[bpf-next,4/4] selftests/bpf: Add a null pointer check for the serial_test_tp_attach_query

Commit Message

Kunwu April 24, 2024, 2:04 a.m. UTC

There is a 'malloc' call, which can be unsuccessful.
Add the malloc failure checking to avoid possible null
dereference.

Signed-off-by: Kunwu Chan <[e-mail]>
---
 tools/testing/selftests/bpf/prog_tests/tp_attach_query.c | 3 +++
 1 file changed, 3 insertions(+)

Comments

Markus Elfring April 24, 2024, 10:48 a.m. UTC | #1

…
> Add the malloc failure checking to avoid possible null
> dereference.
…

How do you think about to use the following wording variant?

   Add a return value check so that a null pointer dereference will be avoided
   after a memory allocation failure.


Would you like to add the tag “Fixes” accordingly?

Regards,
Markus

Daniel Borkmann May 3, 2024, 3:47 p.m. UTC | #2

On 4/24/24 4:04 AM, Kunwu Chan wrote:
> There is a 'malloc' call, which can be unsuccessful.
> Add the malloc failure checking to avoid possible null
> dereference.
> 
> Signed-off-by: Kunwu Chan <[e-mail]>
> ---
>   tools/testing/selftests/bpf/prog_tests/tp_attach_query.c | 3 +++
>   1 file changed, 3 insertions(+)
> 
> diff --git a/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c b/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c
> index 655d69f0ff0b..302b25408a53 100644
> --- a/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c
> +++ b/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c
> @@ -39,6 +39,9 @@ void serial_test_tp_attach_query(void)
>   	attr.wakeup_events = 1;
>   
>   	query = malloc(sizeof(*query) + sizeof(__u32) * num_progs);
> +	if (CHECK(!query, "malloc()", "error:%s\n", strerror(errno)))

Series looks reasonable, small nit on CHECK() : Lets use ASSERT*() macros given they are
preferred over the latter :

if (!ASSERT_OK_PTR(buf, "malloc"))

> +		return;
> +
>   	for (i = 0; i < num_progs; i++) {
>   		err = bpf_prog_test_load(file, BPF_PROG_TYPE_TRACEPOINT, &obj[i],
>   				    &prog_fd[i]);
>

Daniel Borkmann May 3, 2024, 3:51 p.m. UTC | #3

On 5/3/24 5:47 PM, Daniel Borkmann wrote:
> On 4/24/24 4:04 AM, Kunwu Chan wrote:
>> There is a 'malloc' call, which can be unsuccessful.
>> Add the malloc failure checking to avoid possible null
>> dereference.
>>
>> Signed-off-by: Kunwu Chan <[e-mail]>
>> ---
>>   tools/testing/selftests/bpf/prog_tests/tp_attach_query.c | 3 +++
>>   1 file changed, 3 insertions(+)
>>
>> diff --git a/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c b/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c
>> index 655d69f0ff0b..302b25408a53 100644
>> --- a/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c
>> +++ b/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c
>> @@ -39,6 +39,9 @@ void serial_test_tp_attach_query(void)
>>       attr.wakeup_events = 1;
>>       query = malloc(sizeof(*query) + sizeof(__u32) * num_progs);
>> +    if (CHECK(!query, "malloc()", "error:%s\n", strerror(errno)))
> 
> Series looks reasonable, small nit on CHECK() : Lets use ASSERT*() macros given they are
> preferred over the latter :
> 
> if (!ASSERT_OK_PTR(buf, "malloc"))

( Also as a side-note: Fixes tag on all these patches is not needed given this will just
   end up spamming stable tree. If you indeed end up with NULL then the tests will just
   segfault & fail. )

>> +        return;
>> +
>>       for (i = 0; i < num_progs; i++) {
>>           err = bpf_prog_test_load(file, BPF_PROG_TYPE_TRACEPOINT, &obj[i],
>>                       &prog_fd[i]);
>>
>

Kunwu Chan May 10, 2024, 8:21 a.m. UTC | #4

Thanks all for your reply.

On 2024/5/3 23:47, Daniel Borkmann wrote:
> On 4/24/24 4:04 AM, Kunwu Chan wrote:
>> There is a 'malloc' call, which can be unsuccessful.
>> Add the malloc failure checking to avoid possible null
>> dereference.
>>
>> Signed-off-by: Kunwu Chan <[e-mail]>
>> ---
>>   tools/testing/selftests/bpf/prog_tests/tp_attach_query.c | 3 +++
>>   1 file changed, 3 insertions(+)
>>
>> diff --git a/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c 
>> b/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c
>> index 655d69f0ff0b..302b25408a53 100644
>> --- a/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c
>> +++ b/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c
>> @@ -39,6 +39,9 @@ void serial_test_tp_attach_query(void)
>>       attr.wakeup_events = 1;
>>         query = malloc(sizeof(*query) + sizeof(__u32) * num_progs);
>> +    if (CHECK(!query, "malloc()", "error:%s\n", strerror(errno)))
>
> Series looks reasonable, small nit on CHECK() : Lets use ASSERT*() 
> macros given they are
> preferred over the latter :
>
> if (!ASSERT_OK_PTR(buf, "malloc"))

Thanks, I'll update it in v2:

1: Use ASSERT_OK_PTR instead of CHECK

2: Add a suggested-by tag for you

>
>> +        return;
>> +
>>       for (i = 0; i < num_progs; i++) {
>>           err = bpf_prog_test_load(file, BPF_PROG_TYPE_TRACEPOINT, 
>> &obj[i],
>>                       &prog_fd[i]);
>>
>

Patch

diff --git a/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c b/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c
index 655d69f0ff0b..302b25408a53 100644
--- a/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c
+++ b/tools/testing/selftests/bpf/prog_tests/tp_attach_query.c
@@ -39,6 +39,9 @@  void serial_test_tp_attach_query(void)
 	attr.wakeup_events = 1;
 
 	query = malloc(sizeof(*query) + sizeof(__u32) * num_progs);
+	if (CHECK(!query, "malloc()", "error:%s\n", strerror(errno)))
+		return;
+
 	for (i = 0; i < num_progs; i++) {
 		err = bpf_prog_test_load(file, BPF_PROG_TYPE_TRACEPOINT, &obj[i],
 				    &prog_fd[i]);