Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Message-ID: <743452bc4240779e4acafda5c0d3c19482dca38f.camel@linux.ibm.com>
Subject: Re: [PATCH] ima: select ima-buf template for buffer measurement
From:   Mimi Zohar <zohar@linux.ibm.com>
To:     Lakshmi Ramasubramanian <nramas@linux.microsoft.com>
Cc:     tusharsu@linux.microsoft.com, linux-integrity@vger.kernel.org,
        linux-kernel@vger.kernel.org
Date:   Wed, 11 Nov 2020 19:58:30 -0500
In-Reply-To: <20201111205946.503-1-nramas@linux.microsoft.com>
References: <20201111205946.503-1-nramas@linux.microsoft.com>
Content-Type: text/plain; charset="ISO-8859-15"
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hi Lakshmi,

On Wed, 2020-11-11 at 12:59 -0800, Lakshmi Ramasubramanian wrote:
> The default IMA template for measuring buffer should be 'ima-buf' - so
> that the measured buffer is correctly included in the IMA measurement
> log entry. But the default IMA template used for all policy rules is
> the value set for CONFIG_IMA_DEFAULT_TEMPLATE if the policy rule does
> not specify a template. 

The second sentence defines the current status.  The first sentence
describes the problem.  I would reverse the sentence order.

^measuring buffer -> buffer measurements

> IMA does not take into account the template
> requirements of different rules when choosing a default template for
> a given policy rule. This breaks the buffer measurement if the template
> is not provided as part of the rule because the default template could
> be different than 'ima-buf'.

Does the above paragraph add anything new?  Instead describe the
problem.  Perhaps something like:

With the default template format, buffer measurements are added to the
measurement list, but do not include the buffer data, making it
difficult, if not impossible, to validate.  Including "ima-buf"
template records in the measurement list by default, should not impact
existing attestation servers without "ima-buf" template support.

> 
> For example, the following IMA policy rule enables measuring
> the command line arguments passed to the new kernel on kexec system call.
> 
>  measure func=KEXEC_CMDLINE
> 
> The IMA template selected should be 'ima-buf' to have the measured 
> command line arguments included in the IMA measurement log entry.
> Instead the default IMA template is selected, which could be different
> than 'ima-buf'.

When upstreaming a new type of measurement, including an example
provides how to validate the new template data.  Not every patch
description requires an example.

> 
> Initialize a global 'ima-buf' template and select that template,
> by default, for measuring buffer.

Good.
> 
> Signed-off-by: Lakshmi Ramasubramanian <nramas@linux.microsoft.com>
> ---
>  security/integrity/ima/ima.h          |  1 +
>  security/integrity/ima/ima_main.c     | 17 +++++------------
>  security/integrity/ima/ima_policy.c   |  2 +-
>  security/integrity/ima/ima_template.c | 25 +++++++++++++++++++++++++
>  4 files changed, 32 insertions(+), 13 deletions(-)
> 
> diff --git a/security/integrity/ima/ima.h b/security/integrity/ima/ima.h
> index 6ebefec616e4..8e8b1e3cb847 100644
> --- a/security/integrity/ima/ima.h
> +++ b/security/integrity/ima/ima.h
> @@ -156,6 +156,7 @@ int template_desc_init_fields(const char *template_fmt,
>  			      const struct ima_template_field ***fields,
>  			      int *num_fields);
>  struct ima_template_desc *ima_template_desc_current(void);
> +struct ima_template_desc *ima_template_desc_buf(void);
>  struct ima_template_desc *lookup_template_desc(const char *name);
>  bool ima_template_has_modsig(const struct ima_template_desc *ima_template);
>  int ima_restore_measurement_entry(struct ima_template_entry *entry);
> diff --git a/security/integrity/ima/ima_main.c b/security/integrity/ima/ima_main.c
> index a962b23e0429..3646ae763ba9 100644
> --- a/security/integrity/ima/ima_main.c
> +++ b/security/integrity/ima/ima_main.c
> @@ -413,7 +413,7 @@ int ima_file_mmap(struct file *file, unsigned long prot)
>   */
>  int ima_file_mprotect(struct vm_area_struct *vma, unsigned long prot)
>  {
> -	struct ima_template_desc *template;
> +	struct ima_template_desc *template = NULL;
>  	struct file *file = vma->vm_file;
>  	char filename[NAME_MAX];
>  	char *pathbuf = NULL;
> @@ -802,7 +802,7 @@ void process_buffer_measurement(struct inode *inode, const void *buf, int size,
>  					    .filename = eventname,
>  					    .buf = buf,
>  					    .buf_len = size};
> -	struct ima_template_desc *template = NULL;
> +	struct ima_template_desc *template = ima_template_desc_buf();
>  	struct {
>  		struct ima_digest_data hdr;
>  		char digest[IMA_MAX_DIGEST_SIZE];
> @@ -833,16 +833,9 @@ void process_buffer_measurement(struct inode *inode, const void *buf, int size,
>  		pcr = CONFIG_IMA_MEASURE_PCR_IDX;
>  
>  	if (!template) {
> -		template = lookup_template_desc("ima-buf");
> -		ret = template_desc_init_fields(template->fmt,
> -						&(template->fields),
> -						&(template->num_fields));
> -		if (ret < 0) {
> -			pr_err("template %s init failed, result: %d\n",
> -			       (strlen(template->name) ?
> -				template->name : template->fmt), ret);
> -			return;
> -		}
> +		ret = -EINVAL;
> +		audit_cause = "ima_template_desc_buf";
> +		goto out;

