When EVM_ALLOW_METADATA_WRITES is set, EVM allows any operation on metadata. Its main purpose is to allow users to freely set metadata when they are protected by a portable signature, until the HMAC key is loaded.
However, IMA is not notified about metadata changes and, after the first successful appraisal, always allows access to the files without checking metadata again.
This patch introduces the new atomic flag EVM_RESET_STATUS in integrity_iint_cache that is set in the EVM post hooks and cleared in evm_verify_hmac(). IMA checks the new flag in process_measurement() and if it is set, it clears the appraisal flags.
Although the flag could be cleared also by evm_inode_setxattr() and evm_inode_setattr() before IMA sees it, this does not happen if EVM_ALLOW_METADATA_WRITES is set. Since the only remaining caller is evm_verifyxattr(), this ensures that IMA always sees the flag set before it is cleared.
This patch also adds a call to evm_reset_status() in evm_inode_post_setattr() so that EVM won't return the cached status the next time appraisal is performed.
Cc: stable@vger.kernel.org # 4.16.x Fixes: ae1ba1676b88e ("EVM: Allow userland to permit modification of EVM-protected metadata") Signed-off-by: Roberto Sassu roberto.sassu@huawei.com --- security/integrity/evm/evm_main.c | 17 +++++++++++++++-- security/integrity/ima/ima_main.c | 8 ++++++-- security/integrity/integrity.h | 1 + 3 files changed, 22 insertions(+), 4 deletions(-)
diff --git a/security/integrity/evm/evm_main.c b/security/integrity/evm/evm_main.c index 4e9f5e8b21d5..05be1ad3e6f3 100644 --- a/security/integrity/evm/evm_main.c +++ b/security/integrity/evm/evm_main.c @@ -221,8 +221,15 @@ static enum integrity_status evm_verify_hmac(struct dentry *dentry, evm_status = (rc == -ENODATA) ? INTEGRITY_NOXATTRS : INTEGRITY_FAIL; out: - if (iint) + if (iint) { + /* + * EVM_RESET_STATUS can be cleared only by evm_verifyxattr() + * when EVM_ALLOW_METADATA_WRITES is set. This guarantees that + * IMA sees the EVM_RESET_STATUS flag set before it is cleared. + */ + clear_bit(EVM_RESET_STATUS, &iint->atomic_flags); iint->evm_status = evm_status; + } kfree(xattr_data); return evm_status; } @@ -418,8 +425,12 @@ static void evm_reset_status(struct inode *inode) struct integrity_iint_cache *iint;
iint = integrity_iint_find(inode); - if (iint) + if (iint) { + if (evm_initialized & EVM_ALLOW_METADATA_WRITES) + set_bit(EVM_RESET_STATUS, &iint->atomic_flags); + iint->evm_status = INTEGRITY_UNKNOWN; + } }
/** @@ -513,6 +524,8 @@ void evm_inode_post_setattr(struct dentry *dentry, int ia_valid) if (!evm_key_loaded()) return;
+ evm_reset_status(dentry->d_inode); + if (ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID)) evm_update_evmxattr(dentry, NULL, NULL, 0); } diff --git a/security/integrity/ima/ima_main.c b/security/integrity/ima/ima_main.c index 8a91711ca79b..bb9976dc2b74 100644 --- a/security/integrity/ima/ima_main.c +++ b/security/integrity/ima/ima_main.c @@ -246,8 +246,12 @@ static int process_measurement(struct file *file, const struct cred *cred,
mutex_lock(&iint->mutex);
- if (test_and_clear_bit(IMA_CHANGE_ATTR, &iint->atomic_flags)) - /* reset appraisal flags if ima_inode_post_setattr was called */ + if (test_and_clear_bit(IMA_CHANGE_ATTR, &iint->atomic_flags) || + test_bit(EVM_RESET_STATUS, &iint->atomic_flags)) + /* + * Reset appraisal flags if ima_inode_post_setattr was called or + * EVM reset its status and metadata modification was enabled. + */ iint->flags &= ~(IMA_APPRAISE | IMA_APPRAISED | IMA_APPRAISE_SUBMASK | IMA_APPRAISED_SUBMASK | IMA_ACTION_FLAGS); diff --git a/security/integrity/integrity.h b/security/integrity/integrity.h index 413c803c5208..2adec51c0f6e 100644 --- a/security/integrity/integrity.h +++ b/security/integrity/integrity.h @@ -70,6 +70,7 @@ #define IMA_CHANGE_ATTR 2 #define IMA_DIGSIG 3 #define IMA_MUST_MEASURE 4 +#define EVM_RESET_STATUS 5
enum evm_ima_xattr_type { IMA_XATTR_DIGEST = 0x01,
[Cc'ing John Johansen]
Hi Roberto,
On Fri, 2020-09-04 at 11:26 +0200, Roberto Sassu wrote:
When EVM_ALLOW_METADATA_WRITES is set, EVM allows any operation on metadata. Its main purpose is to allow users to freely set metadata when they are protected by a portable signature, until the HMAC key is loaded.
