[PATCH v6 0/3] ima-evm-utils: Add --keyid option

[PATCH v6 0/3] ima-evm-utils: Add --keyid option

[Vitaly Chikunov]

Allow user to set signature's keyid using `--keyid' option. Keyid should
correspond to SKID in certificate. When keyid is calculated using SHA-1
in libimaevm it may mismatch keyid extracted by the kernel from SKID of
certificate (the way public key is presented to the kernel), thus making
signatures not verifiable. This may happen when certificate is using non
SHA-1 SKID (see rfc7093) or just 'unique number' (see rfc5280 4.2.1.2).
As a last resort user may specify arbitrary keyid using the new option.
Certificate @filename could be used instead of the hex number. And,
third option is to read keyid from the cert appended to the key file.

These commits create backward incompatible ABI change for libimaevm,
 thus soname should be incremented on release.

Changes from v5:
- ima_read_keyid renamed to imaevm_read_keyid and its return value
  inverted to match other API functions. Suggested by Stefan Berger
- Rebased over next-testing. Noticed by Mimi Zohar.

Changes from v4:
- ima_read_keyid() API changed. As suggested by Stefan Berger.

Changes from v3:
- ima_read_keyid() is improved to better support both use cases.

Changes from v2:
- Add ima_read_keyid() function to libemaevm and use it in both evmctl
  (for --keyid) and sign_hash_v2() (for concatenated PEMs). Suggested
  by Stefan Berger.
- Autodetect PEM by openssl reading it instead of magic string.
  Suggested by Stefan Berger.
- Trivial change: s/unsigned long int/unsigned long/ for keyid type.

Changes from v1:
- Extract keyid from cert associated to key file.
- Use sizeof instead of constant.

Changes since rfc version:
- `imaevm_params.keyid' now stored as native integer (instead of network
  order). Suggested by Stefan Berger.
- Added support for `--keyid=@filename'.

Vitaly Chikunov (3):
  ima-evm-utils: Allow manual setting keyid for signing
  ima-evm-utils: Allow manual setting keyid from a cert file
  ima-evm-utils: Read keyid from the cert appended to the key file

 README                 |  5 +++
 src/evmctl.c           | 27 ++++++++++++++
 src/imaevm.h           |  2 +
 src/libimaevm.c        | 99 ++++++++++++++++++++++++++++++++++++++++++++++++--
 tests/sign_verify.test |  2 +
 5 files changed, 132 insertions(+), 3 deletions(-)

-- 
2.11.0

[PATCH v6 1/3] ima-evm-utils: Allow manual setting keyid for signing

[Vitaly Chikunov]

Allow user to set signature's keyid using `--keyid' option. Keyid should
correspond to SKID in certificate, when keyid is calculated using SHA-1
in libimaevm it may mismatch keyid extracted by the kernel from SKID of
certificate (the way public key is presented to the kernel), thus making
signatures not verifiable. This may happen when certificate is using non
SHA-1 SKID (see rfc7093) or just 'unique number' (see rfc5280 4.2.1.2).
As a last resort user may specify arbitrary keyid using the new option.

This commit creates ABI change for libimaevm, because of adding
additional parameter to imaevm_params - newer libimaevm cannot work
with older clients.

Signed-off-by: Vitaly Chikunov <vt@altlinux.org>
Reported-by: Elvira Khabirova <lineprinter0@gmail.com>
Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>
---
 README                 |  1 +
 src/evmctl.c           | 15 +++++++++++++++
 src/imaevm.h           |  1 +
 src/libimaevm.c        |  8 +++++---
 tests/sign_verify.test |  1 +
 5 files changed, 23 insertions(+), 3 deletions(-)

diff --git a/README b/README
index 321045d..8cd66e0 100644
--- a/README
+++ b/README
@@ -48,6 +48,7 @@ OPTIONS
       --xattr-user   store xattrs in user namespace (for testing purposes)
       --rsa          use RSA key type and signing scheme v1
   -k, --key          path to signing key (default: /etc/keys/{privkey,pubkey}_evm.pem)
+      --keyid val    overwrite signature keyid with a value (for signing)
   -o, --portable     generate portable EVM signatures
   -p, --pass         password for encrypted signing key
   -r, --recursive    recurse into directories (sign)
diff --git a/src/evmctl.c b/src/evmctl.c
index 7a6f202..d853d34 100644
--- a/src/evmctl.c
+++ b/src/evmctl.c
@@ -2514,6 +2514,7 @@ static void usage(void)
 		"      --xattr-user   store xattrs in user namespace (for testing purposes)\n"
 		"      --rsa          use RSA key type and signing scheme v1\n"
 		"  -k, --key          path to signing key (default: /etc/keys/{privkey,pubkey}_evm.pem)\n"
+		"      --keyid val    overwrite signature keyid with a value (for signing)\n"
 		"  -o, --portable     generate portable EVM signatures\n"
 		"  -p, --pass         password for encrypted signing key\n"
 		"  -r, --recursive    recurse into directories (sign)\n"
@@ -2594,6 +2595,7 @@ static struct option opts[] = {
 	{"ignore-violations", 0, 0, 141},
 	{"pcrs", 1, 0, 142},
 	{"verify-bank", 2, 0, 143},
+	{"keyid", 1, 0, 144},
 	{}
 
 };
@@ -2638,6 +2640,8 @@ int main(int argc, char *argv[])
 {
 	int err = 0, c, lind;
 	ENGINE *eng = NULL;
+	unsigned long keyid;
+	char *eptr;
 
 #if !(OPENSSL_VERSION_NUMBER < 0x10100000)
 	OPENSSL_init_crypto(
@@ -2785,6 +2789,17 @@ int main(int argc, char *argv[])
 		case 143:
 			verify_bank = optarg;
 			break;
+		case 144:
+			errno = 0;
+			keyid = strtoul(optarg, &eptr, 16);
+			if (errno || eptr - optarg != strlen(optarg) ||
+			    keyid == ULONG_MAX || keyid > UINT_MAX ||
+			    keyid == 0) {
+				log_err("Invalid keyid value.\n");
+				exit(1);
+			}
+			imaevm_params.keyid = keyid;
+			break;
 		case '?':
 			exit(1);
 			break;
diff --git a/src/imaevm.h b/src/imaevm.h
index 4503919..9f38059 100644
--- a/src/imaevm.h
+++ b/src/imaevm.h
@@ -196,6 +196,7 @@ struct libimaevm_params {
 	const char *hash_algo;
 	const char *keyfile;
 	const char *keypass;
+	uint32_t keyid;		/* keyid overriding value, unless 0. */
 };
 
