linux-kernel - [RFC][PATCH v2 7/7] selftests/bpf: Change return value in test_libbpf_get_fd_by_id

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed,  7 Dec 2022 18:24:34 +0100
From:   Roberto Sassu <roberto.sassu@...weicloud.com>
To:     ast@...nel.org, daniel@...earbox.net, andrii@...nel.org,
        martin.lau@...ux.dev, song@...nel.org, yhs@...com,
        john.fastabend@...il.com, kpsingh@...nel.org, sdf@...gle.com,
        haoluo@...gle.com, jolsa@...nel.org, revest@...omium.org,
        jackmanb@...omium.org, mykolal@...com, paul@...l-moore.com,
        jmorris@...ei.org, serge@...lyn.com, shuah@...nel.org
Cc:     bpf@...r.kernel.org, linux-security-module@...r.kernel.org,
        linux-kselftest@...r.kernel.org, linux-kernel@...r.kernel.org,
        Roberto Sassu <roberto.sassu@...wei.com>
Subject: [RFC][PATCH v2 7/7] selftests/bpf: Change return value in test_libbpf_get_fd_by_id_opts.c

From: Roberto Sassu <roberto.sassu@...wei.com>

In the no_alu32 version, the eBPF assembly looks like:

       0:	b7 00 00 00 00 00 00 00	r0 = 0
       1:	79 12 00 00 00 00 00 00	r2 = *(u64 *)(r1 + 0)
       2:	18 03 00 00 00 00 00 00 00 00 00 00 00 00 00 00	r3 = 0 ll
       4:	5d 32 04 00 00 00 00 00	if r2 != r3 goto +4 <LBB0_2>
       5:	79 10 08 00 00 00 00 00	r0 = *(u64 *)(r1 + 8)
       6:	67 00 00 00 3e 00 00 00	r0 <<= 62
       7:	c7 00 00 00 3f 00 00 00	r0 s>>= 63
       8:	57 00 00 00 f3 ff ff ff	r0 &= -13

Unfortunately, ANDing of negative numbers is not yet supported in the
verifier. As a consequence, current bounds are lost in the AND operation,
resulting in estimating a positive return value, even if there isn't.

For now, return -EPERM instead of -EACCES, to remove the AND operation and
let the verifier know that the return value is legitimate (negative or
zero).

Signed-off-by: Roberto Sassu <roberto.sassu@...wei.com>
---
 .../selftests/bpf/progs/test_libbpf_get_fd_by_id_opts.c    | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/bpf/progs/test_libbpf_get_fd_by_id_opts.c b/tools/testing/selftests/bpf/progs/test_libbpf_get_fd_by_id_opts.c
index f5ac5f3e8919..a143dbbd5573 100644
--- a/tools/testing/selftests/bpf/progs/test_libbpf_get_fd_by_id_opts.c
+++ b/tools/testing/selftests/bpf/progs/test_libbpf_get_fd_by_id_opts.c
@@ -29,8 +29,13 @@ int BPF_PROG(check_access, struct bpf_map *map, fmode_t fmode)
 	if (map != (struct bpf_map *)&data_input)
 		return 0;
 
+	/*
+	 * Prefer -EPERM to -EACCES to avoid ANDing negative numbers in the
+	 * no_alu32 version, which results in the current register bounds to
+	 * be lost.
+	 */
 	if (fmode & FMODE_WRITE)
-		return -EACCES;
+		return -EPERM;
 
 	return 0;
 }
-- 
2.25.1