ima: Support a mode to appraise signed files only
Subsystems (3)
v3 cover ima: Support a mode to appraise signed files only
Hi,
Currently ima appraises all the files as specified by the rule. So if one wants to create a system where only few executables are signed, that system will not work with IMA.
With secureboot, one needs to disable kexec so that unsigned kernels can't be booted. To avoid this problem, it was proposed that sign /sbin/kexec binary and if signatures are verified successfully, give an special capability to the /sbin/kexec process. And in secureboot mode processes with that special capability can invoke sys_kexec() system call.
So there is a need for IMA to allow appraising only signed binaries. Unsigned binaries will pass the appraisal too, but will not get the special capability. (Capability patches for that are yet to be written).
This patch series adds new option, appraise_type=optional to allow appraisal to pass even if no signatures are present on the file. If signatures are present, then it has to be valid digital signature, otherwise appraisal will fail.
v2: Changed patches based on Mimi's feedback.
v3: - Changed appraise_type=imasig_optional to appraise_type=optional
- Introduced new error codes.
Thanks
Vivek
Vivek Goyal (6):
ima: detect security xattrs not enabled
ima: Return INTEGRITY_FAIL if digital signature can't be verified ima/evm: Differentiate between ima/evm nolabel return code
ima: Introduce new integrity error code INTEGRITY_XATTR_NOTSUPP
ima: Allow appraisal of digitally signed files only
ima: With appraise_type=optional, audit log some messages as info
Documentation/ABI/testing/ima_policy | 2 +- include/linux/integrity.h | 4 +++- security/integrity/evm/evm_main.c | 2 +- security/integrity/ima/ima_appraise.c | 26 ++++++++++++++++++-------- security/integrity/ima/ima_main.c | 14 ++++++++++++-- security/integrity/ima/ima_policy.c | 2 ++ security/integrity/integrity.h | 1 + 7 files changed, 38 insertions(+), 13 deletions(-)
Signature · 1 line
1.7.7.6
PATCH v3 1/6 ima: detect security xattrs not enabled
vfs_getxattr_alloc() returns -EOPNOTSUPP if filesystem does not have security label enabled. In that case there is no point in continuing further and try to fix hashes (if ima_appraise=fix was specified) as that will fail too. Return early
--- security/integrity/ima/ima_appraise.c | 4 ++++ 1 files changed, 4 insertions(+), 0 deletions(-)
security/integrity/ima/ima_appraise.c +4 -0
@@ -134,6 +134,10 @@ int ima_appraise_measurement(int func, struct integrity_iint_cache *iint, rc =vfs_getxattr_alloc (dentry,XATTR_NAME_IMA , (char **)&xattr_value,0 ,GFP_NOFS );if (rc <=0 ) { +/* File system does not support security xattr */ +if (rc == -EOPNOTSUPP ) +return INTEGRITY_UNKNOWN ; +if (rc && rc != -ENODATA ) goto out;
Signature · 1 line
1.7.7.6
PATCH v3 2/6 ima: Return INTEGRITY_FAIL if digital signature can't be verified
In reply to Vivek GoyalDigital signature verification happens using integrity_digsig_verify(). Curently we set integrity to FAIL for all error codes except -EOPNOTSUPP. This sounds out of line.
- If appropriate kernel code is not compiled in to verify signature of a file, then prractically it is a failed signature.
- For so many other possible errors we are setting the status to fail. For example, -EINVAL, -ENOKEY, -ENOMEM, -EINVAL, -ENOTSUPP etc, it beats me that why -EOPNOTSUPP is special.
This patch should make the semantics more consistent. That is, if digital signature is present in security.ima, then any error happened during signature processing leads to status INTEGRITY_FAIL.
AFAICS, it should not have any user visible effect on existing application. In some cases we will start returning INTEGRITY_FAIL instead of INTEGRITY_UNKNOWN. And process_measurement() will deny access to file both in case of INTEGRITY_UNKNOWN and INTEGRITY_FAIL.
--- security/integrity/ima/ima_appraise.c | 4 +--- 1 files changed, 1 insertions(+), 3 deletions(-)
security/integrity/ima/ima_appraise.c +1 -3
@@ -178,9 +178,7 @@ int ima_appraise_measurement(int func, struct integrity_iint_cache *iint, xattr_value->digest, rc -1 , iint->ima_xattr.digest, IMA_DIGEST_SIZE); -if (rc == -EOPNOTSUPP ) { - status =INTEGRITY_UNKNOWN ; - }else if (rc) { +if (rc) { cause ="invalid-signature" ; status =INTEGRITY_FAIL ; }else {
Signature · 1 line
1.7.7.6
Re: [PATCH 2/6] ima: Return INTEGRITY_FAIL if digital signature can't be verified
Quoted 15 lines · Vivek Goyal
Digital signature verification happens using
integrity_digsig_verify(). Curently we set integrity to FAIL for all error codes except-EOPNOTSUPP. This sounds out of line.- If appropriate kernel code is not compiled in to verify signature of a file, then prractically it is a failed signature.
