Re: [PATCH v4 bpf-next 2/3] bpf: Support "%c" in bpf_bprintf_prepare().

From:   Andrii Nakryiko <redacted>
Date:   Wed, 11 Aug 2021 14:15:50 -0700

On Tue, Aug 10, 2021 at 2:29 AM Kuniyuki Iwashima [off-list ref] wrote:

quoted

/proc/net/unix uses "%c" to print a single-byte character to escape '\0' in
the name of the abstract UNIX domain socket.  The following selftest uses
it, so this patch adds support for "%c".  Note that it does not support
wide character ("%lc" and "%llc") for simplicity.

Signed-off-by: Kuniyuki Iwashima <redacted>
---
 kernel/bpf/helpers.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
index 15746f779fe1..6d3aaf94e9ac 100644
--- a/kernel/bpf/helpers.c
+++ b/kernel/bpf/helpers.c

@@ -907,6 +907,20 @@ int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args,
                        tmp_buf += err;
                        num_spec++;

+                       continue;
+               } else if (fmt[i] == 'c') {

you are adding new features to printk-like helpers, please add
corresponding tests as well. I'm particularly curious how something
like "% 9c" (which is now allowed, along with a few other unusual
combinations) will work.

I see. I'll add a test.
I'm now thinking of test like:
  1. pin the bpf prog that outputs "% 9c" and other format strings.
  2. read and validate it

Is there any related test ?
and is there other complicated fomat strings to test ?

Also, "% 9c" worked as is :)

---8<---
$ sudo ./tools/bpftool/bpftool iter pin ./bpf_iter_unix.o /sys/fs/bpf/unix
$ sudo cat /sys/fs/bpf/unix | head -n 1
        a
$ git diff

diff --git a/tools/testing/selftests/bpf/progs/bpf_iter_unix.c b/tools/testing/selftests/bpf/progs/bpf_iter_unix.c
index ad397e2962cf..8a7d5aa4c054 100644
--- a/tools/testing/selftests/bpf/progs/bpf_iter_unix.c
+++ b/tools/testing/selftests/bpf/progs/bpf_iter_unix.c

@@ -34,8 +34,10 @@ int dump_unix(struct bpf_iter__unix *ctx)
 
        seq = ctx->meta->seq;
        seq_num = ctx->meta->seq_num;
-       if (seq_num == 0)
+       if (seq_num == 0) {
+               BPF_SEQ_PRINTF(seq, "% 9c\n", 'a');
                BPF_SEQ_PRINTF(seq, "Num               RefCount Protocol Flags    Type St Inode    Path\n");
+       }
 
        BPF_SEQ_PRINTF(seq, "%pK: %08X %08X %08X %04X %02X %8lu",
                       unix_sk,
---8<---

quoted

+                       if (!tmp_buf)
+                               goto nocopy_fmt;
+
+                       if (tmp_buf_end == tmp_buf) {
+                               err = -ENOSPC;
+                               goto out;
+                       }
+
+                       *tmp_buf = raw_args[num_spec];
+                       tmp_buf++;
+                       num_spec++;
+
                        continue;
                }

--
2.30.2