[PATCH 0/3] support for duplicate measurement of integrity critical data

IMA does not measure duplicate buffer data since TPM extend is a very
expensive operation.  However, in some cases for integrity critical
data, the measurement of duplicate data is necessary to accurately
determine the current state of the system.  Eg, SELinux state changing
from 'audit', to 'enforcing', and back to 'audit' again.  In this
example, currently, IMA will not measure the last state change to
'audit'.  This limits the ability of attestation services to accurately
determine the current state of the integrity critical data on the
system.

This series addresses this gap by providing the ability to measure
duplicate entries for integrity critical data, driven by policy.

This series is based on the following repo/branch/commit:
 repo: https://git.kernel.org/pub/scm/linux/kernel/git/zohar/linux-integrity.git
 branch: next-integrity-testing
 commit b3f82afc1041 ("IMA: Measure kernel version in early boot") 

Tushar Sugandhi (3):
  IMA: add policy condition to measure duplicate critical data
  IMA: update functions to read allow_dup policy condition
  IMA: add support to measure duplicate buffer for critical data hook

 Documentation/ABI/testing/ima_policy  |  4 +++-
 security/integrity/ima/ima.h          |  8 +++----
 security/integrity/ima/ima_api.c      | 15 +++++++------
 security/integrity/ima/ima_appraise.c |  2 +-
 security/integrity/ima/ima_init.c     |  2 +-
 security/integrity/ima/ima_main.c     |  9 ++++----
 security/integrity/ima/ima_policy.c   | 31 ++++++++++++++++++++++++---
 security/integrity/ima/ima_queue.c    |  5 +++--
 8 files changed, 54 insertions(+), 22 deletions(-)

-- 
2.17.1

IMA functions ima_get_action() and ima_match_policy() do not consume the
policy condition to allow measuring duplicate entries for integrity
critical data.

Update ima_get_action() and ima_match_policy() to consume the IMA policy
condition to measure duplicate buffer entries for integrity critical
data.

Signed-off-by: Tushar Sugandhi <redacted>
---
 security/integrity/ima/ima.h          | 4 ++--
 security/integrity/ima/ima_api.c      | 6 ++++--
 security/integrity/ima/ima_appraise.c | 2 +-
 security/integrity/ima/ima_main.c     | 6 +++---
 security/integrity/ima/ima_policy.c   | 7 ++++++-
 5 files changed, 16 insertions(+), 9 deletions(-)

diff --git a/security/integrity/ima/ima.h b/security/integrity/ima/ima.h
index aa312472c7c5..59324173497f 100644
--- a/security/integrity/ima/ima.h
+++ b/security/integrity/ima/ima.h

@@ -257,7 +257,7 @@ static inline void ima_process_queued_keys(void) {}
 int ima_get_action(struct inode *inode, const struct cred *cred, u32 secid,
 		   int mask, enum ima_hooks func, int *pcr,
 		   struct ima_template_desc **template_desc,
-		   const char *func_data);
+		   const char *func_data, bool *allow_dup);
 int ima_must_measure(struct inode *inode, int mask, enum ima_hooks func);
 int ima_collect_measurement(struct integrity_iint_cache *iint,
 			    struct file *file, void *buf, loff_t size,

@@ -286,7 +286,7 @@ const char *ima_d_path(const struct path *path, char **pathbuf, char *filename);
 int ima_match_policy(struct inode *inode, const struct cred *cred, u32 secid,
 		     enum ima_hooks func, int mask, int flags, int *pcr,
 		     struct ima_template_desc **template_desc,
-		     const char *func_data);
+		     const char *func_data, bool *allow_dup);
 void ima_init_policy(void);
 void ima_update_policy(void);
 void ima_update_policy_flag(void);

diff --git a/security/integrity/ima/ima_api.c b/security/integrity/ima/ima_api.c
index 1dd70dc68ffd..d273373e6be9 100644
--- a/security/integrity/ima/ima_api.c
+++ b/security/integrity/ima/ima_api.c

@@ -171,6 +171,8 @@ void ima_add_violation(struct file *file, const unsigned char *filename,
  * @pcr: pointer filled in if matched measure policy sets pcr=
  * @template_desc: pointer filled in if matched measure policy sets template=
  * @func_data: func specific data, may be NULL
+ * @allow_dup: pointer filled in to decide if a duplicate buffer entry
+ *             should be measured
  *
  * The policy is defined in terms of keypairs:
  *		subj=, obj=, type=, func=, mask=, fsmagic=