- For so many other possible errors we are setting the status to fail. For example,
-EINVAL,-ENOKEY,-ENOMEM,-EINVAL,-ENOTSUPPetc, it beats me that why-EOPNOTSUPPis special.This patch should make the semantics more consistent. That is, if digital signature is present in security.ima, then any error happened during signature processing leads to status
INTEGRITY_FAIL.
Quoted 6 lines
AFAICS, it should not have any user visible effect on existing application. In some cases we will start returning
INTEGRITY_FAILinstead ofINTEGRITY_UNKNOWN. Andprocess_measurement()will deny access to file both in case ofINTEGRITY_UNKNOWNandINTEGRITY_FAIL.Signed-off-by Vivek Goyal <[email protected]>
A number of patches in this patchset more finely differentiate return codes, which is good. I agree with you totally that there is no good reason for -EOPNOTSUPP to be handled differently. Unfortunately, the initramfs is CPIO, which doesn't support xattrs. With the proposed change and 'ima_appraise_tcb' flag enabled, we wouldn't be able to boot. I really dislike hard coding policy in the kernel.
thanks,
Mimi
Quoted 19 lines
--- security/integrity/ima/ima_appraise.c | 4 +--- 1 files changed, 1 insertions(+), 3 deletions(-)security/integrity/ima/ima_appraise.c +1
-3@@ -178,9 +178,7 @@ int ima_appraise_measurement(int func, struct integrity_iint_cache *iint, xattr_value->digest, rc -1 , iint->ima_xattr.digest, IMA_DIGEST_SIZE); -if (rc == -EOPNOTSUPP ) { - status =INTEGRITY_UNKNOWN ; - }else if (rc) { +if (rc) { cause ="invalid-signature" ; status =INTEGRITY_FAIL ; }else {
Re: [PATCH 2/6] ima: Return INTEGRITY_FAIL if digital signature can't be verified
Quoted 21 lines · Mimi Zohar
Quoted 20 lines · Vivek Goyal
Digital signature verification happens using
integrity_digsig_verify(). Curently we set integrity to FAIL for all error codes except-EOPNOTSUPP. This sounds out of line.- If appropriate kernel code is not compiled in to verify signature of a file, then prractically it is a failed signature.
- For so many other possible errors we are setting the status to fail. For example,
-EINVAL,-ENOKEY,-ENOMEM,-EINVAL,-ENOTSUPPetc, it beats me that why-EOPNOTSUPPis special.This patch should make the semantics more consistent. That is, if digital signature is present in security.ima, then any error happened during signature processing leads to status
INTEGRITY_FAIL.AFAICS, it should not have any user visible effect on existing application. In some cases we will start returning
INTEGRITY_FAILinstead ofINTEGRITY_UNKNOWN. Andprocess_measurement()will deny access to file both in case ofINTEGRITY_UNKNOWNandINTEGRITY_FAIL.
Quoted 8 lines
Quoted 1 line
Signed-off-by Vivek Goyal <[email protected]>A number of patches in this patchset more finely differentiate return codes, which is good. I agree with you totally that there is no good reason for
-EOPNOTSUPPto be handled differently. Unfortunately, the initramfs is CPIO, which doesn't support xattrs. With the proposed change and 'ima_appraise_tcb' flag enabled, we wouldn't be able to boot. I really dislike hard coding policy in the kernel.
Hi Mimi,
If there are no xattr, then we will not even hit this code. We will bail out early in vfs_getxattr_alloc().
I thought that one of the DON_APPRAISE rules will kick in for initramfs and files in initramfs will not be appraised and boot will continue.
{.action = DONT_APPRAISE,.fsmagic = TMPFS_MAGIC,.flags = IMA_FSMAGIC}, {.action = DONT_APPRAISE,.fsmagic = RAMFS_MAGIC,.flags = IMA_FSMAGIC},
Is that not the case here?
