Re: [PATCH v9 2/8] IMA: add support to measure buffer data hash

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

<snip>

>>>   void process_buffer_measurement(struct inode *inode, const void 
>>> *buf, int size,
>>>                   const char *eventname, enum ima_hooks func,
>>> -                int pcr, const char *func_data);
>>> +                int pcr, const char *func_data,
>>> +                bool measure_buf_hash);
>>
>> Please abbreviate the boolean name to "hash".   The test would then be
>> "if (hash == true)" or "if (hash)".
>>
> Will do.

<snip>

>>> - * process_buffer_measurement - Measure the buffer to ima log.
>>> + * process_buffer_measurement - Measure the buffer or the buffer 
>>> data hash
>>>    * @inode: inode associated with the object being measured (NULL 
>>> for KEY_CHECK)
>>>    * @buf: pointer to the buffer that needs to be added to the log.
>>>    * @size: size of buffer(in bytes).
>>> @@ -787,12 +787,23 @@ int ima_post_load_data(char *buf, loff_t size,
>>>    * @func: IMA hook
>>>    * @pcr: pcr to extend the measurement
>>>    * @func_data: private data specific to @func, can be NULL.
>>> + * @measure_buf_hash: measure buffer hash
>>
>> ^@hash: measure buffer data hash
>>
> Agreed. Will fix.
<snip>
>>>   void process_buffer_measurement(struct inode *inode, const void 
>>> *buf, int size,
>>>                   const char *eventname, enum ima_hooks func,
>>> -                int pcr, const char *func_data)
>>> +                int pcr, const char *func_data,
>>> +                bool measure_buf_hash)
>>>   {
>>>       int ret = 0;
>>>       const char *audit_cause = "ENOMEM";
>>> @@ -807,6 +818,8 @@ void process_buffer_measurement(struct inode 
>>> *inode, const void *buf, int size,
>>>           struct ima_digest_data hdr;
>>>           char digest[IMA_MAX_DIGEST_SIZE];
>>>       } hash = {};
>>> +    char buf_hash[IMA_MAX_DIGEST_SIZE];
>>> +    int buf_hash_len = hash_digest_size[ima_hash_algo];
>>>       int violation = 0;
>>>       int action = 0;
>>>       u32 secid;
>>> @@ -849,13 +862,27 @@ void process_buffer_measurement(struct inode 
>>> *inode, const void *buf, int size,
>>>           goto out;
>>>       }
>>> +    if (measure_buf_hash) {
>>
>> ^ if (hash) {
> Yes.
>>> +        memcpy(buf_hash, hash.hdr.digest, buf_hash_len);
>>> +
>>> +        ret = ima_calc_buffer_hash(buf_hash, buf_hash_len,
>>> +                       iint.ima_hash);
>>> +        if (ret < 0) {
>>> +            audit_cause = "measure_buf_hash_error";


Hi Mimi,
There already exist a local struct variable named "hash" in p_b_m().
I was thinking of using "buf_hash", but that one is taken too.
Maybe I should use "buf_hash" for the input bool, and rename the
existing "buf_hash" local variable to "digest_hash"?
Does it sound ok?

Thanks,
Tushar


<snip>