@@ -186,14 +188,14 @@ void ima_add_violation(struct file *file, const unsigned char *filename,
 int ima_get_action(struct inode *inode, const struct cred *cred, u32 secid,
 		   int mask, enum ima_hooks func, int *pcr,
 		   struct ima_template_desc **template_desc,
-		   const char *func_data)
+		   const char *func_data, bool *allow_dup)
 {
 	int flags = IMA_MEASURE | IMA_AUDIT | IMA_APPRAISE | IMA_HASH;
 
 	flags &= ima_policy_flag;
 
 	return ima_match_policy(inode, cred, secid, func, mask, flags, pcr,
-				template_desc, func_data);
+				template_desc, func_data, allow_dup);
 }
 
 /*

diff --git a/security/integrity/ima/ima_appraise.c b/security/integrity/ima/ima_appraise.c
index 46ffa38bab12..e317a7698a47 100644
--- a/security/integrity/ima/ima_appraise.c
+++ b/security/integrity/ima/ima_appraise.c

@@ -77,7 +77,7 @@ int ima_must_appraise(struct inode *inode, int mask, enum ima_hooks func)
 
 	security_task_getsecid(current, &secid);
 	return ima_match_policy(inode, current_cred(), secid, func, mask,
-				IMA_APPRAISE | IMA_HASH, NULL, NULL, NULL);
+				IMA_APPRAISE | IMA_HASH, NULL, NULL, NULL, NULL);
 }
 
 static int ima_fix_xattr(struct dentry *dentry,

diff --git a/security/integrity/ima/ima_main.c b/security/integrity/ima/ima_main.c
index 6a429846f90a..2774139845b6 100644
--- a/security/integrity/ima/ima_main.c
+++ b/security/integrity/ima/ima_main.c

@@ -219,7 +219,7 @@ static int process_measurement(struct file *file, const struct cred *cred,
 	 * Included is the appraise submask.
 	 */
 	action = ima_get_action(inode, cred, secid, mask, func, &pcr,
-				&template_desc, NULL);
+				&template_desc, NULL, NULL);
 	violation_check = ((func == FILE_CHECK || func == MMAP_CHECK) &&
 			   (ima_policy_flag & IMA_MEASURE));
 	if (!action && !violation_check)

@@ -432,7 +432,7 @@ int ima_file_mprotect(struct vm_area_struct *vma, unsigned long prot)
 	security_task_getsecid(current, &secid);
 	inode = file_inode(vma->vm_file);
 	action = ima_get_action(inode, current_cred(), secid, MAY_EXEC,
-				MMAP_CHECK, &pcr, &template, 0);
+				MMAP_CHECK, &pcr, &template, 0, NULL);
 
 	/* Is the mmap'ed file in policy? */
 	if (!(action & (IMA_MEASURE | IMA_APPRAISE_SUBMASK)))

@@ -865,7 +865,7 @@ void process_buffer_measurement(struct inode *inode, const void *buf, int size,
 	if (func) {
 		security_task_getsecid(current, &secid);
 		action = ima_get_action(inode, current_cred(), secid, 0, func,
-					&pcr, &template, func_data);
+					&pcr, &template, func_data, NULL);
 		if (!(action & IMA_MEASURE))
 			return;
 	}

diff --git a/security/integrity/ima/ima_policy.c b/security/integrity/ima/ima_policy.c
index b89eb768dd05..4147c677eb24 100644
--- a/security/integrity/ima/ima_policy.c
+++ b/security/integrity/ima/ima_policy.c

@@ -644,6 +644,8 @@ static int get_subaction(struct ima_rule_entry *rule, enum ima_hooks func)
  * @pcr: set the pcr to extend
  * @template_desc: the template that should be used for this rule
  * @func_data: func specific data, may be NULL
+ * @allow_dup: pointer filled in to decide if a duplicate buffer entry
+ *             should be measured
  *
  * Measure decision based on func/mask/fsmagic and LSM(subj/obj/type)
  * conditions.

@@ -655,7 +657,7 @@ static int get_subaction(struct ima_rule_entry *rule, enum ima_hooks func)
 int ima_match_policy(struct inode *inode, const struct cred *cred, u32 secid,
 		     enum ima_hooks func, int mask, int flags, int *pcr,
 		     struct ima_template_desc **template_desc,
-		     const char *func_data)
+		     const char *func_data, bool *allow_dup)
 {
 	struct ima_rule_entry *entry;
 	int action = 0, actmask = flags | (flags << 1);

@@ -673,6 +675,9 @@ int ima_match_policy(struct inode *inode, const struct cred *cred, u32 secid,
 				     func_data))
 			continue;
 
+		if ((allow_dup) && (func == CRITICAL_DATA))
+			*allow_dup = entry->allow_dup;
+
 		action |= entry->flags & IMA_ACTION_FLAGS;
 
 		action |= entry->action & IMA_DO_MASK;

-- 
2.17.1

process_buffer_measurement() and the underlying functions do not use the
policy condition to measure duplicate buffer entries for integrity
critical data.