Thanks
Vivek
Re: [PATCH 2/6] ima: Return INTEGRITY_FAIL if digital signature can't be verified
Quoted 36 lines · Vivek Goyal
Quoted 21 lines · Mimi Zohar
Quoted 20 lines · Vivek Goyal
Digital signature verification happens using
integrity_digsig_verify(). Curently we set integrity to FAIL for all error codes except-EOPNOTSUPP. This sounds out of line.- If appropriate kernel code is not compiled in to verify signature of a file, then prractically it is a failed signature.
- For so many other possible errors we are setting the status to fail. For example,
-EINVAL,-ENOKEY,-ENOMEM,-EINVAL,-ENOTSUPPetc, it beats me that why-EOPNOTSUPPis special.This patch should make the semantics more consistent. That is, if digital signature is present in security.ima, then any error happened during signature processing leads to status
INTEGRITY_FAIL.AFAICS, it should not have any user visible effect on existing application. In some cases we will start returning
INTEGRITY_FAILinstead ofINTEGRITY_UNKNOWN. Andprocess_measurement()will deny access to file both in case ofINTEGRITY_UNKNOWNandINTEGRITY_FAIL.Quoted 8 lines
Quoted 1 line
Signed-off-by Vivek Goyal <[email protected]>A number of patches in this patchset more finely differentiate return codes, which is good. I agree with you totally that there is no good reason for
-EOPNOTSUPPto be handled differently. Unfortunately, the initramfs is CPIO, which doesn't support xattrs. With the proposed change and 'ima_appraise_tcb' flag enabled, we wouldn't be able to boot. I really dislike hard coding policy in the kernel.Hi Mimi,
If there are no xattr, then we will not even hit this code. We will bail out early in
vfs_getxattr_alloc().
Quoted 7 lines
I thought that one of the
DON_APPRAISErules will kick in for initramfs and files in initramfs will not be appraised and boot will continue.{.action = DONT_APPRAISE,.fsmagic = TMPFS_MAGIC,.flags =
IMA_FSMAGIC}, {.action = DONT_APPRAISE,.fsmagic = RAMFS_MAGIC,.flags =IMA_FSMAGIC},Is that not the case here?
Right, thanks for the clarification. Perhaps we could abbreviate the patch description like:
Digital signature verification happens using integrity_digsig_verify(). If a digital signature is present in security.ima, then any error, which happens during signature verification, should lead to status INTEGRITY_FAIL. In the future we might want to differentiate between persistent (eg. -ENOMEM) vs. non-persistent errors, in order to cache failures. This patch removes the unnecessary -EOPNOTSUPP test.
thanks,
Mimi
Re: [PATCH 2/6] ima: Return INTEGRITY_FAIL if digital signature can't be verified
Quoted 6 lines · Mimi Zohar
Digital signature verification happens using
integrity_digsig_verify(). If a digital signature is present in security.ima, then any error, which happens during signature verification, should lead to statusINTEGRITY_FAIL. In the future we might want to differentiate between persistent (eg.-ENOMEM) vs. non-persistent errors, in order to cache failures. This patch removes the unnecessary-EOPNOTSUPPtest.
correction, "persistent vs. non-persistent(eg. -ENOMEM)"
Mimi
Re: [PATCH 2/6] ima: Return INTEGRITY_FAIL if digital signature can't be verified
In reply to Mimi ZoharQuoted 48 lines · Mimi Zohar
Quoted 36 lines · Vivek Goyal
Quoted 21 lines · Mimi Zohar
Quoted 20 lines · Vivek Goyal
Digital signature verification happens using
integrity_digsig_verify(). Curently we set integrity to FAIL for all error codes except-EOPNOTSUPP. This sounds out of line.- If appropriate kernel code is not compiled in to verify signature of a file, then prractically it is a failed signature.
- For so many other possible errors we are setting the status to fail. For example,
-EINVAL,-ENOKEY,-ENOMEM,-EINVAL,-ENOTSUPPetc, it beats me that why-EOPNOTSUPPis special.This patch should make the semantics more consistent. That is, if digital signature is present in security.ima, then any error happened during signature processing leads to status
INTEGRITY_FAIL.AFAICS, it should not have any user visible effect on existing application. In some cases we will start returning
INTEGRITY_FAILinstead ofINTEGRITY_UNKNOWN. Andprocess_measurement()will deny access to file both in case ofINTEGRITY_UNKNOWNandINTEGRITY_FAIL.Quoted 8 lines
Quoted 1 line
Signed-off-by Vivek Goyal <[email protected]>A number of patches in this patchset more finely differentiate return codes, which is good. I agree with you totally that there is no good reason for
-EOPNOTSUPPto be handled differently. Unfortunately, the initramfs is CPIO, which doesn't support xattrs. With the proposed change and 'ima_appraise_tcb' flag enabled, we wouldn't be able to boot. I really dislike hard coding policy in the kernel.Hi Mimi,
If there are no xattr, then we will not even hit this code. We will bail out early in
vfs_getxattr_alloc().Quoted 7 lines
I thought that one of the
DON_APPRAISErules will kick in for initramfs and files in initramfs will not be appraised and boot will continue.{.action = DONT_APPRAISE,.fsmagic = TMPFS_MAGIC,.flags =
IMA_FSMAGIC}, {.action = DONT_APPRAISE,.fsmagic = RAMFS_MAGIC,.flags =IMA_FSMAGIC},Is that not the case here?