 struct RSA_ASN1_template {
diff --git a/src/libimaevm.c b/src/libimaevm.c
index fa6c278..481d29d 100644
--- a/src/libimaevm.c
+++ b/src/libimaevm.c
@@ -45,6 +45,7 @@
 #include <sys/param.h>
 #include <sys/stat.h>
 #include <asm/byteorder.h>
+#include <arpa/inet.h>
 #include <unistd.h>
 #include <dirent.h>
 #include <string.h>
@@ -894,7 +895,6 @@ static int sign_hash_v2(const char *algo, const unsigned char *hash,
 	const EVP_MD *md;
 	size_t sigsize;
 	const char *st;
-	uint32_t keyid;
 
 	if (!hash) {
 		log_err("sign_hash_v2: hash is null\n");
@@ -932,8 +932,10 @@ static int sign_hash_v2(const char *algo, const unsigned char *hash,
 		return -1;
 	}
 
-	calc_keyid_v2(&keyid, name, pkey);
-	hdr->keyid = keyid;
+	if (imaevm_params.keyid)
+		hdr->keyid = htonl(imaevm_params.keyid);
+	else
+		calc_keyid_v2(&hdr->keyid, name, pkey);
 
 	st = "EVP_PKEY_CTX_new";
 	if (!(ctx = EVP_PKEY_CTX_new(pkey, NULL)))
diff --git a/tests/sign_verify.test b/tests/sign_verify.test
index 3d7aa51..eccf5fa 100755
--- a/tests/sign_verify.test
+++ b/tests/sign_verify.test
@@ -365,6 +365,7 @@ sign_verify  rsa1024  sha256  0x0301 --rsa
 sign_verify  rsa1024  md5     0x030201:K:0080
 sign_verify  rsa1024  sha1    0x030202:K:0080
 sign_verify  rsa1024  sha224  0x030207:K:0080
+expect_pass check_sign TYPE=ima KEY=rsa1024 ALG=sha256 PREFIX=0x030204aabbccdd0080 OPTS=--keyid=aabbccdd
 sign_verify  rsa1024  sha256  0x030204:K:0080
   try_different_keys
   try_different_sigs
-- 
2.11.0

[PATCH v6 2/3] ima-evm-utils: Allow manual setting keyid from a cert file

[Vitaly Chikunov]

Allow user to specify `--keyid @/path/to/cert.pem' to extract keyid from
SKID of the certificate file. PEM or DER format is auto-detected.

`--keyid' option is reused instead of adding a new option (like possible
`--cert') to signify to the user it's only keyid extraction and nothing
more.

This commit creates ABI change for libimaevm, due to adding new function
ima_read_keyid(). Newer clients cannot work with older libimaevm.
Together with previous commit it creates backward-incompatible ABI
change, thus soname should be incremented on release.

Signed-off-by: Vitaly Chikunov <vt@altlinux.org>
---
 README                 |  1 +
 src/evmctl.c           | 22 ++++++++++---
 src/imaevm.h           |  1 +
 src/libimaevm.c        | 89 ++++++++++++++++++++++++++++++++++++++++++++++++++
 tests/sign_verify.test |  1 +
 5 files changed, 109 insertions(+), 5 deletions(-)

diff --git a/README b/README
index 8cd66e0..0e1f6ba 100644
--- a/README
+++ b/README
@@ -49,6 +49,7 @@ OPTIONS
       --rsa          use RSA key type and signing scheme v1
   -k, --key          path to signing key (default: /etc/keys/{privkey,pubkey}_evm.pem)
       --keyid val    overwrite signature keyid with a value (for signing)
+                     val is a x509 cert file if prefixed with '@'
   -o, --portable     generate portable EVM signatures
   -p, --pass         password for encrypted signing key
   -r, --recursive    recurse into directories (sign)
diff --git a/src/evmctl.c b/src/evmctl.c
index d853d34..9d2bdc3 100644
--- a/src/evmctl.c
+++ b/src/evmctl.c
@@ -42,6 +42,7 @@
 #include <sys/param.h>
 #include <sys/stat.h>
 #include <sys/ioctl.h>
+#include <arpa/inet.h>
 #include <fcntl.h>
 #include <unistd.h>
 #include <stdlib.h>
@@ -57,12 +58,14 @@
 #include <termios.h>
 #include <assert.h>
 
+#include <openssl/asn1.h>
 #include <openssl/sha.h>
 #include <openssl/pem.h>
 #include <openssl/hmac.h>
 #include <openssl/err.h>
 #include <openssl/rsa.h>
 #include <openssl/engine.h>
+#include <openssl/x509v3.h>
 #include "hash_info.h"
 #include "pcr.h"
 #include "utils.h"
@@ -2515,6 +2518,7 @@ static void usage(void)
 		"      --rsa          use RSA key type and signing scheme v1\n"
 		"  -k, --key          path to signing key (default: /etc/keys/{privkey,pubkey}_evm.pem)\n"
 		"      --keyid val    overwrite signature keyid with a value (for signing)\n"
+		"                     val is a x509 cert file if prefixed with '@'\n"
 		"  -o, --portable     generate portable EVM signatures\n"
 		"  -p, --pass         password for encrypted signing key\n"
 		"  -r, --recursive    recurse into directories (sign)\n"
@@ -2641,7 +2645,6 @@ int main(int argc, char *argv[])
 	int err = 0, c, lind;
 	ENGINE *eng = NULL;
 	unsigned long keyid;
-	char *eptr;
 
