On 5/14/2019 7:07 AM, prakhar srivastava wrote:
On Mon, May 13, 2019 at 6:48 AM Roberto Sassu <roberto.sassu@xxxxxxxxxx> wrote:
On 5/11/2019 12:37 AM, Prakhar Srivastava wrote:
From: Prakhar Srivastava <prsriva02@xxxxxxxxx>
The buffer(cmdline args) added to the ima log cannot be attested
without having the actual buffer. Thus to make the measured buffer
available to store/read a new ima template (buf) is added.
Hi Prakhar
please fix the typos. More comments below.
+ buffer_event_data->type = IMA_XATTR_BUFFER;
+ buffer_event_data->buf_length = size;
+ memcpy(buffer_event_data->buf, buf, size);
+
+ event_data.xattr_value = (struct evm_ima_xattr_data *)buffer_event_data;
+ event_data.xattr_len = alloc_length;
I would prefer that you introduce two new fields in the ima_event_data
structure. You can initialize them directly with the parameters of
process_buffer_measurement().
I will make the edits, this will definitely save the kzalloc in this code
path.
ima_write_template_field_data() will make
a copy.
Since event_data->type is used to distinguish what the template field
should contain.
Removing the type and subsequent check in the template_init,
buf template fmt will result in the whole event_Data structure
being added to the log, which is not the expected output.
For buffer entries, the buf templet fmt will contains the buffer itself.
The purpose of ima_event_data is to pass data to the init method of
template fields. Each method takes the data it needs.
If you pass event_data->buf and event_data->buf_len to
ima_write_template_field_data() this should be fine.
Roberto
+ .field_show = ima_show_template_buf},
Please update Documentation/security/IMA-templates.rst
Will update the documentation.
Thanks,
Prakhar Srivastava
Thanks
Roberto
--
HUAWEI TECHNOLOGIES Duesseldorf GmbH, HRB 56063
Managing Director: Bo PENG, Jian LI, Yanli SHI
