Re: [PATCH v3 5/6] IMA: add hook to measure critical data from kernel components

[Tushar Sugandhi]

On 2020-08-31 11:23 a.m., Mimi Zohar wrote:
diff --git a/security/integrity/ima/ima_main.c 
b/security/integrity/ima/ima_main.c
index 52cbbc1f7ea2..a889bf40cb7e 100644
--- a/security/integrity/ima/ima_main.c
+++ b/security/integrity/ima/ima_main.c
@@ -869,6 +869,30 @@ void ima_kexec_cmdline(int kernel_fd, const void *buf, int 
size)
        fdput(f);
  }
  
+/**
+ * ima_measure_critical_data - measure critical data
+ * @event_name: name for the given data
+ * @event_data_source: name of the event data source
+ * @buf: pointer to buffer containing data to measure
+ * @buf_len: length of buffer(in bytes)
+ * @measure_buf_hash: if set to true - will measure hash of the buf,
+ *                    instead of buf
+ *
+ * Buffers can only be measured, not appraised.
+ */
+int ima_measure_critical_data(const char *event_name,
+                             const char *event_data_source,
+                             const void *buf, int buf_len,
+                             bool measure_buf_hash)
+{
+       if (!event_name || !event_data_source || !buf || !buf_len)
+               return -EINVAL;
+
+       return process_buffer_measurement(NULL, buf, buf_len, event_name,
+                                         CRITICAL_DATA, 0, event_data_source,
+                                         measure_buf_hash);

This is exactly what I'm concerned about.  Failure to measure data may
be audited, but should never fail.

Mimi

As I responded in patch 2, I can ignore the result of 
process_buffer_measurement() in ima_measure_critical_data(), and make
ima_measure_critical_data() return type as "void".

But that’s the only place where the results of p_b_m() are being used.
So I might as well just revert the return type of p_b_m() to the
original "void".

+}