 #if !(OPENSSL_VERSION_NUMBER < 0x10100000)
 	OPENSSL_init_crypto(
@@ -2790,10 +2793,19 @@ int main(int argc, char *argv[])
 			verify_bank = optarg;
 			break;
 		case 144:
-			errno = 0;
-			keyid = strtoul(optarg, &eptr, 16);
-			if (errno || eptr - optarg != strlen(optarg) ||
-			    keyid == ULONG_MAX || keyid > UINT_MAX ||
+			if (optarg[0] == '@') {
+				keyid = imaevm_read_keyid(optarg + 1);
+			} else {
+				char *eptr;
+
+				errno = 0;
+				keyid = strtoul(optarg, &eptr, 16);
+				if (eptr - optarg != strlen(optarg) || errno) {
+					log_err("Invalid keyid value.\n");
+					exit(1);
+				}
+			}
+			if (keyid == ULONG_MAX || keyid > UINT_MAX ||
 			    keyid == 0) {
 				log_err("Invalid keyid value.\n");
 				exit(1);
diff --git a/src/imaevm.h b/src/imaevm.h
index 9f38059..f427e8e 100644
--- a/src/imaevm.h
+++ b/src/imaevm.h
@@ -219,6 +219,7 @@ EVP_PKEY *read_pub_pkey(const char *keyfile, int x509);
 void calc_keyid_v1(uint8_t *keyid, char *str, const unsigned char *pkey, int len);
 void calc_keyid_v2(uint32_t *keyid, char *str, EVP_PKEY *pkey);
 int key2bin(RSA *key, unsigned char *pub);
+uint32_t imaevm_read_keyid(const char *certfile);
 
 int sign_hash(const char *algo, const unsigned char *hash, int size, const char *keyfile, const char *keypass, unsigned char *sig);
 int verify_hash(const char *file, const unsigned char *hash, int size, unsigned char *sig, int siglen);
diff --git a/src/libimaevm.c b/src/libimaevm.c
index 481d29d..1c03768 100644
--- a/src/libimaevm.c
+++ b/src/libimaevm.c
@@ -57,6 +57,7 @@
 #include <openssl/pem.h>
 #include <openssl/evp.h>
 #include <openssl/x509.h>
+#include <openssl/x509v3.h>
 #include <openssl/err.h>
 
 #include "imaevm.h"
@@ -748,6 +749,94 @@ void calc_keyid_v2(uint32_t *keyid, char *str, EVP_PKEY *pkey)
 	X509_PUBKEY_free(pk);
 }
 
+enum keyid_file_type {
+	KEYID_FILE_PEM_KEY = 0,
+	KEYID_FILE_UNK_CERT,
+};
+
+/*
+ * @keyid:	Pointer to 32-bit value in network order, can be unaligned.
+ * @file_type:	KEYID_FILE_PEM_KEY - PEM encoded private key file with possible
+ *		appended x509 cert;
+ *		KEYID_FILE_UNK_CERT - x509 cert file (maybe PEM, maybe DER
+ *		encoded).
+ * Return:	0 on success.
+ */
+static int __read_keyid(uint32_t *keyid, const char *certfile,
+			enum keyid_file_type file_type)
+{
+	uint32_t keyid_raw;
+	const ASN1_OCTET_STRING *skid;
+	int skid_len;
+	X509 *x = NULL;
+	FILE *fp;
+
+	if (!(fp = fopen(certfile, "r"))) {
+		log_err("read keyid: %s: Cannot open: %s\n", certfile,
+			strerror(errno));
+		return -1;
+	}
+	if (!PEM_read_X509(fp, &x, NULL, NULL)) {
+		if (ERR_GET_REASON(ERR_peek_last_error()) == PEM_R_NO_START_LINE) {
+			ERR_clear_error();
+			if (file_type == KEYID_FILE_PEM_KEY) {
+				log_debug("%s: x509 certificate not found\n",
+					  certfile);
+				fclose(fp);
+				return -1;
+			}
+			rewind(fp);
+			d2i_X509_fp(fp, &x);
+		}
+		if (!x) {
+			ERR_print_errors_fp(stderr);
+			log_err("read keyid: %s: Error reading x509 certificate\n",
+				certfile);
+			fclose(fp);
+			return -1;
+		}
+	}
+	fclose(fp);
+
+	if (!(skid = X509_get0_subject_key_id(x))) {
+		log_err("read keyid: %s: SKID not found\n", certfile);
+		goto err_free;
+	}
+	skid_len = ASN1_STRING_length(skid);
+	if (skid_len < sizeof(keyid_raw)) {
+		log_err("read keyid: %s: SKID too short (len %d)\n", certfile,
+			skid_len);
+		goto err_free;
+	}
+	memcpy(&keyid_raw, ASN1_STRING_get0_data(skid) + skid_len
+	       - sizeof(keyid_raw), sizeof(keyid_raw));
+	log_info("keyid %04x (from %s)\n", ntohl(keyid_raw), certfile);
+	memcpy(keyid, &keyid_raw, sizeof(*keyid));
+	return 0;
+
+err_free:
+	X509_free(x);
+	return -1;
+}
+
+/**
+ * imaevm_read_keyid() - Read 32-bit keyid from the cert file.
+ * @certfile:	File possibly containing certificate in DER/PEM format.
+ *
+ * Try to read keyid from Subject Key Identifier (SKID) of certificate.
+ * Autodetect if cert is in PEM or DER encoding.
+ *
+ * Return: 0 on error;
+ *         32-bit keyid as unsigned integer in host order.
+ */
+uint32_t imaevm_read_keyid(const char *certfile)
+{
+	uint32_t keyid_raw = 0;
+
+	__read_keyid(&keyid_raw, certfile, KEYID_FILE_UNK_CERT);
+	return ntohl(keyid_raw);
+}
+
 static EVP_PKEY *read_priv_pkey(const char *keyfile, const char *keypass)
 {
 	FILE *fp;
diff --git a/tests/sign_verify.test b/tests/sign_verify.test
index eccf5fa..7f6d421 100755
--- a/tests/sign_verify.test
+++ b/tests/sign_verify.test
@@ -366,6 +366,7 @@ sign_verify  rsa1024  md5     0x030201:K:0080
 sign_verify  rsa1024  sha1    0x030202:K:0080
 sign_verify  rsa1024  sha224  0x030207:K:0080
 expect_pass check_sign TYPE=ima KEY=rsa1024 ALG=sha256 PREFIX=0x030204aabbccdd0080 OPTS=--keyid=aabbccdd
+expect_pass check_sign TYPE=ima KEY=rsa1024 ALG=sha256 PREFIX=0x030204:K:0080 OPTS=--keyid=@test-rsa1024.cer
 sign_verify  rsa1024  sha256  0x030204:K:0080
   try_different_keys
   try_different_sigs
-- 
2.11.0

[PATCH v6 3/3] ima-evm-utils: Read keyid from the cert appended to the key file

[Vitaly Chikunov]

Allow to have certificate appended to the private key of `--key'
specified (PEM) file (for v2 signing) to facilitate reading of keyid
from the associated cert. This will allow users to have private and
public key as a single file. There is no check that public key form the
cert matches associated private key.

