Re: [PATCH v2 2/3] IMA: Add log statements for failure conditions.

[Tushar Sugandhi <tusharsu@linux.microsoft.com>]

Hi Joe,

On 2020-02-10 7:23 p.m., Joe Perches wrote:
> On Mon, 2020-02-10 at 18:47 -0800, Tushar Sugandhi wrote:
>> process_buffer_measurement() and ima_alloc_key_entry()
>> functions do not have log messages for failure conditions.
> 
> trivia:
> 
>> diff --git a/security/integrity/ima/ima_main.c b/security/integrity/ima/ima_main.c
> []
>> @@ -757,6 +757,9 @@ void process_buffer_measurement(const void *buf, int size,
>>   		ima_free_template_entry(entry);
>>   
>>   out:
>> +	if (ret < 0)
>> +		pr_err("Process buffer measurement failed, result: %d\n", ret);
> 
> perhaps use %s, __func__
> 
> 		pr_err("%s: failed, result: %d\n", __func__, ret);
>   
Sounds good. Will make this change in the next update.
>> diff --git a/security/integrity/ima/ima_queue_keys.c b/security/integrity/ima/ima_queue_keys.c
> []
>> @@ -90,6 +90,7 @@ static struct ima_key_entry *ima_alloc_key_entry(struct key *keyring,
>>   
>>   out:
>>   	if (rc) {
>> +		pr_err("Key entry allocation failed, result: %d\n", rc);
>>   		ima_free_key_entry(entry);
>>   		entry = NULL;
>>   	}
> 
> Likely the pr_err is unnecessary here as kmalloc, kstrdup
> and kmemdup all emit a dump_stack() on allocation failure.
Thanks for pointing out kmalloc, kstrdup, and kmemdup emit a 
dump_stack(). But keeping the above pr_err() will help associate the 
failure with IMA.
For instance - "dmesg | grep ima:" will include this error.
Perhaps I should add __func__ here as well.
And since we are redefining the pr_fmt to prefix module and base names, 
it will help further to pinpoint where exactly the failure is coming from.

> 
> Perhaps instead:
> 
> o Remove unnecessary indentation in ima_free_key_entry by
>    returning early on NULL argument
> o Remove unnecessary rc, tests and label in ima_alloc_key_entry
> ---
>   security/integrity/ima/ima_queue_keys.c | 37 +++++++++++++--------------------
>   1 file changed, 15 insertions(+), 22 deletions(-)
> 
> diff --git a/security/integrity/ima/ima_queue_keys.c b/security/integrity/ima/ima_queue_keys.c
> index c87c722..ba449f 100644
> --- a/security/integrity/ima/ima_queue_keys.c
> +++ b/security/integrity/ima/ima_queue_keys.c
> @@ -58,42 +58,35 @@ void ima_init_key_queue(void)
>   
>   static void ima_free_key_entry(struct ima_key_entry *entry)
>   {
> -	if (entry) {
> -		kfree(entry->payload);
> -		kfree(entry->keyring_name);
> -		kfree(entry);
> -	}
> +	if (!entry)
> +		return;
> +
> +	kfree(entry->payload);
> +	kfree(entry->keyring_name);
> +	kfree(entry);
>   }
>   
Thanks for the feedback. Appreciate it. I would like to make this fix.
But I am not sure if this patchset, which mainly focuses on improving 
logging in IMA, is the right patchset for this.
>   static struct ima_key_entry *ima_alloc_key_entry(struct key *keyring,
>   						 const void *payload,
>   						 size_t payload_len)
>   {
> -	int rc = 0;
>   	struct ima_key_entry *entry;
>   
>   	entry = kzalloc(sizeof(*entry), GFP_KERNEL);
> -	if (entry) {
> -		entry->payload = kmemdup(payload, payload_len, GFP_KERNEL);
> -		entry->keyring_name = kstrdup(keyring->description,
> -					      GFP_KERNEL);
> -		entry->payload_len = payload_len;
> -	}
> -
> -	if ((entry == NULL) || (entry->payload == NULL) ||
> -	    (entry->keyring_name == NULL)) {
> -		rc = -ENOMEM;
> -		goto out;
> -	}
> +	if (!entry)
> +		return NULL;
>   
> -	INIT_LIST_HEAD(&entry->list);
> +	entry->payload = kmemdup(payload, payload_len, GFP_KERNEL);
> +	entry->payload_len = payload_len;
> +	entry->keyring_name = kstrdup(keyring->description, GFP_KERNEL);
>   
> -out:
> -	if (rc) {
> +	if (!entry->payload || !entry->keyring_name) {
>   		ima_free_key_entry(entry);
> -		entry = NULL;
> +		return NULL;
>   	}
>   
> +	INIT_LIST_HEAD(&entry->list);
> +
>   	return entry;
>   }
>   
> 
Thanks again. This recommended change certainly makes the code more 
readable. But again, I am not sure if this patchset is the right one for 
this proposed change.
Perhaps I can create another patchset for the above two recommended 
changes, and only focus on improving logging in this patchset?

Thanks,
Tushar