mips, bpf: Add JIT workarounds for CPU errata

This patch adds workarounds for the following CPU errata to the MIPS
eBPF JIT, if enabled in the kernel configuration.

  - R10000 ll/sc weak ordering
  - Loongson-3 ll/sc weak ordering
  - Loongson-2F jump hang

The Loongson-2F nop errata is implemented in uasm, which the JIT uses,
so no additional mitigations are needed for that.
Signed-off-by: Johan Almbladh <johan.almbladh@anyfinetworks.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Link: https://lore.kernel.org/bpf/20211005165408.2305108-6-johan.almbladh@anyfinetworks.com

arch/mips/net/bpf_jit_comp.c

@@ -404,6 +404,7 @@ void emit_alu_r(struct jit_context *ctx, u8 dst, u8 src, u8 op)
 /* Atomic read-modify-write (32-bit) */
 void emit_atomic_r(struct jit_context *ctx, u8 dst, u8 src, s16 off, u8 code)
 {
+	LLSC_sync(ctx);
 	emit(ctx, ll, MIPS_R_T9, off, dst);
 	switch (code) {
 	case BPF_ADD:
@@ -427,7 +428,7 @@ void emit_atomic_r(struct jit_context *ctx, u8 dst, u8 src, s16 off, u8 code)
 		break;
 	}
 	emit(ctx, sc, MIPS_R_T8, off, dst);
-	emit(ctx, beqz, MIPS_R_T8, -16);
+	emit(ctx, LLSC_beqz, MIPS_R_T8, -16 - LLSC_offset);
 	emit(ctx, nop); /* Delay slot */
 
 	if (code & BPF_FETCH) {
@@ -439,11 +440,12 @@ void emit_atomic_r(struct jit_context *ctx, u8 dst, u8 src, s16 off, u8 code)
 /* Atomic compare-and-exchange (32-bit) */
 void emit_cmpxchg_r(struct jit_context *ctx, u8 dst, u8 src, u8 res, s16 off)
 {
+	LLSC_sync(ctx);
 	emit(ctx, ll, MIPS_R_T9, off, dst);
 	emit(ctx, bne, MIPS_R_T9, res, 12);
 	emit(ctx, move, MIPS_R_T8, src);     /* Delay slot */
 	emit(ctx, sc, MIPS_R_T8, off, dst);
-	emit(ctx, beqz, MIPS_R_T8, -20);
+	emit(ctx, LLSC_beqz, MIPS_R_T8, -20 - LLSC_offset);
 	emit(ctx, move, res, MIPS_R_T9);     /* Delay slot */
 	clobber_reg(ctx, res);
 }

arch/mips/net/bpf_jit_comp.h

@@ -87,7 +87,7 @@ struct jit_context {
 };
 
 /* Emit the instruction if the JIT memory space has been allocated */
-#define emit(ctx, func, ...)					\
+#define __emit(ctx, func, ...)					\
 do {								\
 	if ((ctx)->target != NULL) {				\
 		u32 *p = &(ctx)->target[ctx->jit_index];	\
@@ -95,6 +95,30 @@ do {								\
 	}							\
 	(ctx)->jit_index++;					\
 } while (0)
+#define emit(...) __emit(__VA_ARGS__)
+
+/* Workaround for R10000 ll/sc errata */
+#ifdef CONFIG_WAR_R10000
+#define LLSC_beqz	beqzl
+#else
+#define LLSC_beqz	beqz
+#endif
+
+/* Workaround for Loongson-3 ll/sc errata */
+#ifdef CONFIG_CPU_LOONGSON3_WORKAROUNDS
+#define LLSC_sync(ctx)	emit(ctx, sync, 0)
+#define LLSC_offset	4
+#else
+#define LLSC_sync(ctx)
+#define LLSC_offset	0
+#endif
+
+/* Workaround for Loongson-2F jump errata */
+#ifdef CONFIG_CPU_JUMP_WORKAROUNDS
+#define JALR_MASK	0xffffffffcfffffffULL
+#else
+#define JALR_MASK	(~0ULL)
+#endif
 
 /*
  * Mark a BPF register as accessed, it needs to be

arch/mips/net/bpf_jit_comp64.c

@@ -375,6 +375,7 @@ static void emit_atomic_r64(struct jit_context *ctx,
 	u8 t1 = MIPS_R_T6;
 	u8 t2 = MIPS_R_T7;
 
+	LLSC_sync(ctx);
 	emit(ctx, lld, t1, off, dst);
 	switch (code) {
 	case BPF_ADD:
@@ -398,7 +399,7 @@ static void emit_atomic_r64(struct jit_context *ctx,
 		break;
 	}
 	emit(ctx, scd, t2, off, dst);
-	emit(ctx, beqz, t2, -16);
+	emit(ctx, LLSC_beqz, t2, -16 - LLSC_offset);
 	emit(ctx, nop); /* Delay slot */
 
 	if (code & BPF_FETCH) {
@@ -414,12 +415,13 @@ static void emit_cmpxchg_r64(struct jit_context *ctx, u8 dst, u8 src, s16 off)
 	u8 t1 = MIPS_R_T6;
 	u8 t2 = MIPS_R_T7;
 
+	LLSC_sync(ctx);
 	emit(ctx, lld, t1, off, dst);
 	emit(ctx, bne, t1, r0, 12);
 	emit(ctx, move, t2, src);      /* Delay slot */
 	emit(ctx, scd, t2, off, dst);
-	emit(ctx, beqz, t2, -20);
-	emit(ctx, move, r0, t1);      /* Delay slot */
+	emit(ctx, LLSC_beqz, t2, -20 - LLSC_offset);
+	emit(ctx, move, r0, t1);       /* Delay slot */
 
 	clobber_reg(ctx, r0);
 }
@@ -443,7 +445,7 @@ static int emit_call(struct jit_context *ctx, const struct bpf_insn *insn)
 	push_regs(ctx, ctx->clobbered & JIT_CALLER_REGS, 0, 0);
 
 	/* Emit function call */
-	emit_mov_i64(ctx, tmp, addr);
+	emit_mov_i64(ctx, tmp, addr & JALR_MASK);
 	emit(ctx, jalr, MIPS_R_RA, tmp);
 	emit(ctx, nop); /* Delay slot */