Right, thanks for the clarification. Perhaps we could abbreviate the patch description like:
Quoted 6 lines
Digital signature verification happens using
integrity_digsig_verify(). If a digital signature is present in security.ima, then any error, which happens during signature verification, should lead to statusINTEGRITY_FAIL. In the future we might want to differentiate between persistent (eg.-ENOMEM) vs. non-persistent errors, in order to cache failures. This patch removes the unnecessary-EOPNOTSUPPtest.
Sure. I will modify the changelog for future posting.
Thanks
Vivek
PATCH v3 3/6 ima/evm: Differentiate between ima/evm nolabel return code
In reply to Vivek GoyalCurrently if no IMA/EVM label was present, both return error code INTEGRITY_NOLABEL. This works fine so far as nobody cares to differentiate between two error paths.
But with appraise_type=optional, we want to allow file access if IMA security label is not present. But still disallow access if EVM security label is not present.
Hence, start using different error codes for two different paths.
--- include/linux/integrity.h | 3 ++- security/integrity/evm/evm_main.c | 2 +- security/integrity/ima/ima_appraise.c | 4 ++-- 3 files changed, 5 insertions(+), 4 deletions(-)
include/linux/integrity.h +2 -1
@@ -15,7 +15,8 @@enum integrity_status {INTEGRITY_PASS =0 ,INTEGRITY_FAIL , -INTEGRITY_NOLABEL , +INTEGRITY_IMA_NOLABEL , +INTEGRITY_EVM_NOLABEL ,INTEGRITY_NOXATTRS ,INTEGRITY_UNKNOWN , };
security/integrity/evm/evm_main.c +1 -1
@@ -111,7 +111,7 @@ static enum integrity_status evm_verify_hmac(struct dentry *dentry,else if (rc == -ENODATA ) { rc =evm_find_protected_xattrs (dentry);if (rc >0 ) - evm_status =INTEGRITY_NOLABEL ; + evm_status =INTEGRITY_EVM_NOLABEL ;else if (rc ==0 ) evm_status =INTEGRITY_NOXATTRS ;/* new file */ }
security/integrity/ima/ima_appraise.c +2 -2
@@ -143,13 +143,13 @@ int ima_appraise_measurement(int func, struct integrity_iint_cache *iint, cause ="missing-hash" ; status = - (inode->i_size ==0 ) ?INTEGRITY_PASS :INTEGRITY_NOLABEL ; + (inode->i_size ==0 ) ?INTEGRITY_PASS :INTEGRITY_IMA_NOLABEL ; goto out; } status =evm_verifyxattr (dentry,XATTR_NAME_IMA , xattr_value, rc, iint);if ((status !=INTEGRITY_PASS ) && (status !=INTEGRITY_UNKNOWN )) { -if ((status ==INTEGRITY_NOLABEL ) +if ((status ==INTEGRITY_EVM_NOLABEL ) || (status ==INTEGRITY_NOXATTRS )) cause ="missing-HMAC" ;else if (status ==INTEGRITY_FAIL )
Signature · 1 line
1.7.7.6
PATCH v3 4/6 ima: Introduce new integrity error code INTEGRITY_XATTR_NOTSUPP
In reply to Vivek GoyalCurrently file system does not support xattr or security xattr are not enabled, we return INTEGRITY_UNKNOWN. INTEGRITY_UNKNOWN is returned in so many other conditions too, evm not initialized etc.
So far nobody cared but with appraise_type=optional, I need to to differentiate between different error paths. For example, I want to allow access to file when xattr are not enabled or specific security attr is not enabled by file system. But I don't want to allow access for all cases of INTEGRITY_UNKNOWN. For example,
if vfs_getxattr_alloc() returns -ENOMEM, then also INTEGRITY_UNKNOWN will be returned and I don't think we want to allow access to file in that case.