Update process_buffer_measurement(), ima_add_template_entry(), and
ima_store_template() to use the policy condition to decide if a
duplicate buffer entry for integrity critical data should be measured.

Signed-off-by: Tushar Sugandhi <redacted>
---
 security/integrity/ima/ima.h       | 4 ++--
 security/integrity/ima/ima_api.c   | 9 +++++----
 security/integrity/ima/ima_init.c  | 2 +-
 security/integrity/ima/ima_main.c  | 5 +++--
 security/integrity/ima/ima_queue.c | 5 +++--
 5 files changed, 14 insertions(+), 11 deletions(-)

diff --git a/security/integrity/ima/ima.h b/security/integrity/ima/ima.h
index 59324173497f..b06732560949 100644
--- a/security/integrity/ima/ima.h
+++ b/security/integrity/ima/ima.h

@@ -139,7 +139,7 @@ int ima_init(void);
 int ima_fs_init(void);
 int ima_add_template_entry(struct ima_template_entry *entry, int violation,
 			   const char *op, struct inode *inode,
-			   const unsigned char *filename);
+			   const unsigned char *filename, bool allow_dup);
 int ima_calc_file_hash(struct file *file, struct ima_digest_data *hash);
 int ima_calc_buffer_hash(const void *buf, loff_t len,
 			 struct ima_digest_data *hash);

@@ -278,7 +278,7 @@ int ima_alloc_init_template(struct ima_event_data *event_data,
 			    struct ima_template_desc *template_desc);
 int ima_store_template(struct ima_template_entry *entry, int violation,
 		       struct inode *inode,
-		       const unsigned char *filename, int pcr);
+		       const unsigned char *filename, int pcr, bool allow_dup);
 void ima_free_template_entry(struct ima_template_entry *entry);
 const char *ima_d_path(const struct path *path, char **pathbuf, char *filename);
 

diff --git a/security/integrity/ima/ima_api.c b/security/integrity/ima/ima_api.c
index d273373e6be9..f84369f9905e 100644
--- a/security/integrity/ima/ima_api.c
+++ b/security/integrity/ima/ima_api.c

@@ -101,7 +101,7 @@ int ima_alloc_init_template(struct ima_event_data *event_data,
  */
 int ima_store_template(struct ima_template_entry *entry,
 		       int violation, struct inode *inode,
-		       const unsigned char *filename, int pcr)
+		       const unsigned char *filename, int pcr, bool allow_dup)
 {
 	static const char op[] = "add_template_measure";
 	static const char audit_cause[] = "hashing_error";

@@ -119,7 +119,8 @@ int ima_store_template(struct ima_template_entry *entry,
 		}
 	}
 	entry->pcr = pcr;
-	result = ima_add_template_entry(entry, violation, op, inode, filename);
+	result = ima_add_template_entry(entry, violation, op, inode, filename,
+					allow_dup);
 	return result;
 }
 

@@ -152,7 +153,7 @@ void ima_add_violation(struct file *file, const unsigned char *filename,
 		goto err_out;
 	}
 	result = ima_store_template(entry, violation, inode,
-				    filename, CONFIG_IMA_MEASURE_PCR_IDX);
+				    filename, CONFIG_IMA_MEASURE_PCR_IDX, false);
 	if (result < 0)
 		ima_free_template_entry(entry);
 err_out:

@@ -330,7 +331,7 @@ void ima_store_measurement(struct integrity_iint_cache *iint,
 		return;
 	}
 
-	result = ima_store_template(entry, violation, inode, filename, pcr);
+	result = ima_store_template(entry, violation, inode, filename, pcr, false);
 	if ((!result || result == -EEXIST) && !(file->f_flags & O_DIRECT)) {
 		iint->flags |= IMA_MEASURED;
 		iint->measured_pcrs |= (0x1 << pcr);

diff --git a/security/integrity/ima/ima_init.c b/security/integrity/ima/ima_init.c
index 6e8742916d1d..d0a79d7b8d89 100644
--- a/security/integrity/ima/ima_init.c
+++ b/security/integrity/ima/ima_init.c

