To prevent timing attacks, HMAC value comparison needs to be constant time. Replace the memcmp() with the correct function, crypto_memneq().
Fixes: d00a1c72f7f4 ("keys: add new trusted key-type") Cc: stable@vger.kernel.org Signed-off-by: Eric Biggers ebiggers@kernel.org --- security/keys/trusted-keys/trusted_tpm1.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/security/keys/trusted-keys/trusted_tpm1.c b/security/keys/trusted-keys/trusted_tpm1.c index 89c9798d18007..e73f2c6c817a0 100644 --- a/security/keys/trusted-keys/trusted_tpm1.c +++ b/security/keys/trusted-keys/trusted_tpm1.c @@ -5,10 +5,11 @@ * * See Documentation/security/keys/trusted-encrypted.rst */
#include <crypto/hash_info.h> +#include <crypto/utils.h> #include <linux/init.h> #include <linux/slab.h> #include <linux/parser.h> #include <linux/string.h> #include <linux/err.h> @@ -239,11 +240,11 @@ int TSS_checkhmac1(unsigned char *buffer, TPM_NONCE_SIZE, enonce, TPM_NONCE_SIZE, ononce, 1, continueflag, 0, 0); if (ret < 0) goto out;
- if (memcmp(testhmac, authdata, SHA1_DIGEST_SIZE)) + if (crypto_memneq(testhmac, authdata, SHA1_DIGEST_SIZE)) ret = -EINVAL; out: kfree_sensitive(sdesc); return ret; } @@ -332,20 +333,20 @@ static int TSS_checkhmac2(unsigned char *buffer, ret = TSS_rawhmac(testhmac1, key1, keylen1, SHA1_DIGEST_SIZE, paramdigest, TPM_NONCE_SIZE, enonce1, TPM_NONCE_SIZE, ononce, 1, continueflag1, 0, 0); if (ret < 0) goto out; - if (memcmp(testhmac1, authdata1, SHA1_DIGEST_SIZE)) { + if (crypto_memneq(testhmac1, authdata1, SHA1_DIGEST_SIZE)) { ret = -EINVAL; goto out; } ret = TSS_rawhmac(testhmac2, key2, keylen2, SHA1_DIGEST_SIZE, paramdigest, TPM_NONCE_SIZE, enonce2, TPM_NONCE_SIZE, ononce, 1, continueflag2, 0, 0); if (ret < 0) goto out; - if (memcmp(testhmac2, authdata2, SHA1_DIGEST_SIZE)) + if (crypto_memneq(testhmac2, authdata2, SHA1_DIGEST_SIZE)) ret = -EINVAL; out: kfree_sensitive(sdesc); return ret; }
base-commit: d6084bb815c453de27af8071a23163a711586a6c
On Thu, Jul 31, 2025 at 02:23:52PM -0700, Eric Biggers wrote:
To prevent timing attacks, HMAC value comparison needs to be constant time. Replace the memcmp() with the correct function, crypto_memneq().
Fixes: d00a1c72f7f4 ("keys: add new trusted key-type") Cc: stable@vger.kernel.org Signed-off-by: Eric Biggers ebiggers@kernel.org
Was crypto_memneq() available at the time?