However, IMA is not notified about metadata changes and, after the first successful appraisal, always allows access to the files without checking metadata again.
This patch introduces the new atomic flag EVM_RESET_STATUS in integrity_iint_cache that is set in the EVM post hooks and cleared in evm_verify_hmac(). IMA checks the new flag in process_measurement() and if it is set, it clears the appraisal flags.
Although the flag could be cleared also by evm_inode_setxattr() and evm_inode_setattr() before IMA sees it, this does not happen if EVM_ALLOW_METADATA_WRITES is set. Since the only remaining caller is evm_verifyxattr(), this ensures that IMA always sees the flag set before it is cleared.
This patch also adds a call to evm_reset_status() in evm_inode_post_setattr() so that EVM won't return the cached status the next time appraisal is performed.
Cc: stable@vger.kernel.org # 4.16.x Fixes: ae1ba1676b88e ("EVM: Allow userland to permit modification of EVM-protected metadata") Signed-off-by: Roberto Sassu roberto.sassu@huawei.com
security/integrity/evm/evm_main.c | 17 +++++++++++++++-- security/integrity/ima/ima_main.c | 8 ++++++-- security/integrity/integrity.h | 1 + 3 files changed, 22 insertions(+), 4 deletions(-)
diff --git a/security/integrity/evm/evm_main.c b/security/integrity/evm/evm_main.c index 4e9f5e8b21d5..05be1ad3e6f3 100644 --- a/security/integrity/evm/evm_main.c +++ b/security/integrity/evm/evm_main.c @@ -221,8 +221,15 @@ static enum integrity_status evm_verify_hmac(struct dentry *dentry, evm_status = (rc == -ENODATA) ? INTEGRITY_NOXATTRS : INTEGRITY_FAIL; out:
- if (iint)
- if (iint) {
/*
* EVM_RESET_STATUS can be cleared only by evm_verifyxattr()
* when EVM_ALLOW_METADATA_WRITES is set. This guarantees that
* IMA sees the EVM_RESET_STATUS flag set before it is cleared.
*/
iint->evm_status = evm_status;clear_bit(EVM_RESET_STATUS, &iint->atomic_flags);
True IMA is currently the only caller of evm_verifyxattr() in the upstreamed kernel, but it is an exported function, which may be called from elsewhere. The previous version crossed the boundary between EVM & IMA with EVM modifying the IMA flag directly. This version assumes that IMA will be the only caller. Otherwise, I like this version.
Mimi
- } kfree(xattr_data); return evm_status;
} @@ -418,8 +425,12 @@ static void evm_reset_status(struct inode *inode) struct integrity_iint_cache *iint; iint = integrity_iint_find(inode);
- if (iint)
- if (iint) {
if (evm_initialized & EVM_ALLOW_METADATA_WRITES)
set_bit(EVM_RESET_STATUS, &iint->atomic_flags);
- iint->evm_status = INTEGRITY_UNKNOWN;
- }
} /** @@ -513,6 +524,8 @@ void evm_inode_post_setattr(struct dentry *dentry, int ia_valid) if (!evm_key_loaded()) return;
- evm_reset_status(dentry->d_inode);
- if (ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID)) evm_update_evmxattr(dentry, NULL, NULL, 0);
}
From: Mimi Zohar [mailto:zohar@linux.ibm.com] Sent: Thursday, September 17, 2020 2:01 PM [Cc'ing John Johansen]
Hi Roberto,
On Fri, 2020-09-04 at 11:26 +0200, Roberto Sassu wrote:
When EVM_ALLOW_METADATA_WRITES is set, EVM allows any operation
on
metadata. Its main purpose is to allow users to freely set metadata when they are protected by a portable signature, until the HMAC key is loaded.
However, IMA is not notified about metadata changes and, after the first successful appraisal, always allows access to the files without checking metadata again.
This patch introduces the new atomic flag EVM_RESET_STATUS in integrity_iint_cache that is set in the EVM post hooks and cleared in evm_verify_hmac(). IMA checks the new flag in process_measurement()
and if
it is set, it clears the appraisal flags.
Although the flag could be cleared also by evm_inode_setxattr() and evm_inode_setattr() before IMA sees it, this does not happen if EVM_ALLOW_METADATA_WRITES is set. Since the only remaining caller is evm_verifyxattr(), this ensures that IMA always sees the flag set before it is cleared.
This patch also adds a call to evm_reset_status() in evm_inode_post_setattr() so that EVM won't return the cached status
the
next time appraisal is performed.