@@ -88,7 +88,7 @@ static int __init ima_add_boot_aggregate(void)
 
 	result = ima_store_template(entry, violation, NULL,
 				    boot_aggregate_name,
-				    CONFIG_IMA_MEASURE_PCR_IDX);
+				    CONFIG_IMA_MEASURE_PCR_IDX, false);
 	if (result < 0) {
 		ima_free_template_entry(entry);
 		audit_cause = "store_entry";

diff --git a/security/integrity/ima/ima_main.c b/security/integrity/ima/ima_main.c
index 2774139845b6..ff6d15d7594c 100644
--- a/security/integrity/ima/ima_main.c
+++ b/security/integrity/ima/ima_main.c

@@ -843,6 +843,7 @@ void process_buffer_measurement(struct inode *inode, const void *buf, int size,
 	int digest_hash_len = hash_digest_size[ima_hash_algo];
 	int violation = 0;
 	int action = 0;
+	bool allow_dup = false;
 	u32 secid;
 
 	if (!ima_policy_flag)

@@ -865,7 +866,7 @@ void process_buffer_measurement(struct inode *inode, const void *buf, int size,
 	if (func) {
 		security_task_getsecid(current, &secid);
 		action = ima_get_action(inode, current_cred(), secid, 0, func,
-					&pcr, &template, func_data, NULL);
+					&pcr, &template, func_data, &allow_dup);
 		if (!(action & IMA_MEASURE))
 			return;
 	}

@@ -903,7 +904,7 @@ void process_buffer_measurement(struct inode *inode, const void *buf, int size,
 		goto out;
 	}
 
-	ret = ima_store_template(entry, violation, NULL, event_data.buf, pcr);
+	ret = ima_store_template(entry, violation, NULL, event_data.buf, pcr, allow_dup);
 	if (ret < 0) {
 		audit_cause = "store_entry";
 		ima_free_template_entry(entry);

diff --git a/security/integrity/ima/ima_queue.c b/security/integrity/ima/ima_queue.c
index c096ef8945c7..fbf359495fa8 100644
--- a/security/integrity/ima/ima_queue.c
+++ b/security/integrity/ima/ima_queue.c

@@ -158,7 +158,7 @@ static int ima_pcr_extend(struct tpm_digest *digests_arg, int pcr)
  */
 int ima_add_template_entry(struct ima_template_entry *entry, int violation,
 			   const char *op, struct inode *inode,
-			   const unsigned char *filename)
+			   const unsigned char *filename, bool allow_dup)
 {
 	u8 *digest = entry->digests[ima_hash_algo_idx].digest;
 	struct tpm_digest *digests_arg = entry->digests;

@@ -169,7 +169,8 @@ int ima_add_template_entry(struct ima_template_entry *entry, int violation,
 
 	mutex_lock(&ima_extend_list_mutex);
 	if (!violation) {
-		if (ima_lookup_digest_entry(digest, entry->pcr)) {
+		if (!allow_dup &&
+		    ima_lookup_digest_entry(digest, entry->pcr)) {
 			audit_cause = "hash_exists";
 			result = -EEXIST;
 			goto out;

-- 
2.17.1

IMA needs to support duplicate measurements of integrity
critical data to accurately determine the current state of that data
on the system.  Further, since measurement of duplicate data is not
required for all the use cases, it needs to be policy driven.

Define "allow_dup", a new IMA policy condition, for the IMA func
CRITICAL_DATA to allow duplicate buffer measurement of integrity
critical data.

Limit the ability to measure duplicate buffer data when action is
"measure" and func is CRITICAL_DATA.

Signed-off-by: Tushar Sugandhi <redacted>
---
 Documentation/ABI/testing/ima_policy |  4 +++-
 security/integrity/ima/ima_policy.c  | 24 ++++++++++++++++++++++--
 2 files changed, 25 insertions(+), 3 deletions(-)

diff --git a/Documentation/ABI/testing/ima_policy b/Documentation/ABI/testing/ima_policy
index bc8e1cbe5e61..9598287e3bbf 100644
--- a/Documentation/ABI/testing/ima_policy
+++ b/Documentation/ABI/testing/ima_policy

@@ -27,7 +27,7 @@ Description:
 			lsm:	[[subj_user=] [subj_role=] [subj_type=]
 				 [obj_user=] [obj_role=] [obj_type=]]
 			option:	[[appraise_type=]] [template=] [permit_directio]
-				[appraise_flag=] [keyrings=]
+				[appraise_flag=] [keyrings=] [allow_dup]
 		  base:
 			func:= [BPRM_CHECK][MMAP_CHECK][CREDS_CHECK][FILE_CHECK]MODULE_CHECK]
 			        [FIRMWARE_CHECK]