So to differentiate the errors where we can allow acccess to file, introduce the new error code.
--- include/linux/integrity.h | 1 + security/integrity/ima/ima_appraise.c | 4 ++-- 2 files changed, 3 insertions(+), 2 deletions(-)
include/linux/integrity.h +1 -0
@@ -19,6 +19,7 @@ enum integrity_status {INTEGRITY_EVM_NOLABEL ,INTEGRITY_NOXATTRS ,INTEGRITY_UNKNOWN , + INTEGRITY_XATTR_NOTSUPP, };/* List of EVM protected security xattrs */
security/integrity/ima/ima_appraise.c +2 -2
@@ -129,14 +129,14 @@ int ima_appraise_measurement(int func, struct integrity_iint_cache *iint,if (!ima_appraise)return 0 ;if (!inode->i_op->getxattr) -return INTEGRITY_UNKNOWN ; +return INTEGRITY_XATTR_NOTSUPP ; rc =vfs_getxattr_alloc (dentry,XATTR_NAME_IMA , (char **)&xattr_value,0 ,GFP_NOFS );if (rc <=0 ) {/* File system does not support security xattr */ if (rc == -EOPNOTSUPP ) -return INTEGRITY_UNKNOWN ; +return INTEGRITY_XATTR_NOTSUPP ;if (rc && rc != -ENODATA ) goto out;
Signature · 1 line
1.7.7.6
PATCH v3 5/6 ima: Allow appraisal of digitally signed files only
In reply to Vivek GoyalCurrently ima appraises all the files as specified by the rule. So if one wants to create a system where only few executables are signed, that system will not work with IMA.
With secureboot, one needs to disable kexec so that unsigned kernels can't be booted. To avoid this problem, it was proposed that sign /sbin/kexec binary and if signatures are verified successfully, give an special capability to the /sbin/kexec process. And in secureboot mode processes with that special capability can invoke sys_kexec() system call.
So there is a need for IMA to allow appraising only signed binaries. Unsigned binaries will pass the appraisal too, but will not get the special capability. (Capability patches for that are yet to be written).
This patch adds new option, appraise_type=imasig_optional to allow appraisal to pass even if no signatures are present on the file. If signatures are present, then it has to be valid digital signature, otherwise appraisal will fail.
--- Documentation/ABI/testing/ima_policy | 2 +- security/integrity/ima/ima_main.c | 14 ++++++++++++-- security/integrity/ima/ima_policy.c | 2 ++ security/integrity/integrity.h | 1 + 4 files changed, 16 insertions(+), 3 deletions(-)
Documentation/ABI/testing/ima_policy +1 -1
@@ -30,7 +30,7 @@ Description: uid:= decimal value fowner:=decimal value lsm: are LSM specific - option: appraise_type:= [imasig] + option: appraise_type:= [imasig] | [optional] default policy: # PROC_SUPER_MAGIC
security/integrity/ima/ima_main.c +12 -2
@@ -207,8 +207,18 @@ out_digsig: rc = -EACCES ; out:mutex_unlock (&inode->i_mutex); -if ((rc && must_appraise) && (ima_appraise &IMA_APPRAISE_ENFORCE )) -return -EACCES ; +if ((rc && must_appraise) && (ima_appraise &IMA_APPRAISE_ENFORCE )) { +/* +* If IMA_APPRAISAL_OPT is set, then access is allowed +* even if hash or digital signatures are not present. +*/ +if ((iint->flags &IMA_APPRAISAL_OPT ) && + (rc ==INTEGRITY_XATTR_NOTSUPP || + rc ==INTEGRITY_IMA_NOLABEL )) +return 0 ; +else +return -EACCES ; + }return 0 ; }
security/integrity/ima/ima_policy.c +2 -0
@@ -598,6 +598,8 @@ static int ima_parse_rule(char *rule, struct ima_rule_entry *entry)ima_log_string (ab,"appraise_type" , args[0 ].from);if ((strcmp (args[0 ].from,"imasig" )) ==0 ) entry->flags |=IMA_DIGSIG_REQUIRED ; +else if ((strcmp (args[0 ].from,"optional" )) ==0 ) + entry->flags |=IMA_APPRAISAL_OPT ;else result = -EINVAL ;break ;
security/integrity/integrity.h +1 -0
@@ -29,6 +29,7 @@#define IMA_ACTION_FLAGS 0xff000000#define IMA_DIGSIG 0x01000000#define IMA_DIGSIG_REQUIRED 0x02000000 +#define IMA_APPRAISAL_OPT 0x04000000#define IMA_DO_MASK (IMA_MEASURE | IMA_APPRAISE | IMA_AUDIT | \ IMA_APPRAISE_SUBMASK)
Signature · 1 line
1.7.7.6
Re: [PATCH 5/6] ima: Allow appraisal of digitally signed files only
Quoted 52 lines · Vivek Goyal
Currently ima appraises all the files as specified by the rule. So if one wants to create a system where only few executables are signed, that system will not work with IMA.