Signed-off-by: Vitaly Chikunov <vt@altlinux.org>
---
 README          | 3 +++
 src/libimaevm.c | 8 +++++---
 2 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/README b/README
index 0e1f6ba..ea11bde 100644
--- a/README
+++ b/README
@@ -127,6 +127,9 @@ for signing and importing the key.
 Second key format uses X509 DER encoded public key certificates and uses asymmetric key support
 in the kernel (since kernel 3.9). CONFIG_INTEGRITY_ASYMMETRIC_KEYS must be enabled (default).
 
+For v2 signatures x509 certificate with the public key could be appended to the private
+key (both are in PEM format) to properly determine its Subject Key Identifier (SKID).
+
 
 Integrity keyrings
 ----------------
diff --git a/src/libimaevm.c b/src/libimaevm.c
index 1c03768..bfce7ef 100644
--- a/src/libimaevm.c
+++ b/src/libimaevm.c
@@ -1021,10 +1021,12 @@ static int sign_hash_v2(const char *algo, const unsigned char *hash,
 		return -1;
 	}
 
-	if (imaevm_params.keyid)
+	if (imaevm_params.keyid) {
 		hdr->keyid = htonl(imaevm_params.keyid);
-	else
-		calc_keyid_v2(&hdr->keyid, name, pkey);
+	} else {
+		if (__read_keyid(&hdr->keyid, keyfile, KEYID_FILE_PEM_KEY))
+			calc_keyid_v2(&hdr->keyid, name, pkey);
+	}
 
 	st = "EVP_PKEY_CTX_new";
 	if (!(ctx = EVP_PKEY_CTX_new(pkey, NULL)))
-- 
2.11.0

Re: [PATCH v6 1/3] ima-evm-utils: Allow manual setting keyid for signing

[Mimi Zohar]

On Tue, 2021-05-11 at 14:56 +0300, Vitaly Chikunov wrote:
> Allow user to set signature's keyid using `--keyid' option. Keyid should
> correspond to SKID in certificate, when keyid is calculated using SHA-1
> in libimaevm it may mismatch keyid extracted by the kernel from SKID of
> certificate (the way public key is presented to the kernel), thus making
> signatures not verifiable. This may happen when certificate is using non
> SHA-1 SKID (see rfc7093) or just 'unique number' (see rfc5280 4.2.1.2).
> As a last resort user may specify arbitrary keyid using the new option.
> 
> This commit creates ABI change for libimaevm, because of adding
> additional parameter to imaevm_params - newer libimaevm cannot work
> with older clients.

Thank you for highlighting the ABI change.

> 
> Signed-off-by: Vitaly Chikunov <vt@altlinux.org>
> Reported-by: Elvira Khabirova <lineprinter0@gmail.com>
> Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>
> ---
>  README                 |  1 +
>  src/evmctl.c           | 15 +++++++++++++++
>  src/imaevm.h           |  1 +
>  src/libimaevm.c        |  8 +++++---
>  tests/sign_verify.test |  1 +
>  5 files changed, 23 insertions(+), 3 deletions(-)
> 
> diff --git a/README b/README
> index 321045d..8cd66e0 100644
> --- a/README
> +++ b/README
> @@ -48,6 +48,7 @@ OPTIONS
>        --xattr-user   store xattrs in user namespace (for testing purposes)
>        --rsa          use RSA key type and signing scheme v1
>    -k, --key          path to signing key (default: /etc/keys/{privkey,pubkey}_evm.pem)
> +      --keyid val    overwrite signature keyid with a value (for signing)
>    -o, --portable     generate portable EVM signatures
>    -p, --pass         password for encrypted signing key
>    -r, --recursive    recurse into directories (sign)
> diff --git a/src/evmctl.c b/src/evmctl.c
> index 7a6f202..d853d34 100644
> --- a/src/evmctl.c
> +++ b/src/evmctl.c
> @@ -2514,6 +2514,7 @@ static void usage(void)
>  		"      --xattr-user   store xattrs in user namespace (for testing purposes)\n"
>  		"      --rsa          use RSA key type and signing scheme v1\n"
>  		"  -k, --key          path to signing key (default: /etc/keys/{privkey,pubkey}_evm.pem)\n"
> +		"      --keyid val    overwrite signature keyid with a value (for signing)\n"
>  		"  -o, --portable     generate portable EVM signatures\n"
>  		"  -p, --pass         password for encrypted signing key\n"
>  		"  -r, --recursive    recurse into directories (sign)\n"
> @@ -2594,6 +2595,7 @@ static struct option opts[] = {
>  	{"ignore-violations", 0, 0, 141},
>  	{"pcrs", 1, 0, 142},
>  	{"verify-bank", 2, 0, 143},
> +	{"keyid", 1, 0, 144},
>  	{}
> 
>  };
> @@ -2638,6 +2640,8 @@ int main(int argc, char *argv[])
>  {
>  	int err = 0, c, lind;
>  	ENGINE *eng = NULL;
> +	unsigned long keyid;
> +	char *eptr;
> 
>  #if !(OPENSSL_VERSION_NUMBER < 0x10100000)
>  	OPENSSL_init_crypto(
> @@ -2785,6 +2789,17 @@ int main(int argc, char *argv[])
>  		case 143:
>  			verify_bank = optarg;
>  			break;
> +		case 144:
> +			errno = 0;
> +			keyid = strtoul(optarg, &eptr, 16);
> +			if (errno || eptr - optarg != strlen(optarg) ||

Before using eptr, shouldn't there be a null test?  It's unclear what
verifying the resulting string length against the user provided length
provides.   Requiring the optarg value to be prefixed with "0x" would
simplify the strlen test.  The length would be 2 * sizeof uint + 2.  
(The subsequent patch wouldn't need a contrived prefix.)

> +			    keyid == ULONG_MAX || keyid > UINT_MAX ||

Why are both of these tests required?  When ULONG_MAX and UINT_MAX
aren't the same, wouldn't ULONG_MAX always be bigger than UINT_MAX?

> +			    keyid == 0) {
> +				log_err("Invalid keyid value.\n");
> +				exit(1);
> +			}
> +			imaevm_params.keyid = keyid;
> +			break;
>  		case '?':
>  			exit(1);
>  			break;
> diff --git a/src/imaevm.h b/src/imaevm.h
> index 4503919..9f38059 100644
> --- a/src/imaevm.h
> +++ b/src/imaevm.h
> @@ -196,6 +196,7 @@ struct libimaevm_params {
>  	const char *hash_algo;
>  	const char *keyfile;
>  	const char *keypass;
> +	uint32_t keyid;		/* keyid overriding value, unless 0. */

If keyid is suppose to be big endiran format, please annotate it in the
comment.