security/keys/trusted-keys/trusted_tpm1.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/security/keys/trusted-keys/trusted_tpm1.c b/security/keys/trusted-keys/trusted_tpm1.c index 89c9798d18007..e73f2c6c817a0 100644 --- a/security/keys/trusted-keys/trusted_tpm1.c +++ b/security/keys/trusted-keys/trusted_tpm1.c @@ -5,10 +5,11 @@
- See Documentation/security/keys/trusted-encrypted.rst
*/ #include <crypto/hash_info.h> +#include <crypto/utils.h> #include <linux/init.h> #include <linux/slab.h> #include <linux/parser.h> #include <linux/string.h> #include <linux/err.h> @@ -239,11 +240,11 @@ int TSS_checkhmac1(unsigned char *buffer, TPM_NONCE_SIZE, enonce, TPM_NONCE_SIZE, ononce, 1, continueflag, 0, 0); if (ret < 0) goto out;
- if (memcmp(testhmac, authdata, SHA1_DIGEST_SIZE))
- if (crypto_memneq(testhmac, authdata, SHA1_DIGEST_SIZE)) ret = -EINVAL;
out: kfree_sensitive(sdesc); return ret; } @@ -332,20 +333,20 @@ static int TSS_checkhmac2(unsigned char *buffer, ret = TSS_rawhmac(testhmac1, key1, keylen1, SHA1_DIGEST_SIZE, paramdigest, TPM_NONCE_SIZE, enonce1, TPM_NONCE_SIZE, ononce, 1, continueflag1, 0, 0); if (ret < 0) goto out;
- if (memcmp(testhmac1, authdata1, SHA1_DIGEST_SIZE)) {
- if (crypto_memneq(testhmac1, authdata1, SHA1_DIGEST_SIZE)) { ret = -EINVAL; goto out; } ret = TSS_rawhmac(testhmac2, key2, keylen2, SHA1_DIGEST_SIZE, paramdigest, TPM_NONCE_SIZE, enonce2, TPM_NONCE_SIZE, ononce, 1, continueflag2, 0, 0); if (ret < 0) goto out;
- if (memcmp(testhmac2, authdata2, SHA1_DIGEST_SIZE))
- if (crypto_memneq(testhmac2, authdata2, SHA1_DIGEST_SIZE)) ret = -EINVAL;
out: kfree_sensitive(sdesc); return ret; }
base-commit: d6084bb815c453de27af8071a23163a711586a6c
2.50.1
BR, Jarkko
On Tue, Aug 05, 2025 at 04:44:27PM +0300, Jarkko Sakkinen wrote:
On Thu, Jul 31, 2025 at 02:23:52PM -0700, Eric Biggers wrote:
To prevent timing attacks, HMAC value comparison needs to be constant time. Replace the memcmp() with the correct function, crypto_memneq().
Fixes: d00a1c72f7f4 ("keys: add new trusted key-type") Cc: stable@vger.kernel.org Signed-off-by: Eric Biggers ebiggers@kernel.org
Was crypto_memneq() available at the time?
No. The Fixes commit is still correct, though, as it's the commit that introduced the memcmp(). Technically it was still a bug at that time, even if there wasn't a helper function available yet.
- Eric
On Tue, Aug 05, 2025 at 10:32:27AM -0700, Eric Biggers wrote:
On Tue, Aug 05, 2025 at 04:44:27PM +0300, Jarkko Sakkinen wrote:
On Thu, Jul 31, 2025 at 02:23:52PM -0700, Eric Biggers wrote:
To prevent timing attacks, HMAC value comparison needs to be constant time. Replace the memcmp() with the correct function, crypto_memneq().
Fixes: d00a1c72f7f4 ("keys: add new trusted key-type") Cc: stable@vger.kernel.org Signed-off-by: Eric Biggers ebiggers@kernel.org
Was crypto_memneq() available at the time?
No. The Fixes commit is still correct, though, as it's the commit that introduced the memcmp(). Technically it was still a bug at that time, even if there wasn't a helper function available yet.
Add a remark to the commit message.
- Eric
BR, Jarkko
On Sat, Aug 09, 2025 at 01:37:49PM +0300, Jarkko Sakkinen wrote:
On Tue, Aug 05, 2025 at 10:32:27AM -0700, Eric Biggers wrote:
On Tue, Aug 05, 2025 at 04:44:27PM +0300, Jarkko Sakkinen wrote:
On Thu, Jul 31, 2025 at 02:23:52PM -0700, Eric Biggers wrote:
To prevent timing attacks, HMAC value comparison needs to be constant time. Replace the memcmp() with the correct function, crypto_memneq().
Fixes: d00a1c72f7f4 ("keys: add new trusted key-type") Cc: stable@vger.kernel.org Signed-off-by: Eric Biggers ebiggers@kernel.org
Was crypto_memneq() available at the time?
No. The Fixes commit is still correct, though, as it's the commit that introduced the memcmp(). Technically it was still a bug at that time, even if there wasn't a helper function available yet.
Add a remark to the commit message.
I don't know what the point is (both commits are over a decade old), but sure I did that in v2.
- Eric
linux-stable-mirror@lists.linaro.org
-
Eric Biggers -
Jarkko Sakkinen