With secureboot, one needs to disable kexec so that unsigned kernels can't be booted. To avoid this problem, it was proposed that sign
/sbin/kexecbinary and if signatures are verified successfully, give an special capability to the/sbin/kexecprocess. And in secureboot mode processes with that special capability can invokesys_kexec()system call.So there is a need for IMA to allow appraising only signed binaries. Unsigned binaries will pass the appraisal too, but will not get the special capability. (Capability patches for that are yet to be written).
This patch adds new option, appraise_type=imasig_optional to allow appraisal to pass even if no signatures are present on the file. If signatures are present, then it has to be valid digital signature, otherwise appraisal will fail.
Signed-off-by Vivek Goyal <[email protected]>--- Documentation/ABI/testing/ima_policy | 2 +- security/integrity/ima/ima_main.c | 14 ++++++++++++-- security/integrity/ima/ima_policy.c | 2 ++ security/integrity/integrity.h | 1 + 4 files changed, 16 insertions(+), 3 deletions(-)Documentation/ABI/testing/ima_policy +1
-1@@ -30,7 +30,7 @@ Description: uid:= decimal value fowner:=decimal value lsm: are LSM specific - option: appraise_type:= [imasig] + option: appraise_type:= [imasig] | [optional] default policy: # PROC_SUPER_MAGICsecurity/integrity/ima/ima_main.c +1
-2@@ -207,8 +207,18 @@ out_digsig: rc = -EACCES ; out:mutex_unlock (&inode->i_mutex); -if ((rc && must_appraise) && (ima_appraise &IMA_APPRAISE_ENFORCE )) -return -EACCES ; +if ((rc && must_appraise) && (ima_appraise &IMA_APPRAISE_ENFORCE )) {
security/integrity/ima/ima_main.c out_digsig: · line 220
Quoted 10 lines
security/integrity/ima/ima_main.c +10
-0+/* +* If IMA_APPRAISAL_OPT is set, then access is allowed +* even if hash or digital signatures are not present. +*/ +if ((iint->flags &IMA_APPRAISAL_OPT ) && + (rc ==INTEGRITY_XATTR_NOTSUPP || + rc ==INTEGRITY_IMA_NOLABEL )) +return 0 ; +else +return -EACCES ;
I think there is problem here. "appraise_type=optional" can be specified per rule/hook. So two different hooks can specify two different rules.
appraise func=MMAP_CHECK appraise_type=optional
appraise func=BPRM_CHECK
I think if a file is first mmaped(), then appraisal will take place and IMA_APPRAISAL_OPT will be set in iint->flags.
Later when BPRM_CHECK hook gets executed, and it will return success based on IMA_APPRAISAL_OPT even if there was no label. And that's not what exec() expects.
So storing IMA_APPRAISAL_OPT in iint->flags seems wrong (espectially as part of IMA_ACTION_FLAGS bits). I think only bits which are valid across all rules/hooks should be stored here.
Any property which is hook/rule specific should either not be stored or should be stroed in hook specific property area.
We don't have enough space to store more hook specific properties, so I will explore the option of passing around this flag when hook is being executed and then discard it.
Thanks
Vivek
Quoted 31 lines
security/integrity/ima/ima_main.c +1
-0+ }return 0 ; }security/integrity/ima/ima_policy.c +2
-0@@ -598,6 +598,8 @@ static int ima_parse_rule(char *rule, struct ima_rule_entry *entry)ima_log_string (ab,"appraise_type" , args[0 ].from);if ((strcmp (args[0 ].from,"imasig" )) ==0 ) entry->flags |=IMA_DIGSIG_REQUIRED ; +else if ((strcmp (args[0 ].from,"optional" )) ==0 ) + entry->flags |=IMA_APPRAISAL_OPT ;else result = -EINVAL ;break ;security/integrity/integrity.h +1
-0@@ -29,6 +29,7 @@#define IMA_ACTION_FLAGS 0xff000000#define IMA_DIGSIG 0x01000000#define IMA_DIGSIG_REQUIRED 0x02000000 +#define IMA_APPRAISAL_OPT 0x04000000#define IMA_DO_MASK (IMA_MEASURE | IMA_APPRAISE | IMA_AUDIT | \ IMA_APPRAISE_SUBMASK)Signature · 1 line
1.7.7.6
Re: [PATCH 5/6] ima: Allow appraisal of digitally signed files only
Quoted 81 lines · Vivek Goyal
Quoted 52 lines · Vivek Goyal
Currently ima appraises all the files as specified by the rule. So if one wants to create a system where only few executables are signed, that system will not work with IMA.