>  };
> 
>  struct RSA_ASN1_template {
> diff --git a/src/libimaevm.c b/src/libimaevm.c
> index fa6c278..481d29d 100644
> --- a/src/libimaevm.c
> +++ b/src/libimaevm.c
> @@ -45,6 +45,7 @@
>  #include <sys/param.h>
>  #include <sys/stat.h>
>  #include <asm/byteorder.h>
> +#include <arpa/inet.h>
>  #include <unistd.h>
>  #include <dirent.h>
>  #include <string.h>
> @@ -894,7 +895,6 @@ static int sign_hash_v2(const char *algo, const unsigned char *hash,
>  	const EVP_MD *md;
>  	size_t sigsize;
>  	const char *st;
> -	uint32_t keyid;
> 
>  	if (!hash) {
>  		log_err("sign_hash_v2: hash is null\n");
> @@ -932,8 +932,10 @@ static int sign_hash_v2(const char *algo, const unsigned char *hash,
>  		return -1;
>  	}
> 
> -	calc_keyid_v2(&keyid, name, pkey);
> -	hdr->keyid = keyid;
> +	if (imaevm_params.keyid)
> +		hdr->keyid = htonl(imaevm_params.keyid);
> +	else
> +		calc_keyid_v2(&hdr->keyid, name, pkey);
> 
At what point should imaevm_params.keyid be converted to big endian -
on usage or setting it?

Mimi

>  	st = "EVP_PKEY_CTX_new";
>  	if (!(ctx = EVP_PKEY_CTX_new(pkey, NULL)))
> diff --git a/tests/sign_verify.test b/tests/sign_verify.test
> index 3d7aa51..eccf5fa 100755
> --- a/tests/sign_verify.test
> +++ b/tests/sign_verify.test
> @@ -365,6 +365,7 @@ sign_verify  rsa1024  sha256  0x0301 --rsa
>  sign_verify  rsa1024  md5     0x030201:K:0080
>  sign_verify  rsa1024  sha1    0x030202:K:0080
>  sign_verify  rsa1024  sha224  0x030207:K:0080
> +expect_pass check_sign TYPE=ima KEY=rsa1024 ALG=sha256 PREFIX=0x030204aabbccdd0080 OPTS=--keyid=aabbccdd
>  sign_verify  rsa1024  sha256  0x030204:K:0080
>    try_different_keys
>    try_different_sigs

Re: [PATCH v6 2/3] ima-evm-utils: Allow manual setting keyid from a cert file

[Mimi Zohar]

On Tue, 2021-05-11 at 14:56 +0300, Vitaly Chikunov wrote:
> 
> +/**
> + * imaevm_read_keyid() - Read 32-bit keyid from the cert file.
> + * @certfile:	File possibly containing certificate in DER/PEM format.
> + *
> + * Try to read keyid from Subject Key Identifier (SKID) of certificate.
> + * Autodetect if cert is in PEM or DER encoding.
> + *
> + * Return: 0 on error;
> + *         32-bit keyid as unsigned integer in host order.
> + */
> +uint32_t imaevm_read_keyid(const char *certfile)
> +{
> +	uint32_t keyid_raw = 0;
> +
> +	__read_keyid(&keyid_raw, certfile, KEYID_FILE_UNK_CERT);
> +	return ntohl(keyid_raw);

On error, __read_keyid() returns < 0, but the return value isn't
checked before calling ntohl.  In either case, is imaevm_params.keyid
suppose to be big endian byte order?  If so, the call to htonl in
sign_hash_v2() would be unnecessary.

thanks,

Mimi

Re: [PATCH v6 3/3] ima-evm-utils: Read keyid from the cert appended to the key file

[Mimi Zohar]

On Tue, 2021-05-11 at 14:56 +0300, Vitaly Chikunov wrote:
> Allow to have certificate appended to the private key of `--key'
> specified (PEM) file (for v2 signing) to facilitate reading of keyid
> from the associated cert. This will allow users to have private and
> public key as a single file. There is no check that public key form the
> cert matches associated private key.

Is this a standard formats for storing the public and private key in
the same file?

Mimi

Re: [PATCH v6 3/3] ima-evm-utils: Read keyid from the cert appended to the key file

[Vitaly Chikunov]

Mimi,

On Fri, Jun 25, 2021 at 08:22:15AM -0400, Mimi Zohar wrote:
> On Tue, 2021-05-11 at 14:56 +0300, Vitaly Chikunov wrote:
> > Allow to have certificate appended to the private key of `--key'
> > specified (PEM) file (for v2 signing) to facilitate reading of keyid
> > from the associated cert. This will allow users to have private and
> > public key as a single file. There is no check that public key form the
> > cert matches associated private key.
> 
> Is this a standard formats for storing the public and private key in
> the same file?

I am not aware of any standard to keep ASCII armored text x509 cert
together with the private key in the same file. But, it's common usage
for some web servers, such as NGINX. People commonly suggest doing
it in that context:

  https://stackoverflow.com/questions/991758/how-to-get-pem-file-from-key-and-crt-files
  "cat server.crt server.key > server.includesprivatekey.pem"
  "cat server.crt server.key > server.pem"

Thanks,


> 
> Mimi

Re: [PATCH v6 2/3] ima-evm-utils: Allow manual setting keyid from a cert file

[Vitaly Chikunov]

On Fri, Jun 25, 2021 at 08:22:06AM -0400, Mimi Zohar wrote:
> On Tue, 2021-05-11 at 14:56 +0300, Vitaly Chikunov wrote:
> > 
> > +/**
> > + * imaevm_read_keyid() - Read 32-bit keyid from the cert file.
> > + * @certfile:	File possibly containing certificate in DER/PEM format.
> > + *
> > + * Try to read keyid from Subject Key Identifier (SKID) of certificate.
> > + * Autodetect if cert is in PEM or DER encoding.
> > + *
> > + * Return: 0 on error;
> > + *         32-bit keyid as unsigned integer in host order.
> > + */
> > +uint32_t imaevm_read_keyid(const char *certfile)
> > +{
> > +	uint32_t keyid_raw = 0;
> > +
> > +	__read_keyid(&keyid_raw, certfile, KEYID_FILE_UNK_CERT);
> > +	return ntohl(keyid_raw);
> 
> On error, __read_keyid() returns < 0, but the return value isn't
> checked before calling ntohl.

