From: Ilya Leoshkevich <iii@linux.ibm.com>
Date: Thu, 10 Sep 2020 01:21:41 +0200
Subject: s390/bpf: Fix multiple tail calls
Git-commit: d72714c1da138e6755d3bd14662dc5b7f17fae7f
Patch-mainline: v5.10-rc1
References: git-fixes

In order to branch around tail calls (due to out-of-bounds index,
exceeding tail call count or missing tail call target), JIT uses
label[0] field, which contains the address of the instruction following
the tail call. When there are multiple tail calls, label[0] value comes
from handling of a previous tail call, which is incorrect.

Fix by getting rid of label array and resolving the label address
locally: for all 3 branches that jump to it, emit 0 offsets at the
beginning, and then backpatch them with the correct value.

Also, do not use the long jump infrastructure: the tail call sequence
is known to be short, so make all 3 jumps short.

Fixes: 6651ee070b31 ("s390/bpf: implement bpf_tail_call() helper")
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20200909232141.3099367-1-iii@linux.ibm.com
[ ptesarik: Some changes from 4e9b4a6883dd97aff53ae3b08eb900716a5469dc
  had to be added to make the code work. ]
Signed-off-by: Petr Tesarik <ptesarik@suse.com>
---
 arch/s390/net/bpf_jit_comp.c |   53 +++++++++++++++++++++++++++++--------------
 1 file changed, 36 insertions(+), 17 deletions(-)

--- a/arch/s390/net/bpf_jit_comp.c
+++ b/arch/s390/net/bpf_jit_comp.c
@@ -46,7 +46,6 @@ struct bpf_jit {
 	int r1_thunk_ip;	/* Address of expoline thunk for 'br %r1' */
 	int r14_thunk_ip;	/* Address of expoline thunk for 'br %r14' */
 	int tail_call_start;	/* Tail call start offset */
-	int labels[1];		/* Labels for local jumps */
 };
 
 #define BPF_SIZE_MAX	0xffff	/* Max size for program (16 bit branches) */
@@ -191,6 +190,12 @@ static inline void reg_set_seen(struct b
 	_EMIT4(op | __pcrel);					\
 })
 
+#define EMIT4_PCREL_RIC(op, mask, target)			\
+({								\
+	int __rel = ((target) - jit->prg) / 2;			\
+	_EMIT4((op) | (mask) << 20 | (__rel & 0xffff));		\
+})
+
 #define _EMIT6(op1, op2)					\
 ({								\
 	if (jit->prg_buf) {					\
@@ -223,18 +228,18 @@ static inline void reg_set_seen(struct b
 	REG_SET_SEEN(b3);					\
 })
 
-#define EMIT6_PCREL_LABEL(op1, op2, b1, b2, label, mask)	\
+#define EMIT6_PCREL_RIEB(op1, op2, b1, b2, mask, target)	\
 ({								\
-	int rel = (jit->labels[label] - jit->prg) >> 1;		\
+	unsigned int rel = (int)((target) - jit->prg) / 2;	\
 	_EMIT6(op1 | reg(b1, b2) << 16 | (rel & 0xffff),	\
 	       op2 | mask << 12);				\
 	REG_SET_SEEN(b1);					\
 	REG_SET_SEEN(b2);					\
 })
 