With secureboot, one needs to disable kexec so that unsigned kernels can't be booted. To avoid this problem, it was proposed that sign
/sbin/kexecbinary and if signatures are verified successfully, give an special capability to the/sbin/kexecprocess. And in secureboot mode processes with that special capability can invokesys_kexec()system call.So there is a need for IMA to allow appraising only signed binaries. Unsigned binaries will pass the appraisal too, but will not get the special capability. (Capability patches for that are yet to be written).
This patch adds new option, appraise_type=imasig_optional to allow appraisal to pass even if no signatures are present on the file. If signatures are present, then it has to be valid digital signature, otherwise appraisal will fail.
Signed-off-by Vivek Goyal <[email protected]>--- Documentation/ABI/testing/ima_policy | 2 +- security/integrity/ima/ima_main.c | 14 ++++++++++++-- security/integrity/ima/ima_policy.c | 2 ++ security/integrity/integrity.h | 1 + 4 files changed, 16 insertions(+), 3 deletions(-)Documentation/ABI/testing/ima_policy +1
-1@@ -30,7 +30,7 @@ Description: uid:= decimal value fowner:=decimal value lsm: are LSM specific - option: appraise_type:= [imasig] + option: appraise_type:= [imasig] | [optional] default policy: # PROC_SUPER_MAGICsecurity/integrity/ima/ima_main.c +1
-2@@ -207,8 +207,18 @@ out_digsig: rc = -EACCES ; out:mutex_unlock (&inode->i_mutex); -if ((rc && must_appraise) && (ima_appraise &IMA_APPRAISE_ENFORCE )) -return -EACCES ; +if ((rc && must_appraise) && (ima_appraise &IMA_APPRAISE_ENFORCE )) {security/integrity/ima/ima_main.c out_digsig: · line 220
Quoted 10 lines
security/integrity/ima/ima_main.c +10
-0+/* +* If IMA_APPRAISAL_OPT is set, then access is allowed +* even if hash or digital signatures are not present. +*/ +if ((iint->flags &IMA_APPRAISAL_OPT ) && + (rc ==INTEGRITY_XATTR_NOTSUPP || + rc ==INTEGRITY_IMA_NOLABEL )) +return 0 ; +else +return -EACCES ;I think there is problem here. "appraise_type=optional" can be specified per rule/hook. So two different hooks can specify two different rules.
appraise func=MMAP_CHECK appraise_type=optional
appraise func=BPRM_CHECKI think if a file is first
mmaped(), then appraisal will take place andIMA_APPRAISAL_OPTwill be set iniint->flags.Later when
BPRM_CHECKhook gets executed, and it will return success based onIMA_APPRAISAL_OPTeven if there was no label. And that's not whatexec()expects.So storing
IMA_APPRAISAL_OPTiniint->flagsseems wrong (espectially as part ofIMA_ACTION_FLAGSbits). I think only bits which are valid across all rules/hooks should be stored here.
Quoted 7 lines
Any property which is hook/rule specific should either not be stored or should be stroed in hook specific property area.
We don't have enough space to store more hook specific properties, so I will explore the option of passing around this flag when hook is being executed and then discard it.
Hi,
I think there is no need to store optional in iint->flags. Just test "action" which is returned by ima_get_action(). Then it will be rule specific...
- Dmitry
Quoted 34 lines
Thanks
VivekQuoted 31 lines
security/integrity/ima/ima_main.c +1
-0+ }return 0 ; }security/integrity/ima/ima_policy.c +2
-0@@ -598,6 +598,8 @@ static int ima_parse_rule(char *rule, struct ima_rule_entry *entry)ima_log_string (ab,"appraise_type" , args[0 ].from);if ((strcmp (args[0 ].from,"imasig" )) ==0 ) entry->flags |=IMA_DIGSIG_REQUIRED ; +else if ((strcmp (args[0 ].from,"optional" )) ==0 ) + entry->flags |=IMA_APPRAISAL_OPT ;else result = -EINVAL ;break ;security/integrity/integrity.h +1
-1@@ -29,6 +29,7 @@#define IMA_ACTION_FLAGS 0xff000000#define IMA_DIGSIG 0x01000000#define IMA_DIGSIG_REQUIRED 0x02000000 +#define IMA_APPRAISAL_OPT 0x04000000#define IMA_DO_MASK (IMA_MEASURE | IMA_APPRAISE | IMA_AUDIT | \ IMA_APPRAISE_SUBMASK) --1.7.7.6
PATCH v3 6/6 ima: With appraise_type=optional, audit log some messages as info
In reply to Vivek GoyalCurrently, if there integrity status is not INTEGRITY_PASS, it is logged in audit log (as non info). This is fine because we always deny access to file for anything other than INTEGRITY_PASS.