It does not need to be checked, because it also does not change value
pointed by `&keyid_raw' (which is 0 by default). Yes, this is not
mentioned in the `__read_keyid' description. But, otherwise (to change
pointed value on error) would be a bug.

> In either case, is imaevm_params.keyid
> suppose to be big endian byte order? 

No, it's in native order. This looked reasonable, since it could be
presented to the user as a hex number.

Thanks,

> If so, the call to htonl in
> sign_hash_v2() would be unnecessary.
> 
> thanks,
> 
> Mimi

Re: [PATCH v6 1/3] ima-evm-utils: Allow manual setting keyid for signing

[Vitaly Chikunov]

Mimi,

On Fri, Jun 25, 2021 at 08:21:46AM -0400, Mimi Zohar wrote:
> On Tue, 2021-05-11 at 14:56 +0300, Vitaly Chikunov wrote:
> > Allow user to set signature's keyid using `--keyid' option. Keyid should
> > correspond to SKID in certificate, when keyid is calculated using SHA-1
> > in libimaevm it may mismatch keyid extracted by the kernel from SKID of
> > certificate (the way public key is presented to the kernel), thus making
> > signatures not verifiable. This may happen when certificate is using non
> > SHA-1 SKID (see rfc7093) or just 'unique number' (see rfc5280 4.2.1.2).
> > As a last resort user may specify arbitrary keyid using the new option.
> > 
> > This commit creates ABI change for libimaevm, because of adding
> > additional parameter to imaevm_params - newer libimaevm cannot work
> > with older clients.
> 
> Thank you for highlighting the ABI change.
> 
> > 
> > Signed-off-by: Vitaly Chikunov <vt@altlinux.org>
> > Reported-by: Elvira Khabirova <lineprinter0@gmail.com>
> > Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>
> > ---
> >  README                 |  1 +
> >  src/evmctl.c           | 15 +++++++++++++++
> >  src/imaevm.h           |  1 +
> >  src/libimaevm.c        |  8 +++++---
> >  tests/sign_verify.test |  1 +
> >  5 files changed, 23 insertions(+), 3 deletions(-)
> > 
> > diff --git a/README b/README
> > index 321045d..8cd66e0 100644
> > --- a/README
> > +++ b/README
> > @@ -48,6 +48,7 @@ OPTIONS
> >        --xattr-user   store xattrs in user namespace (for testing purposes)
> >        --rsa          use RSA key type and signing scheme v1
> >    -k, --key          path to signing key (default: /etc/keys/{privkey,pubkey}_evm.pem)
> > +      --keyid val    overwrite signature keyid with a value (for signing)
> >    -o, --portable     generate portable EVM signatures
> >    -p, --pass         password for encrypted signing key
> >    -r, --recursive    recurse into directories (sign)
> > diff --git a/src/evmctl.c b/src/evmctl.c
> > index 7a6f202..d853d34 100644
> > --- a/src/evmctl.c
> > +++ b/src/evmctl.c
> > @@ -2514,6 +2514,7 @@ static void usage(void)
> >  		"      --xattr-user   store xattrs in user namespace (for testing purposes)\n"
> >  		"      --rsa          use RSA key type and signing scheme v1\n"
> >  		"  -k, --key          path to signing key (default: /etc/keys/{privkey,pubkey}_evm.pem)\n"
> > +		"      --keyid val    overwrite signature keyid with a value (for signing)\n"
> >  		"  -o, --portable     generate portable EVM signatures\n"
> >  		"  -p, --pass         password for encrypted signing key\n"
> >  		"  -r, --recursive    recurse into directories (sign)\n"
> > @@ -2594,6 +2595,7 @@ static struct option opts[] = {
> >  	{"ignore-violations", 0, 0, 141},
> >  	{"pcrs", 1, 0, 142},
> >  	{"verify-bank", 2, 0, 143},
> > +	{"keyid", 1, 0, 144},
> >  	{}
> > 
> >  };
> > @@ -2638,6 +2640,8 @@ int main(int argc, char *argv[])
> >  {
> >  	int err = 0, c, lind;
> >  	ENGINE *eng = NULL;
> > +	unsigned long keyid;
> > +	char *eptr;
> > 
> >  #if !(OPENSSL_VERSION_NUMBER < 0x10100000)
> >  	OPENSSL_init_crypto(
> > @@ -2785,6 +2789,17 @@ int main(int argc, char *argv[])
> >  		case 143:
> >  			verify_bank = optarg;
> >  			break;
> > +		case 144:
> > +			errno = 0;
> > +			keyid = strtoul(optarg, &eptr, 16);
> > +			if (errno || eptr - optarg != strlen(optarg) ||
> 
> Before using eptr, shouldn't there be a null test?

AFAIK, strtoul() should never return NULL there. strtoul(3):

  If endptr is not NULL, strtoul() stores the address of the first
  invalid character in *endptr.  If there were no digits at all,
  strtoul() stores the original value of nptr in *endptr (and returns
  0). In particular, if *nptr is not '\0' but **endptr is '\0' on
  return, the entire string is valid.

> It's unclear what
> verifying the resulting string length against the user provided length
> provides.

It shows that whole string is parsed by strtoul, otherwise there is some
unparsed garbage, which mean this was not valid number.

> Requiring the optarg value to be prefixed with "0x" would
> simplify the strlen test.  The length would be 2 * sizeof uint + 2.  
> (The subsequent patch wouldn't need a contrived prefix.)

(I do not understand this remark at the moment.)

Base 16 will let user pass keyid just as a string, copy-pasting from
somewhere else.

> 
> > +			    keyid == ULONG_MAX || keyid > UINT_MAX ||
> 
> Why are both of these tests required?  When ULONG_MAX and UINT_MAX
> aren't the same, wouldn't ULONG_MAX always be bigger than UINT_MAX?

ULONG_MAX is special return value of strtoul. UINT_MAX is a maximum
capacity of our keyid variable. So, in the case they are different there
is two tests. This also shows that both conditions are checked, without
need to solve them in the head.

> 
> > +			    keyid == 0) {
> > +				log_err("Invalid keyid value.\n");
> > +				exit(1);
> > +			}
> > +			imaevm_params.keyid = keyid;
> > +			break;
> >  		case '?':
> >  			exit(1);
> >  			break;
> > diff --git a/src/imaevm.h b/src/imaevm.h
> > index 4503919..9f38059 100644
> > --- a/src/imaevm.h
> > +++ b/src/imaevm.h
> > @@ -196,6 +196,7 @@ struct libimaevm_params {
> >  	const char *hash_algo;
> >  	const char *keyfile;
> >  	const char *keypass;
> > +	uint32_t keyid;		/* keyid overriding value, unless 0. */
> 
> If keyid is suppose to be big endiran format, please annotate it in the
> comment.