Normally a test follows the variable assignment, but in this case, the
check is being deferred in case there isn't a policy rule.

>  	}
>  
>  	iint.ima_hash = &hash.hdr;
> diff --git a/security/integrity/ima/ima_policy.c b/security/integrity/ima/ima_policy.c
> index 9b5adeaa47fc..823a0c1379cb 100644
> --- a/security/integrity/ima/ima_policy.c
> +++ b/security/integrity/ima/ima_policy.c
> @@ -628,7 +628,7 @@ int ima_match_policy(struct inode *inode, const struct cred *cred, u32 secid,
>  	struct ima_rule_entry *entry;
>  	int action = 0, actmask = flags | (flags << 1);
>  
> -	if (template_desc)
> +	if (template_desc && !*template_desc)
>  		*template_desc = ima_template_desc_current();
>  
>  	rcu_read_lock();
> diff --git a/security/integrity/ima/ima_template.c b/security/integrity/ima/ima_template.c
> index 1e89e2d3851f..e53fce2c1610 100644
> --- a/security/integrity/ima/ima_template.c
> +++ b/security/integrity/ima/ima_template.c
> @@ -55,6 +55,7 @@ static const struct ima_template_field supported_fields[] = {
>  #define MAX_TEMPLATE_NAME_LEN sizeof("d-ng|n-ng|sig|buf|d-modisg|modsig")
>  
>  static struct ima_template_desc *ima_template;
> +static struct ima_template_desc *ima_buf_template;
>  
>  /**
>   * ima_template_has_modsig - Check whether template has modsig-related fields.
> @@ -252,6 +253,30 @@ struct ima_template_desc *ima_template_desc_current(void)
>  	return ima_template;
>  }
>  
> +struct ima_template_desc *ima_template_desc_buf(void)
> +{
> +	struct ima_template_desc *template = NULL;
> +	int ret = 0;
> +
> +	if (ima_buf_template)
> +		return ima_buf_template;
> +
> +	ima_init_template_list();
> +	template = lookup_template_desc("ima-buf");
> +	if (!template)
> +		return NULL;
> +
> +	ret = template_desc_init_fields(template->fmt,
> +					&(template->fields),
> +					&(template->num_fields));
> +	if (ret)
> +		return NULL;

Instead of initializing the fields here, maybe it should be done in
ima_init_template()?  That would remove the deferred !template test in
process_buffer_measurement() and would also simplify this function.

thanks,

Mimi

> +
> +	ima_buf_template = template;
> +
> +	return ima_buf_template;
> +}
> +
>  int __init ima_init_template(void)
>  {
>  	struct ima_template_desc *template = ima_template_desc_current();