Cc: stable@vger.kernel.org # 4.16.x Fixes: ae1ba1676b88e ("EVM: Allow userland to permit modification of
EVM-protected metadata")
Signed-off-by: Roberto Sassu roberto.sassu@huawei.com
security/integrity/evm/evm_main.c | 17 +++++++++++++++-- security/integrity/ima/ima_main.c | 8 ++++++-- security/integrity/integrity.h | 1 + 3 files changed, 22 insertions(+), 4 deletions(-)
diff --git a/security/integrity/evm/evm_main.c
b/security/integrity/evm/evm_main.c
index 4e9f5e8b21d5..05be1ad3e6f3 100644 --- a/security/integrity/evm/evm_main.c +++ b/security/integrity/evm/evm_main.c @@ -221,8 +221,15 @@ static enum integrity_status
evm_verify_hmac(struct dentry *dentry,
evm_status = (rc == -ENODATA) ? INTEGRITY_NOXATTRS : INTEGRITY_FAIL;
out:
- if (iint)
- if (iint) {
/*
* EVM_RESET_STATUS can be cleared only by
evm_verifyxattr()
* when EVM_ALLOW_METADATA_WRITES is set. This
guarantees that
* IMA sees the EVM_RESET_STATUS flag set before it is
cleared.
*/
iint->evm_status = evm_status;clear_bit(EVM_RESET_STATUS, &iint->atomic_flags);
True IMA is currently the only caller of evm_verifyxattr() in the upstreamed kernel, but it is an exported function, which may be called from elsewhere. The previous version crossed the boundary between EVM & IMA with EVM modifying the IMA flag directly. This version assumes that IMA will be the only caller. Otherwise, I like this version.
Ok, I think it is better, as you suggested, to export a new EVM function that tells if evm_reset_status() will be executed in the EVM post hooks, and to call this function from IMA. IMA would then call ima_reset_appraise_flags() also depending on the result of the new EVM function.
ima_reset_appraise_flags() should be called in a post hook in IMA. Should I introduce it?
Thanks
Roberto
HUAWEI TECHNOLOGIES Duesseldorf GmbH, HRB 56063 Managing Director: Li Peng, Li Jian, Shi Yanli
Mimi
- } kfree(xattr_data); return evm_status;
} @@ -418,8 +425,12 @@ static void evm_reset_status(struct inode *inode) struct integrity_iint_cache *iint;
iint = integrity_iint_find(inode);
- if (iint)
- if (iint) {
if (evm_initialized & EVM_ALLOW_METADATA_WRITES)
set_bit(EVM_RESET_STATUS, &iint->atomic_flags);
- iint->evm_status = INTEGRITY_UNKNOWN;
- }
}
/** @@ -513,6 +524,8 @@ void evm_inode_post_setattr(struct dentry
*dentry, int ia_valid)
if (!evm_key_loaded()) return;
- evm_reset_status(dentry->d_inode);
- if (ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID)) evm_update_evmxattr(dentry, NULL, NULL, 0);
}
On Thu, 2020-09-17 at 17:36 +0000, Roberto Sassu wrote:
diff --git a/security/integrity/evm/evm_main.c
b/security/integrity/evm/evm_main.c
index 4e9f5e8b21d5..05be1ad3e6f3 100644 --- a/security/integrity/evm/evm_main.c +++ b/security/integrity/evm/evm_main.c @@ -221,8 +221,15 @@ static enum integrity_status
evm_verify_hmac(struct dentry *dentry,
evm_status = (rc == -ENODATA) ? INTEGRITY_NOXATTRS : INTEGRITY_FAIL;
out:
- if (iint)
- if (iint) {
/*
* EVM_RESET_STATUS can be cleared only by
evm_verifyxattr()
* when EVM_ALLOW_METADATA_WRITES is set. This
guarantees that
* IMA sees the EVM_RESET_STATUS flag set before it is
cleared.
*/
iint->evm_status = evm_status;clear_bit(EVM_RESET_STATUS, &iint->atomic_flags);
True IMA is currently the only caller of evm_verifyxattr() in the upstreamed kernel, but it is an exported function, which may be called from elsewhere. The previous version crossed the boundary between EVM & IMA with EVM modifying the IMA flag directly. This version assumes that IMA will be the only caller. Otherwise, I like this version.
Ok, I think it is better, as you suggested, to export a new EVM function that tells if evm_reset_status() will be executed in the EVM post hooks, and to call this function from IMA. IMA would then call ima_reset_appraise_flags() also depending on the result of the new EVM function.
ima_reset_appraise_flags() should be called in a post hook in IMA. Should I introduce it?
Yes, so any callers of evm_verifyxattr() will need to implement the post hook as well. As much as possible, please limit code duplication.
The last time I looked, there didn't seem to be a locking concern, but please make sure.
thanks,
Mimi
linux-stable-mirror@lists.linaro.org