It isn't.

> 
> >  };
> > 
> >  struct RSA_ASN1_template {
> > diff --git a/src/libimaevm.c b/src/libimaevm.c
> > index fa6c278..481d29d 100644
> > --- a/src/libimaevm.c
> > +++ b/src/libimaevm.c
> > @@ -45,6 +45,7 @@
> >  #include <sys/param.h>
> >  #include <sys/stat.h>
> >  #include <asm/byteorder.h>
> > +#include <arpa/inet.h>
> >  #include <unistd.h>
> >  #include <dirent.h>
> >  #include <string.h>
> > @@ -894,7 +895,6 @@ static int sign_hash_v2(const char *algo, const unsigned char *hash,
> >  	const EVP_MD *md;
> >  	size_t sigsize;
> >  	const char *st;
> > -	uint32_t keyid;
> > 
> >  	if (!hash) {
> >  		log_err("sign_hash_v2: hash is null\n");
> > @@ -932,8 +932,10 @@ static int sign_hash_v2(const char *algo, const unsigned char *hash,
> >  		return -1;
> >  	}
> > 
> > -	calc_keyid_v2(&keyid, name, pkey);
> > -	hdr->keyid = keyid;
> > +	if (imaevm_params.keyid)
> > +		hdr->keyid = htonl(imaevm_params.keyid);
> > +	else
> > +		calc_keyid_v2(&hdr->keyid, name, pkey);
> > 
> At what point should imaevm_params.keyid be converted to big endian -
> on usage or setting it?

On usage (when storing into header).

Thanks,

> 
> Mimi
> 
> >  	st = "EVP_PKEY_CTX_new";
> >  	if (!(ctx = EVP_PKEY_CTX_new(pkey, NULL)))
> > diff --git a/tests/sign_verify.test b/tests/sign_verify.test
> > index 3d7aa51..eccf5fa 100755
> > --- a/tests/sign_verify.test
> > +++ b/tests/sign_verify.test
> > @@ -365,6 +365,7 @@ sign_verify  rsa1024  sha256  0x0301 --rsa
> >  sign_verify  rsa1024  md5     0x030201:K:0080
> >  sign_verify  rsa1024  sha1    0x030202:K:0080
> >  sign_verify  rsa1024  sha224  0x030207:K:0080
> > +expect_pass check_sign TYPE=ima KEY=rsa1024 ALG=sha256 PREFIX=0x030204aabbccdd0080 OPTS=--keyid=aabbccdd
> >  sign_verify  rsa1024  sha256  0x030204:K:0080
> >    try_different_keys
> >    try_different_sigs
> 

Re: [PATCH v6 1/3] ima-evm-utils: Allow manual setting keyid for signing

[Mimi Zohar]

Hi Vitaly,

Thank you for the detailed explanation.

On Sat, 2021-06-26 at 03:42 +0300, Vitaly Chikunov wrote:

> > Requiring the optarg value to be prefixed with "0x" would
> > simplify the strlen test.
> > (The subsequent patch wouldn't need a contrived prefix.)
> 
> (I do not understand this remark at the moment.)
> 
> Base 16 will let user pass keyid just as a string, copy-pasting from
> somewhere else.

strtoul() supports prefixing the ascii-hex string with "0x".  To
differentiate between a keyid and pathname, why not require the keyid
be prefixed with "0x", as opposed to requiring the pathname to be
prefixed with '@', like "--keyid @/path/to/cert.pem".

The new test would be "OPTS=--keyid=0xaabbccdd"

thanks,

Mimi

Re: [PATCH v6 1/3] ima-evm-utils: Allow manual setting keyid for signing

[Vitaly Chikunov]

Mimi,

On Mon, Jun 28, 2021 at 04:50:42PM -0400, Mimi Zohar wrote:
> 
> Thank you for the detailed explanation.
> 
> On Sat, 2021-06-26 at 03:42 +0300, Vitaly Chikunov wrote:
> 
> > > Requiring the optarg value to be prefixed with "0x" would
> > > simplify the strlen test.
> > > (The subsequent patch wouldn't need a contrived prefix.)
> > 
> > (I do not understand this remark at the moment.)
> > 
> > Base 16 will let user pass keyid just as a string, copy-pasting from
> > somewhere else.
> 
> strtoul() supports prefixing the ascii-hex string with "0x".  To
> differentiate between a keyid and pathname, why not require the keyid
> be prefixed with "0x", as opposed to requiring the pathname to be
> prefixed with '@', like "--keyid @/path/to/cert.pem".

I wanted to avoid (filename vs keyid) ambiguity of the argument to
`--keyid' - if user have file named "0x00112233" they would have hard
time passing it to `--keyid'. But, it's impossible to have keyid string
starting with "@". So, "@" perfectly distinguish type of `--keyid'
argument but "0x" isn't.

Also, in some software (zip, rar) "@" is common prefix meaning value
should be taken from the specified file. But, yes, "@" is not common
in Unix environments. Do you want me to create separate option like
`--keyid-from-file'?

Thanks,

> 
> The new test would be "OPTS=--keyid=0xaabbccdd"
> 
> thanks,
> 
> Mimi

Re: [PATCH v6 1/3] ima-evm-utils: Allow manual setting keyid for signing

[Mimi Zohar]

Hi Vitaly,

On Tue, 2021-06-29 at 04:32 +0300, Vitaly Chikunov wrote:
> Mimi,
> 
> On Mon, Jun 28, 2021 at 04:50:42PM -0400, Mimi Zohar wrote:
> > 
> > Thank you for the detailed explanation.
> > 
> > On Sat, 2021-06-26 at 03:42 +0300, Vitaly Chikunov wrote:
> > 
> > > > Requiring the optarg value to be prefixed with "0x" would
> > > > simplify the strlen test.
> > > > (The subsequent patch wouldn't need a contrived prefix.)
> > > 
> > > (I do not understand this remark at the moment.)
> > > 
> > > Base 16 will let user pass keyid just as a string, copy-pasting from
> > > somewhere else.
> > 
> > strtoul() supports prefixing the ascii-hex string with "0x".  To
> > differentiate between a keyid and pathname, why not require the keyid
> > be prefixed with "0x", as opposed to requiring the pathname to be
> > prefixed with '@', like "--keyid @/path/to/cert.pem".
> 
> I wanted to avoid (filename vs keyid) ambiguity of the argument to
> `--keyid' - if user have file named "0x00112233" they would have hard
> time passing it to `--keyid'. But, it's impossible to have keyid string
> starting with "@". So, "@" perfectly distinguish type of `--keyid'
> argument but "0x" isn't.
> 
> Also, in some software (zip, rar) "@" is common prefix meaning value
> should be taken from the specified file. But, yes, "@" is not common
> in Unix environments. Do you want me to create separate option like
> `--keyid-from-file'?