But with appraise_type=optional, we will allow access to file even if appraisal status is not INTEGRITY_PASS. For example, in the case of INTEGRITY_IMA_NOLABEL. And on this system we don't want to log each and every executed file which is not signed.
In a typical system we are anticipating that only 1-2 files will be signed.
So don't flood the audit logs if appraise_type=optional and no IMA label is present. These messages will still show up if somebody chooses to enable audit info messages.
--- security/integrity/ima/ima_appraise.c | 12 ++++++++++-- 1 files changed, 10 insertions(+), 2 deletions(-)
security/integrity/ima/ima_appraise.c +10 -2
@@ -124,7 +124,7 @@ int ima_appraise_measurement(int func, struct integrity_iint_cache *iint,enum integrity_status status =INTEGRITY_UNKNOWN ;const char *op ="appraise_data" ;char *cause ="unknown" ; -int rc; +int rc, audit_info =0 ;if (!ima_appraise)return 0 ;
@@ -199,8 +199,16 @@ out:if (!ima_fix_xattr (dentry, iint)) status =INTEGRITY_PASS ; } + +/* +* If appraisal is optional, and if no label is present, +* log it is info. Don't flood audit logs. +*/ +if ((iint->flags &IMA_APPRAISAL_OPT ) && + status ==INTEGRITY_IMA_NOLABEL ) + audit_info =1 ;integrity_audit_msg (AUDIT_INTEGRITY_DATA , inode, filename, - op, cause, rc,0 ); + op, cause, rc, audit_info); }else {ima_cache_flags (iint, func); }
Signature · 1 line
1.7.7.6
Quoted 3 lines · Vivek Goyal
Hi,
Currently ima appraises all the files as specified by the rule.
Currently IMA appraises files based on policy.
Quoted 3 lines
So
if one wants to create a system where only few executables are signed, that system will not work with IMA.
This statement misrepresents the IMA policy. You can definitely define a policy that only measures/appraises a few specific files. In your usecase scenario, you are not willing to rely on LSM labels. Policy rules can also be based on file owner. We could also add support for gid.
Quoted 6 lines
With secureboot, one needs to disable kexec so that unsigned kernels can't be booted. To avoid this problem, it was proposed that sign
/sbin/kexecbinary and if signatures are verified successfully, give an special capability to the/sbin/kexecprocess. And in secureboot mode processes with that special capability can invokesys_kexec()system call.
Please add here that you then rely on /sbin/kexec to verify the integrity of the kernel image.
thanks,
Mimi
Quoted 6 lines · Mimi Zohar
Quoted 3 lines · Vivek Goyal
Hi,
Currently ima appraises all the files as specified by the rule.
Currently IMA appraises files based on policy.
And policy is composed of multiple rules. Ok, will change it.
Quoted 10 lines
Quoted 4 lines
So
if one wants to create a system where only few executables are signed, that system will not work with IMA.This statement misrepresents the IMA policy. You can definitely define a policy that only measures/appraises a few specific files. In your usecase scenario, you are not willing to rely on LSM labels. Policy rules can also be based on file owner. We could also add support for gid.
Ok, will change it. How about following.
We want to create a system where only few executables are signed. This patch extends IMA policy syntax so that one can specify that signatures are optional.
Quoted 10 lines
Quoted 7 lines
With secureboot, one needs to disable kexec so that unsigned kernels can't be booted. To avoid this problem, it was proposed that sign
/sbin/kexecbinary and if signatures are verified successfully, give an special capability to the/sbin/kexecprocess. And in secureboot mode processes with that special capability can invokesys_kexec()system call.Please add here that you then rely on
/sbin/kexecto verify the integrity of the kernel image.
Ok, will do that. This is infact a grey area. Yet to be figured out how /sbin/kexec will ensure a signed kernel is being loaded.
Thanks
Vivek