@@ -55,6 +55,8 @@ Description:
 			label:= [selinux]|[kernel_info]|[data_label]
 			data_label:= a unique string used for grouping and limiting critical data.
 			For example, "selinux" to measure critical data for SELinux.
+			allow_dup allows measurement of duplicate data.  Only valid
+			when action is "measure" and func is CRITICAL_DATA.
 
 		  default policy:
 			# PROC_SUPER_MAGIC

diff --git a/security/integrity/ima/ima_policy.c b/security/integrity/ima/ima_policy.c
index 9b45d064a87d..b89eb768dd05 100644
--- a/security/integrity/ima/ima_policy.c
+++ b/security/integrity/ima/ima_policy.c

@@ -35,6 +35,7 @@
 #define IMA_FSNAME	0x0200
 #define IMA_KEYRINGS	0x0400
 #define IMA_LABEL	0x0800
+#define IMA_ALLOW_DUP	0x1000
 
 #define UNKNOWN		0
 #define MEASURE		0x0001	/* same as IMA_MEASURE */

@@ -87,6 +88,7 @@ struct ima_rule_entry {
 	char *fsname;
 	struct ima_rule_opt_list *keyrings; /* Measure keys added to these keyrings */
 	struct ima_rule_opt_list *label; /* Measure data grouped under this label */
+	bool allow_dup; /* Allow duplicate buffer entry measurement */
 	struct ima_template_desc *template;
 };
 

@@ -942,7 +944,7 @@ enum {
 	Opt_uid_lt, Opt_euid_lt, Opt_fowner_lt,
 	Opt_appraise_type, Opt_appraise_flag,
 	Opt_permit_directio, Opt_pcr, Opt_template, Opt_keyrings,
-	Opt_label, Opt_err
+	Opt_label, Opt_allow_dup, Opt_err
 };
 
 static const match_table_t policy_tokens = {

@@ -980,6 +982,7 @@ static const match_table_t policy_tokens = {
 	{Opt_template, "template=%s"},
 	{Opt_keyrings, "keyrings=%s"},
 	{Opt_label, "label=%s"},
+	{Opt_allow_dup, "allow_dup"},
 	{Opt_err, NULL}
 };
 

@@ -1148,7 +1151,7 @@ static bool ima_validate_rule(struct ima_rule_entry *entry)
 			return false;
 
 		if (entry->flags & ~(IMA_FUNC | IMA_UID | IMA_PCR |
-				     IMA_LABEL))
+				     IMA_LABEL | IMA_ALLOW_DUP))
 			return false;
 
 		if (ima_rule_contains_lsm_cond(entry))

@@ -1184,6 +1187,7 @@ static int ima_parse_rule(char *rule, struct ima_rule_entry *entry)
 	entry->uid_op = &uid_eq;
 	entry->fowner_op = &uid_eq;
 	entry->action = UNKNOWN;
+	entry->allow_dup = false;
 	while ((p = strsep(&rule, " \t")) != NULL) {
 		substring_t args[MAX_OPT_ARGS];
 		int token;

@@ -1375,6 +1379,19 @@ static int ima_parse_rule(char *rule, struct ima_rule_entry *entry)
 
 			entry->flags |= IMA_LABEL;
 			break;
+		case Opt_allow_dup:
+			ima_log_string(ab, "allow_dup", NULL);
+
+			if ((entry->func != CRITICAL_DATA) ||
+			    (entry->action != MEASURE)) {
+				result = -EINVAL;
+				break;
+			}
+
+			entry->allow_dup = true;
+
+			entry->flags |= IMA_ALLOW_DUP;
+			break;
 		case Opt_fsuuid:
 			ima_log_string(ab, "fsuuid", args[0].from);
 

@@ -1761,6 +1778,9 @@ int ima_policy_show(struct seq_file *m, void *v)
 		seq_puts(m, " ");
 	}
 
+	if (entry->flags & IMA_ALLOW_DUP)
+		seq_puts(m, "allow_dup");
+
 	if (entry->flags & IMA_PCR) {
 		snprintf(tbuf, sizeof(tbuf), "%d", entry->pcr);
 		seq_printf(m, pt(Opt_pcr), tbuf);

-- 
2.17.1

Hi Tushar,

On Fri, 2021-01-29 at 16:45 -0800, Tushar Sugandhi wrote:

IMA does not measure duplicate buffer data since TPM extend is a very
expensive operation.  However, in some cases for integrity critical
data, the measurement of duplicate data is necessary to accurately
determine the current state of the system.  Eg, SELinux state changing
from 'audit', to 'enforcing', and back to 'audit' again.  In this
example, currently, IMA will not measure the last state change to
'audit'.  This limits the ability of attestation services to accurately
determine the current state of the integrity critical data on the
system.

