[RFC,bpf-next] libbpf: print character arrays as strings if possible

Commit Message

Quentin Deslandes April 13, 2024, 9:39 p.m. UTC

Introduce the new print_strings flag in btf_dump_type_data_opts. If
enabled, libbpf will print character arrays as strings if they meet the
following conditions:
- Contains a nul-termination character ('\0')
- Contains only printable characters before the nul-termination character

If print_strings is set to false (default value), the existing
behavior remains unchanged.

With print_strings=false:
.str_array = (__u8[14])[
    'H',
    'e',
    'l',
    'l',
    'o',
],

With print_strings=true:
.str_array = (__u8[14])"Hello",

Signed-off-by: Quentin Deslandes <[e-mail]>
---
 tools/lib/bpf/btf.h      |  3 ++-
 tools/lib/bpf/btf_dump.c | 25 ++++++++++++++++++++++++-
 2 files changed, 26 insertions(+), 2 deletions(-)

--
2.44.0

Comments

Eduard Zingerman April 24, 2024, 11:33 p.m. UTC | #1

On Sat, 2024-04-13 at 23:39 +0200, Quentin Deslandes wrote:
> Introduce the new print_strings flag in btf_dump_type_data_opts. If
> enabled, libbpf will print character arrays as strings if they meet the
> following conditions:
> - Contains a nul-termination character ('\0')
> - Contains only printable characters before the nul-termination character
> 
> If print_strings is set to false (default value), the existing
> behavior remains unchanged.
> 
> With print_strings=false:
> .str_array = (__u8[14])[
>     'H',
>     'e',
>     'l',
>     'l',
>     'o',
> ],
> 
> With print_strings=true:
> .str_array = (__u8[14])"Hello",
> 
> Signed-off-by: Quentin Deslandes <[e-mail]>
> ---

Hi Quentin,

Thank you for this patch, sorry for the delay reviewing it.
Could you please also add a few tests in
tools/testing/selftests/bpf/prog_tests/btf_dump.c ?

[...]

> @@ -2021,6 +2022,21 @@ static int btf_dump_var_data(struct btf_dump *d,
>  	return btf_dump_dump_type_data(d, NULL, t, type_id, data, 0, 0);
>  }
> 
> +static bool btf_dump_isprint_str(const char *data, unsigned int len)
> +{
> +	unsigned int i;
> +
> +	for (i = 0; i < len; ++i) {
> +		if (data[i] == '\0')
> +			return true;
> +
> +		if (!isprint(data[i]))
> +			return false;

Would it make sense to use isprint_l() and specify something like C locale? 

> +	}
> +
> +	return false;
> +}
> +
>  static int btf_dump_array_data(struct btf_dump *d,
>  			       const struct btf_type *t,
>  			       __u32 id,
> @@ -2047,8 +2063,14 @@ static int btf_dump_array_data(struct btf_dump *d,
>  		 * char arrays, so if size is 1 and element is
>  		 * printable as a char, we'll do that.
>  		 */
> -		if (elem_size == 1)
> +		if (elem_size == 1) {
>  			d->typed_dump->is_array_char = true;
> +			if (d->typed_dump->print_strings &&
> +					btf_dump_isprint_str(data, array->nelems)) {
> +				btf_dump_type_values(d, "\"%s\"", data);

Note: this would have to deal with escape sequences,
otherwise strings containing '\' would be printed incorrectly.

> +				return 0;
> +			}
> +		}
>  	}
> 
>  	/* note that we increment depth before calling btf_dump_print() below;

[...]

Patch

diff --git a/tools/lib/bpf/btf.h b/tools/lib/bpf/btf.h
index 8e6880d91c84..cf190900d483 100644
--- a/tools/lib/bpf/btf.h
+++ b/tools/lib/bpf/btf.h
@@ -284,9 +284,10 @@  struct btf_dump_type_data_opts {
 	bool compact;		/* no newlines/indentation */
 	bool skip_names;	/* skip member/type names */
 	bool emit_zeroes;	/* show 0-valued fields */
+	bool print_strings; /* print char arrays as string */
 	size_t :0;
 };
-#define btf_dump_type_data_opts__last_field emit_zeroes
+#define btf_dump_type_data_opts__last_field print_strings

 LIBBPF_API int
 btf_dump__dump_type_data(struct btf_dump *d, __u32 id,
diff --git a/tools/lib/bpf/btf_dump.c b/tools/lib/bpf/btf_dump.c
index 4d9f30bf7f01..d396a10f37b2 100644
--- a/tools/lib/bpf/btf_dump.c
+++ b/tools/lib/bpf/btf_dump.c
@@ -67,6 +67,7 @@  struct btf_dump_data {
 	bool compact;
 	bool skip_names;
 	bool emit_zeroes;
+	bool print_strings;
 	__u8 indent_lvl;	/* base indent level */
 	char indent_str[BTF_DATA_INDENT_STR_LEN];
 	/* below are used during iteration */
@@ -2021,6 +2022,21 @@  static int btf_dump_var_data(struct btf_dump *d,
 	return btf_dump_dump_type_data(d, NULL, t, type_id, data, 0, 0);
 }

+static bool btf_dump_isprint_str(const char *data, unsigned int len)
+{
+	unsigned int i;
+
+	for (i = 0; i < len; ++i) {
+		if (data[i] == '\0')
+			return true;
+
+		if (!isprint(data[i]))
+			return false;
+	}
+
+	return false;
+}
+
 static int btf_dump_array_data(struct btf_dump *d,
 			       const struct btf_type *t,
 			       __u32 id,
@@ -2047,8 +2063,14 @@  static int btf_dump_array_data(struct btf_dump *d,
 		 * char arrays, so if size is 1 and element is
 		 * printable as a char, we'll do that.
 		 */
-		if (elem_size == 1)
+		if (elem_size == 1) {
 			d->typed_dump->is_array_char = true;
+			if (d->typed_dump->print_strings &&
+					btf_dump_isprint_str(data, array->nelems)) {
+				btf_dump_type_values(d, "\"%s\"", data);
+				return 0;
+			}
+		}
 	}

 	/* note that we increment depth before calling btf_dump_print() below;
@@ -2533,6 +2555,7 @@  int btf_dump__dump_type_data(struct btf_dump *d, __u32 id,
 	d->typed_dump->compact = OPTS_GET(opts, compact, false);
 	d->typed_dump->skip_names = OPTS_GET(opts, skip_names, false);
 	d->typed_dump->emit_zeroes = OPTS_GET(opts, emit_zeroes, false);
+	d->typed_dump->print_strings = OPTS_GET(opts, print_strings, false);

 	ret = btf_dump_dump_type_data(d, NULL, t, id, data, 0, 0);