bpf, x64: Fix tailcall infinite loop

From commit ebf7d1f5 ("bpf, x64: rework pro/epilogue and tailcall
handling in JIT"), the tailcall on x64 works better than before.

From commit e411901c ("bpf: allow for tailcalls in BPF subprograms
for x64 JIT"), tailcall is able to run in BPF subprograms on x64.

From commit 5b92a28a ("bpf: Support attaching tracing BPF program
to other BPF programs"), BPF program is able to trace other BPF programs.

How about combining them all together?

1. FENTRY/FEXIT on a BPF subprogram.
2. A tailcall runs in the BPF subprogram.
3. The tailcall calls the subprogram's caller.

As a result, a tailcall infinite loop comes up. And the loop would halt
the machine.

As we know, in tail call context, the tail_call_cnt propagates by stack
and rax register between BPF subprograms. So do in trampolines.

Fixes: ebf7d1f5 ("bpf, x64: rework pro/epilogue and tailcall handling in JIT")
Fixes: e411901c ("bpf: allow for tailcalls in BPF subprograms for x64 JIT")
Reviewed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Signed-off-by: Leon Hwang <hffilwlqm@gmail.com>
Link: https://lore.kernel.org/r/20230912150442.2009-3-hffilwlqm@gmail.comSigned-off-by: Alexei Starovoitov <ast@kernel.org>

arch/x86/net/bpf_jit_comp.c

@@ -1022,6 +1022,10 @@ static void emit_shiftx(u8 **pprog, u32 dst_reg, u8 src_reg, bool is64, u8 op)
 
 #define INSN_SZ_DIFF (((addrs[i] - addrs[i - 1]) - (prog - temp)))
 
+/* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */
+#define RESTORE_TAIL_CALL_CNT(stack)				\
+	EMIT3_off32(0x48, 0x8B, 0x85, -round_up(stack, 8) - 8)
+
 static int do_jit(struct bpf_prog *bpf_prog, int *addrs, u8 *image, u8 *rw_image,
 		  int oldproglen, struct jit_context *ctx, bool jmp_padding)
 {
@@ -1627,9 +1631,7 @@ st:			if (is_imm8(insn->off))
 
 			func = (u8 *) __bpf_call_base + imm32;
 			if (tail_call_reachable) {
-				/* mov rax, qword ptr [rbp - rounded_stack_depth - 8] */
-				EMIT3_off32(0x48, 0x8B, 0x85,
-					    -round_up(bpf_prog->aux->stack_depth, 8) - 8);
+				RESTORE_TAIL_CALL_CNT(bpf_prog->aux->stack_depth);
 				if (!imm32)
 					return -EINVAL;
 				offs = 7 + x86_call_depth_emit_accounting(&prog, func);
@@ -2404,6 +2406,7 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i
 	 *                     [ ...        ]
 	 *                     [ stack_arg2 ]
 	 * RBP - arg_stack_off [ stack_arg1 ]
+	 * RSP                 [ tail_call_cnt ] BPF_TRAMP_F_TAIL_CALL_CTX
 	 */
 
 	/* room for return value of orig_call or fentry prog */
@@ -2468,6 +2471,8 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i
 	else
 		/* sub rsp, stack_size */
 		EMIT4(0x48, 0x83, 0xEC, stack_size);
+	if (flags & BPF_TRAMP_F_TAIL_CALL_CTX)
+		EMIT1(0x50);		/* push rax */
 	/* mov QWORD PTR [rbp - rbx_off], rbx */
 	emit_stx(&prog, BPF_DW, BPF_REG_FP, BPF_REG_6, -rbx_off);
 
@@ -2520,9 +2525,15 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i
 		restore_regs(m, &prog, regs_off);
 		save_args(m, &prog, arg_stack_off, true);
 
+		if (flags & BPF_TRAMP_F_TAIL_CALL_CTX)
+			/* Before calling the original function, restore the
+			 * tail_call_cnt from stack to rax.
+			 */
+			RESTORE_TAIL_CALL_CNT(stack_size);
+
 		if (flags & BPF_TRAMP_F_ORIG_STACK) {
-			emit_ldx(&prog, BPF_DW, BPF_REG_0, BPF_REG_FP, 8);
-			EMIT2(0xff, 0xd0); /* call *rax */
+			emit_ldx(&prog, BPF_DW, BPF_REG_6, BPF_REG_FP, 8);
+			EMIT2(0xff, 0xd3); /* call *rbx */
 		} else {
 			/* call original function */
 			if (emit_rsb_call(&prog, orig_call, prog)) {
@@ -2573,7 +2584,12 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i
 			ret = -EINVAL;
 			goto cleanup;
 		}
-	}
+	} else if (flags & BPF_TRAMP_F_TAIL_CALL_CTX)
+		/* Before running the original function, restore the
+		 * tail_call_cnt from stack to rax.
+		 */
+		RESTORE_TAIL_CALL_CNT(stack_size);
+
 	/* restore return value of orig_call or fentry prog back into RAX */
 	if (save_ret)
 		emit_ldx(&prog, BPF_DW, BPF_REG_0, BPF_REG_FP, -8);

include/linux/bpf.h

@@ -1035,6 +1035,11 @@ struct btf_func_model {
  */
 #define BPF_TRAMP_F_SHARE_IPMODIFY	BIT(6)
 
+/* Indicate that current trampoline is in a tail call context. Then, it has to
+ * cache and restore tail_call_cnt to avoid infinite tail call loop.
+ */
+#define BPF_TRAMP_F_TAIL_CALL_CTX	BIT(7)
+
 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
  * bytes on x86.
  */

kernel/bpf/trampoline.c

@@ -415,8 +415,8 @@ static int bpf_trampoline_update(struct bpf_trampoline *tr, bool lock_direct_mut
 		goto out;
 	}
 
-	/* clear all bits except SHARE_IPMODIFY */
-	tr->flags &= BPF_TRAMP_F_SHARE_IPMODIFY;
+	/* clear all bits except SHARE_IPMODIFY and TAIL_CALL_CTX */
+	tr->flags &= (BPF_TRAMP_F_SHARE_IPMODIFY | BPF_TRAMP_F_TAIL_CALL_CTX);
 
 	if (tlinks[BPF_TRAMP_FEXIT].nr_links ||
 	    tlinks[BPF_TRAMP_MODIFY_RETURN].nr_links) {

kernel/bpf/verifier.c

@@ -19774,6 +19774,9 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)
 	if (!tr)
 		return -ENOMEM;
 
+	if (tgt_prog && tgt_prog->aux->tail_call_reachable)
+		tr->flags = BPF_TRAMP_F_TAIL_CALL_CTX;
+
 	prog->aux->dst_trampoline = tr;
 	return 0;
 }