This series addresses this gap by providing the ability to measure
duplicate entries for integrity critical data, driven by policy.

The same reason for re-measuring buffer data is equally applicable to
files.  In both cases, the file or the buffer isn't re-measured if it
already exists in the htable.   Please don't limit this patch set to
just buffer data.

thanks,

Mimi

Hi Tushar,

On Fri, 2021-01-29 at 16:45 -0800, Tushar Sugandhi wrote:

 	struct tpm_digestate_entry(struct ima_template_entry *entry, int violation,

Can't this change be simplified to "if (!violation && !allow_dup)"?

Also perhaps instead of passing another variable "allow_dup" to each of
these functions, pass a mask containing violation and allow_dup.

 			audit_cause = "hash_exists";
 			result = -EEXIST;
 			goto out;

thanks,

Mimi

Hi Tushar,

On Fri, 2021-01-29 at 16:45 -0800, Tushar Sugandhi wrote:

IMA needs to support duplicate measurements of integrity
critical data to accurately determine the current state of that data
on the system.  Further, since measurement of duplicate data is not
required for all the use cases, it needs to be policy driven.

Define "allow_dup", a new IMA policy condition, for the IMA func
CRITICAL_DATA to allow duplicate buffer measurement of integrity
critical data.

Limit the ability to measure duplicate buffer data when action is
"measure" and func is CRITICAL_DATA.

Why?!

Defining a new boolean entry shouldn't be necessary.    The other
boolean values are just stored in "flags".

 	struct ima_template_desc *template;
 };

thanks,

Mimi

Hi Tushar,


On Mon, 2021-02-08 at 15:22 -0500, Mimi Zohar wrote:

On Fri, 2021-01-29 at 16:45 -0800, Tushar Sugandhi wrote:

quoted

IMA does not measure duplicate buffer data since TPM extend is a very
expensive operation.  However, in some cases for integrity critical
data, the measurement of duplicate data is necessary to accurately
determine the current state of the system.  Eg, SELinux state changing
from 'audit', to 'enforcing', and back to 'audit' again.  In this
example, currently, IMA will not measure the last state change to
'audit'.  This limits the ability of attestation services to accurately
determine the current state of the integrity critical data on the
system.

This series addresses this gap by providing the ability to measure
duplicate entries for integrity critical data, driven by policy.

The same reason for re-measuring buffer data is equally applicable to
files.  In both cases, the file or the buffer isn't re-measured if it
already exists in the htable.   Please don't limit this patch set to
just buffer data.

Instead of making the change on a per measurement rule basis, disabling
"htable" would be the simplest way of forcing re-measurements.  All
that would be needed is a new Kconfig (e.g. CONFIG_IMA_DISABLE_HTABLE)
and the associated test in ima_add_template_entry().

thanks,

Mimi

Thank you Mimi for reviewing this series.

On 2021-02-08 1:10 p.m., Mimi Zohar wrote:

Hi Tushar,


On Mon, 2021-02-08 at 15:22 -0500, Mimi Zohar wrote:

quoted

On Fri, 2021-01-29 at 16:45 -0800, Tushar Sugandhi wrote:

quoted

IMA does not measure duplicate buffer data since TPM extend is a very
expensive operation.  However, in some cases for integrity critical
data, the measurement of duplicate data is necessary to accurately
determine the current state of the system.  Eg, SELinux state changing
from 'audit', to 'enforcing', and back to 'audit' again.  In this
example, currently, IMA will not measure the last state change to
'audit'.  This limits the ability of attestation services to accurately
determine the current state of the integrity critical data on the
system.

This series addresses this gap by providing the ability to measure
duplicate entries for integrity critical data, driven by policy.

The same reason for re-measuring buffer data is equally applicable to
files.  In both cases, the file or the buffer isn't re-measured if it
already exists in the htable.   Please don't limit this patch set to
just buffer data.

Agreed.  I wasn't sure if you wanted the support for files, or other buffer measurement scenarios, except critical data.  So I started the implementation with supporting just critical data.  Happy to extend it to files and other buffer measurement scenarios as you suggested.

Instead of making the change on a per measurement rule basis, disabling
"htable" would be the simplest way of forcing re-measurements.  All
that would be needed is a new Kconfig (e.g. CONFIG_IMA_DISABLE_HTABLE)
and the associated test in ima_add_template_entry().