-#define EMIT6_PCREL_IMM_LABEL(op1, op2, b1, imm, label, mask)	\
+#define EMIT6_PCREL_RIEC(op1, op2, b1, imm, mask, target)	\
 ({								\
-	int rel = (jit->labels[label] - jit->prg) >> 1;		\
+	unsigned int rel = (int)((target) - jit->prg) / 2;	\
 	_EMIT6(op1 | (reg_high(b1) | mask) << 16 |		\
 		(rel & 0xffff), op2 | (imm & 0xff) << 8);	\
 	REG_SET_SEEN(b1);					\
@@ -1033,7 +1038,9 @@ static noinline int bpf_jit_insn(struct
 		EMIT4(0xb9040000, BPF_REG_0, REG_2);
 		break;
 	}
-	case BPF_JMP | BPF_TAIL_CALL:
+	case BPF_JMP | BPF_TAIL_CALL: {
+		int patch_1_clrj, patch_2_clij, patch_3_brc;
+
 		/*
 		 * Implicit input:
 		 *  B1: pointer to ctx
@@ -1050,9 +1057,11 @@ static noinline int bpf_jit_insn(struct
 		/* llgf %w1,map.max_entries(%b2) */
 		EMIT6_DISP_LH(0xe3000000, 0x0016, REG_W1, REG_0, BPF_REG_2,
 			      offsetof(struct bpf_array, map.max_entries));
-		/* clrj %b3,%w1,0xa,label0: if (u32)%b3 >= (u32)%w1 goto out */
-		EMIT6_PCREL_LABEL(0xec000000, 0x0077, BPF_REG_3,
-				  REG_W1, 0, 0xa);
+		/* if ((u32)%b3 >= (u32)%w1) goto out; */
+		/* clrj %b3,%w1,0xa,out */
+		patch_1_clrj = jit->prg;
+		EMIT6_PCREL_RIEB(0xec000000, 0x0077, BPF_REG_3, REG_W1, 0xa,
+				 jit->prg);
 
 		/*
 		 * if (tail_call_cnt++ > MAX_TAIL_CALL_CNT)
@@ -1067,9 +1076,10 @@ static noinline int bpf_jit_insn(struct
 		EMIT4_IMM(0xa7080000, REG_W0, 1);
 		/* laal %w1,%w0,off(%r15) */
 		EMIT6_DISP_LH(0xeb000000, 0x00fa, REG_W1, REG_W0, REG_15, off);
-		/* clij %w1,MAX_TAIL_CALL_CNT,0x2,label0 */
-		EMIT6_PCREL_IMM_LABEL(0xec000000, 0x007f, REG_W1,
-				      MAX_TAIL_CALL_CNT, 0, 0x2);
+		/* clij %w1,MAX_TAIL_CALL_CNT,0x2,out */
+		patch_2_clij = jit->prg;
+		EMIT6_PCREL_RIEC(0xec000000, 0x007f, REG_W1, MAX_TAIL_CALL_CNT,
+				 2, jit->prg);
 
 		/*
 		 * prog = array->ptrs[index];
@@ -1081,11 +1091,12 @@ static noinline int bpf_jit_insn(struct
 		EMIT4(0xb9160000, REG_1, BPF_REG_3);
 		/* sllg %r1,%r1,3: %r1 *= 8 */
 		EMIT6_DISP_LH(0xeb000000, 0x000d, REG_1, REG_1, REG_0, 3);
-		/* lg %r1,prog(%b2,%r1) */
-		EMIT6_DISP_LH(0xe3000000, 0x0004, REG_1, BPF_REG_2,
+		/* ltg %r1,prog(%b2,%r1) */
+		EMIT6_DISP_LH(0xe3000000, 0x0002, REG_1, BPF_REG_2,
 			      REG_1, offsetof(struct bpf_array, ptrs));
-		/* clgij %r1,0,0x8,label0 */
-		EMIT6_PCREL_IMM_LABEL(0xec000000, 0x007d, REG_1, 0, 0, 0x8);
+		/* brc 0x8,out */
+		patch_3_brc = jit->prg;
+		EMIT4_PCREL_RIC(0xa7040000, 8, jit->prg);
 
 		/*
 		 * Restore registers before calling function
@@ -1102,8 +1113,16 @@ static noinline int bpf_jit_insn(struct
 		/* bc 0xf,tail_call_start(%r1) */
 		_EMIT4(0x47f01000 + jit->tail_call_start);
 		/* out: */
-		jit->labels[0] = jit->prg;
+		if (jit->prg_buf) {
+			*(u16 *)(jit->prg_buf + patch_1_clrj + 2) =
+				(jit->prg - patch_1_clrj) >> 1;
+			*(u16 *)(jit->prg_buf + patch_2_clij + 2) =
+				(jit->prg - patch_2_clij) >> 1;
+			*(u16 *)(jit->prg_buf + patch_3_brc + 2) =
+				(jit->prg - patch_3_brc) >> 1;
+		}
 		break;
+	}
 	case BPF_JMP | BPF_EXIT: /* return b0 */
 		last = (i == fp->len - 1) ? 1 : 0;
 		if (last && !(jit->seen & SEEN_RET0))