[PATCH bpf-next] tools, bpf_asm: warn when jumps are out of range

[PATCH bpf-next] tools, bpf_asm: warn when jumps are out of range

[Ilya Leoshkevich]

When compiling larger programs with bpf_asm, it's possible to
accidentally exceed jt/jf range, in which case it won't complain, but
rather silently emit a truncated offset, leading to a "happy debugging"
situation.

Add a warning to help detecting such issues. It could be made an error
instead, but this might break compilation of existing code (which might
be working by accident).

Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
---
 tools/bpf/bpf_exp.y | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/tools/bpf/bpf_exp.y b/tools/bpf/bpf_exp.y
index 56ba1de50784..8d48e896be50 100644
--- a/tools/bpf/bpf_exp.y
+++ b/tools/bpf/bpf_exp.y
@@ -545,6 +545,16 @@ static void bpf_reduce_k_jumps(void)
 	}
 }
 
+static uint8_t bpf_encode_jt_jf_offset(int off, int i)
+{
+	int delta = off - i - 1;
+
+	if (delta < 0 || delta > 255)
+		fprintf(stderr, "warning: insn #%d jumps to insn #%d, "
+				"which is out of range\n", i, off);
+	return (uint8_t) delta;
+}
+
 static void bpf_reduce_jt_jumps(void)
 {
 	int i;
@@ -552,7 +562,7 @@ static void bpf_reduce_jt_jumps(void)
 	for (i = 0; i < curr_instr; i++) {
 		if (labels_jt[i]) {
 			int off = bpf_find_insns_offset(labels_jt[i]);
-			out[i].jt = (uint8_t) (off - i -1);
+			out[i].jt = bpf_encode_jt_jf_offset(off, i);
 		}
 	}
 }
@@ -564,7 +574,7 @@ static void bpf_reduce_jf_jumps(void)
 	for (i = 0; i < curr_instr; i++) {
 		if (labels_jf[i]) {
 			int off = bpf_find_insns_offset(labels_jf[i]);
-			out[i].jf = (uint8_t) (off - i - 1);
+			out[i].jf = bpf_encode_jt_jf_offset(off, i);
 		}
 	}
 }
-- 
2.23.0

Re: [PATCH bpf-next] tools, bpf_asm: warn when jumps are out of range

[Daniel Borkmann]

On 11/7/19 11:03 AM, Ilya Leoshkevich wrote:
> When compiling larger programs with bpf_asm, it's possible to
> accidentally exceed jt/jf range, in which case it won't complain, but
> rather silently emit a truncated offset, leading to a "happy debugging"
> situation.
> 
> Add a warning to help detecting such issues. It could be made an error
> instead, but this might break compilation of existing code (which might
> be working by accident).
> 
> Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>

LGTM, applied, thanks!