Agreed.  Earlier I wasn't sure if you wanted allow_dup support for all the scenarios.  Now that it is clear,  I will implement it as you suggested.  Thank you so much for the pointers.  Appreciate it.

Thanks,
Tushar

thanks,

Mimi

On 2021-02-08 12:45 p.m., Mimi Zohar wrote:

Hi Tushar,

On Fri, 2021-01-29 at 16:45 -0800, Tushar Sugandhi wrote:

quoted

IMA needs to support duplicate measurements of integrity
critical data to accurately determine the current state of that data
on the system.  Further, since measurement of duplicate data is not
required for all the use cases, it needs to be policy driven.

Define "allow_dup", a new IMA policy condition, for the IMA func
CRITICAL_DATA to allow duplicate buffer measurement of integrity
critical data.

Limit the ability to measure duplicate buffer data when action is
"measure" and func is CRITICAL_DATA.

Why?!

I wasn't sure if it would break any use-case by supporting this for all the files / buffers.  That's why I only wanted to address the scenario that we discussed in the last series (critical data measurement).
But as you suggested in this series' cover letter response, I am happy to extend it to other scenarios (by disabling "htable" using new Kconfig (e.g. CONFIG_IMA_DISABLE_HTABLE)

quoted

Signed-off-by: Tushar Sugandhi <redacted>
---

diff --git a/security/integrity/ima/ima_policy.c b/security/integrity/ima/ima_policy.c
index 9b45d064a87d..b89eb768dd05 100644
--- a/security/integrity/ima/ima_policy.c
+++ b/security/integrity/ima/ima_policy.c

@@ -35,6 +35,7 @@
  #define IMA_FSNAME	0x0200
  #define IMA_KEYRINGS	0x0400
  #define IMA_LABEL	0x0800
+#define IMA_ALLOW_DUP	0x1000
  
  #define UNKNOWN		0
  #define MEASURE		0x0001	/* same as IMA_MEASURE */

@@ -87,6 +88,7 @@ struct ima_rule_entry {
  	char *fsname;
  	struct ima_rule_opt_list *keyrings; /* Measure keys added to these keyrings */
  	struct ima_rule_opt_list *label; /* Measure data grouped under this label */

Defining a new boolean entry shouldn't be necessary.    The other
boolean values are just stored in "flags".

Thanks.  Will do the same here.
Thanks,
Tushar

quoted

  	struct ima_template_desc *template;
  };

thanks,

Mimi

On Tue, 2021-02-09 at 10:23 -0800, Tushar Sugandhi wrote:

quoted

On Mon, 2021-02-08 at 15:22 -0500, Mimi Zohar wrote:

quoted

On Fri, 2021-01-29 at 16:45 -0800, Tushar Sugandhi wrote:

quoted

IMA does not measure duplicate buffer data since TPM extend is a very
expensive operation.  However, in some cases for integrity critical
data, the measurement of duplicate data is necessary to accurately
determine the current state of the system.  Eg, SELinux state changing
from 'audit', to 'enforcing', and back to 'audit' again.  In this
example, currently, IMA will not measure the last state change to
'audit'.  This limits the ability of attestation services to accurately
determine the current state of the integrity critical data on the
system.

This series addresses this gap by providing the ability to measure
duplicate entries for integrity critical data, driven by policy.

The same reason for re-measuring buffer data is equally applicable to
files.  In both cases, the file or the buffer isn't re-measured if it
already exists in the htable.   Please don't limit this patch set to
just buffer data.

Agreed.  I wasn't sure if you wanted the support for files, or other 
buffer measurement scenarios, except critical data.  So I started the 
implementation with supporting just critical data.  Happy to extend it 
to files and other buffer measurement scenarios as you suggested.

quoted

Instead of making the change on a per measurement rule basis, disabling
"htable" would be the simplest way of forcing re-measurements.  All
that would be needed is a new Kconfig (e.g. CONFIG_IMA_DISABLE_HTABLE)
and the associated test in ima_add_template_entry().

Agreed.  Earlier I wasn't sure if you wanted allow_dup support for all 
the scenarios.  Now that it is clear,  I will implement it as you 
suggested.  Thank you so much for the pointers.  Appreciate it.

