct jit_ctx ctx;
 	u8 *image_ptr;
+	int body_offset;
+	int exentry_idx;
 
 	if (!prog->jit_requested)
 		return orig_prog;
@@ -1515,6 +1534,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog)
 		goto out_off;
 	}
 
+	/* Get the max image size */
 	if (build_body(&ctx, extra_pass)) {
 		prog = orig_prog;
 		goto out_off;
@@ -1528,7 +1548,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog)
 	extable_size = prog->aux->num_exentries *
 		sizeof(struct exception_table_entry);
 
-	/* Now we know the actual image size. */
+	/* Now we know the max image size. */
 	prog_size = sizeof(u32) * ctx.idx;
 	/* also allocate space for plt target */
 	extable_offset = round_up(prog_size + PLT_TARGET_SIZE, extable_align);
@@ -1548,15 +1568,37 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog)
 skip_init_ctx:
 	ctx.idx = 0;
 	ctx.exentry_idx = 0;
+	ctx.write = true;
 
 	build_prologue(&ctx, was_classic);
 
+	/* Record exentry_idx and ctx.idx before first build_body */
+	exentry_idx = ctx.exentry_idx;
+	body_offset = ctx.idx;
+	/* Don't write instruction to memory for now */
+	ctx.write = false;
+
+	/* Determine call distance and instruction position */
 	if (build_body(&ctx, extra_pass)) {
 		bpf_jit_binary_free(header);
 		prog = orig_prog;
 		goto out_off;
 	}
 
+	ctx.epilogue_offset = ctx.idx;
+
+	ctx.exentry_idx = exentry_idx;
+	ctx.idx = body_offset;
+	ctx.write = true;
+
+	/* Determine jump offset and write result to memory */
+	if (build_body(&ctx, extra_pass) ||
+		WARN_ON_ONCE(ctx.idx != ctx.epilogue_offset)) {
+		bpf_jit_binary_free(header);
+		prog = orig_prog;
+		goto out_off;
+	}
+
 	build_epilogue(&ctx);
 	build_plt(&ctx);
 
@@ -1567,6 +1609,8 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog)
 		goto out_off;
 	}
 
+	/* Update prog size */
+	prog_size = sizeof(u32) * ctx.idx;
 	/* And we're done. */
 	if (bpf_jit_enable > 1)
 		bpf_jit_dump(prog->len, prog_size, 2, ctx.image);
@@ -1574,8 +1618,8 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog)
 	bpf_flush_icache(header, ctx.image + ctx.idx);
 
 	if (!prog->is_func || extra_pass) {
-		if (extra_pass && ctx.idx != jit_data->ctx.idx) {
-			pr_err_once("multi-func JIT bug %d != %d\n",
+		if (extra_pass && ctx.idx > jit_data->ctx.idx) {
+			pr_err_once("multi-func JIT bug %d > %d\n",
 				    ctx.idx, jit_data->ctx.idx);
 			bpf_jit_binary_free(header);
 			prog->bpf_func = NULL;
@@ -1976,6 +2020,7 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image,
 	struct jit_ctx ctx = {
 		.image = NULL,
 		.idx = 0,
+		.write = true,
 	};
 
 	/* the first 8 arguments are passed by registers */
-- 
2.30.2


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

  reply	other threads:[~2022-09-02 15:14 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-09-02 15:20 [PATCH bpf-next 0/2] Jit BPF_CALL to direct call when possible Xu Kuohai
2022-09-02 15:20 ` Xu Kuohai [this message]
2022-09-02 15:20 ` [PATCH bpf-next 2/2] bpf, arm64: Eliminate false -EFBIG error in bpf trampoline Xu Kuohai

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220902152043.721806-2-xukuohai@huaweicloud.com \
    --to=xukuohai@huaweicloud.com \
    --cc=andrii@kernel.org \
    --cc=ast@kernel.org \
    --cc=bpf@vger.kernel.org \
    --cc=catalin.marinas@arm.com \
    --cc=daniel@iogearbox.net \
    --cc=haoluo@google.com \
    --cc=jean-philippe@linaro.org \
    --cc=john.fastabend@gmail.com \
    --cc=jolsa@kernel.org \
    --cc=kpsingh@kernel.org \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=martin.lau@linux.dev \
    --cc=sdf@google.com \
    --cc=song@kernel.org \
    --cc=will@kernel.org \
    --cc=yhs@fb.com \
    --cc=zlim.lnx@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).