There are two different solutions - per measurement rule, disabling
htable - being discussed.   Disabling htable requires miminumal
changes.  Which version are you thinking of implementing?

thanks,

Mimi

On 2021-02-08 12:24 p.m., Mimi Zohar wrote:

Hi Tushar,

On Fri, 2021-01-29 at 16:45 -0800, Tushar Sugandhi wrote:

quoted

diff --git a/security/integrity/ima/ima_queue.c b/security/integrity/ima/ima_queue.c

index c096ef8945c7..fbf359495fa8 100644
--- a/security/integrity/ima/ima_queue.c
+++ b/security/integrity/ima/ima_queue.c

@@ -158,7 +158,7 @@ static int ima_pcr_extend(struct tpm_digest *digests_arg, int pcr)
   */
  int ima_add_template_entry(struct ima_template_entry *entry, int violation,
  			   const char *op, struct inode *inode,
-			   const unsigned char *filename)
+			   const unsigned char *filename, bool allow_dup)
  {
  	u8 *digest = entry->digests[ima_hash_algo_idx].digest;

  	struct tpm_digestate_entry(struct ima_template_entry *entry, int violation,

Not sure I understand this.  Maybe a typo?  Could you please explain?

quoted

  
  	mutex_lock(&ima_extend_list_mutex);
  	if (!violation) {
-		if (ima_lookup_digest_entry(digest, entry->pcr)) {
+		if (!allow_dup &&
+		    ima_lookup_digest_entry(digest, entry->pcr)) {

Can't this change be simplified to "if (!violation && !allow_dup)"?

Sure.  Will do.

Earlier I wasn't sure if 'violation' would touch any other use-cases inadvertently.  That's why I localized the change as above.

But now since we are supporting other scenarios as well,
I believe "if (!violation && !allow_dup)" would be cleaner.

Also perhaps instead of passing another variable "allow_dup" to each of
these functions, pass a mask containing violation and allow_dup.

There were examples of both approaches in ima_match_policy().
 - int *pcr/ima_template_desc **template_desc as an out-param;
 - and various actions as flags in return int.

Earlier I couldn't decide one way or the other, so I picked the out-param approach.

But since allow_dup is just a single bit info, returning it as a bit in the action flag is a cleaner solution.
Will implement it with flag in the next iteration.

Thanks again for reviewing the series.  Really appreciate it.

Thanks,
Tushar

quoted

  			audit_cause = "hash_exists";
  			result = -EEXIST;
  			goto out;

thanks,

Mimi

On 2021-02-09 10:53 a.m., Mimi Zohar wrote:

On Tue, 2021-02-09 at 10:23 -0800, Tushar Sugandhi wrote:

quoted

quoted

On Mon, 2021-02-08 at 15:22 -0500, Mimi Zohar wrote:

quoted

On Fri, 2021-01-29 at 16:45 -0800, Tushar Sugandhi wrote:

quoted

IMA does not measure duplicate buffer data since TPM extend is a very
expensive operation.  However, in some cases for integrity critical
data, the measurement of duplicate data is necessary to accurately
determine the current state of the system.  Eg, SELinux state changing
from 'audit', to 'enforcing', and back to 'audit' again.  In this
example, currently, IMA will not measure the last state change to
'audit'.  This limits the ability of attestation services to accurately
determine the current state of the integrity critical data on the
system.

This series addresses this gap by providing the ability to measure
duplicate entries for integrity critical data, driven by policy.

The same reason for re-measuring buffer data is equally applicable to
files.  In both cases, the file or the buffer isn't re-measured if it
already exists in the htable.   Please don't limit this patch set to
just buffer data.

Agreed.  I wasn't sure if you wanted the support for files, or other
buffer measurement scenarios, except critical data.  So I started the
implementation with supporting just critical data.  Happy to extend it
to files and other buffer measurement scenarios as you suggested.

quoted

Instead of making the change on a per measurement rule basis, disabling
"htable" would be the simplest way of forcing re-measurements.  All
that would be needed is a new Kconfig (e.g. CONFIG_IMA_DISABLE_HTABLE)
and the associated test in ima_add_template_entry().

Agreed.  Earlier I wasn't sure if you wanted allow_dup support for all
the scenarios.  Now that it is clear,  I will implement it as you
suggested.  Thank you so much for the pointers.  Appreciate it.

There are two different solutions - per measurement rule, disabling
htable - being discussed.   Disabling htable requires miminumal
changes.  Which version are you thinking of implementing?

I am thinking of implementing "disabling 'htable' using a new Kconfig (e.g. CONFIG_IMA_DISABLE_HTABLE)".  That is, not using the var ima_htable or ima_lookup_digest_entry() if that CONFIG is set.
So the duplicate measurements are allowed when the CONFIG is set.
This would cover all the measurement scenarios through a single CONFIG setting.

I am not planning to implement it as a "per measurement rule".

Sorry it wasn't clear in my earlier response.

Thanks,
Tushar

thanks,

Mimi