It's highly unlikely that both the filename and pathname would be
prefixed with "0x".  Defining a new option might be a good idea. 
Possibly naming it --extract-cert-keyid,  ---cert-keyid, or --keyid-
from-cert.  

thanks,

Mimi

Re: [PATCH v6 2/3] ima-evm-utils: Allow manual setting keyid from a cert file

[Mimi Zohar]

Hi Vitaly,

On Sat, 2021-06-26 at 03:27 +0300, Vitaly Chikunov wrote:
> > In either case, is imaevm_params.keyid
> > suppose to be big endian byte order? 
> 
> No, it's in native order. This looked reasonable, since it could be
> presented to the user as a hex number.

My confusion must have come from the __read_keyid() comment.

+ * @keyid:     Pointer to 32-bit value in network order, can be
unaligned.

Please annotate the comment and the imaevm_params.keyid appropriately.

thanks,

Mimi

Re: [PATCH v6 3/3] ima-evm-utils: Read keyid from the cert appended to the key file

[Mimi Zohar]

Hi Vitaly,

On Tue, 2021-05-11 at 14:56 +0300, Vitaly Chikunov wrote:
> 
> diff --git a/src/libimaevm.c b/src/libimaevm.c
> index 1c03768..bfce7ef 100644
> --- a/src/libimaevm.c
> +++ b/src/libimaevm.c
> @@ -1021,10 +1021,12 @@ static int sign_hash_v2(const char *algo, const unsigned char *hash,
>  		return -1;
>  	}
> 
> -	if (imaevm_params.keyid)
> +	if (imaevm_params.keyid) {
>  		hdr->keyid = htonl(imaevm_params.keyid);
> -	else
> -		calc_keyid_v2(&hdr->keyid, name, pkey);
> +	} else {
> +		if (__read_keyid(&hdr->keyid, keyfile, KEYID_FILE_PEM_KEY))
> +			calc_keyid_v2(&hdr->keyid, name, pkey);
> +	}

If the call to  __read_keyid() was independent of the test and call to
calc_keyid_v2(), it would make reading code the easier to understand.

result = __read_keyid(...);
if (result  < 0)
    calc_keyid_v2(...);

thanks,

Mimi

> 
>  	st = "EVP_PKEY_CTX_new";
>  	if (!(ctx = EVP_PKEY_CTX_new(pkey, NULL)))

Re: [PATCH v6 3/3] ima-evm-utils: Read keyid from the cert appended to the key file

[Vitaly Chikunov]

Mimi,

On Wed, Jun 30, 2021 at 01:38:02PM -0400, Mimi Zohar wrote:
> On Tue, 2021-05-11 at 14:56 +0300, Vitaly Chikunov wrote:
> > 
> > diff --git a/src/libimaevm.c b/src/libimaevm.c
> > index 1c03768..bfce7ef 100644
> > --- a/src/libimaevm.c
> > +++ b/src/libimaevm.c
> > @@ -1021,10 +1021,12 @@ static int sign_hash_v2(const char *algo, const unsigned char *hash,
> >  		return -1;
> >  	}
> > 
> > -	if (imaevm_params.keyid)
> > +	if (imaevm_params.keyid) {
> >  		hdr->keyid = htonl(imaevm_params.keyid);
> > -	else
> > -		calc_keyid_v2(&hdr->keyid, name, pkey);
> > +	} else {
> > +		if (__read_keyid(&hdr->keyid, keyfile, KEYID_FILE_PEM_KEY))
> > +			calc_keyid_v2(&hdr->keyid, name, pkey);
> > +	}
> 
> If the call to  __read_keyid() was independent of the test and call to
> calc_keyid_v2(), it would make reading code the easier to understand.
> 
> result = __read_keyid(...);
> if (result  < 0)
>     calc_keyid_v2(...);

No problem to add additional variable and check __read_keyid indirectly.
But, how is this independent? Call to calc_keyid_v2 is dependent to
__read_keyid result.

Thanks,

> 
> thanks,
> 
> Mimi
> 
> > 
> >  	st = "EVP_PKEY_CTX_new";
> >  	if (!(ctx = EVP_PKEY_CTX_new(pkey, NULL)))
> 

Re: [PATCH v6 3/3] ima-evm-utils: Read keyid from the cert appended to the key file

[Vitaly Chikunov]

Mimi,

On Wed, Jun 30, 2021 at 10:10:55PM +0300, Vitaly Chikunov wrote:
> On Wed, Jun 30, 2021 at 01:38:02PM -0400, Mimi Zohar wrote:
> > On Tue, 2021-05-11 at 14:56 +0300, Vitaly Chikunov wrote:
> > > 
> > > diff --git a/src/libimaevm.c b/src/libimaevm.c
> > > index 1c03768..bfce7ef 100644
> > > --- a/src/libimaevm.c
> > > +++ b/src/libimaevm.c
> > > @@ -1021,10 +1021,12 @@ static int sign_hash_v2(const char *algo, const unsigned char *hash,
> > >  		return -1;
> > >  	}
> > > 
> > > -	if (imaevm_params.keyid)
> > > +	if (imaevm_params.keyid) {
> > >  		hdr->keyid = htonl(imaevm_params.keyid);
> > > -	else
> > > -		calc_keyid_v2(&hdr->keyid, name, pkey);
> > > +	} else {
> > > +		if (__read_keyid(&hdr->keyid, keyfile, KEYID_FILE_PEM_KEY))
> > > +			calc_keyid_v2(&hdr->keyid, name, pkey);
> > > +	}
> > 
> > If the call to  __read_keyid() was independent of the test and call to
> > calc_keyid_v2(), it would make reading code the easier to understand.
> > 
> > result = __read_keyid(...);
> > if (result  < 0)
> >     calc_keyid_v2(...);
> 
> No problem to add additional variable and check __read_keyid indirectly.
> But, how is this independent? Call to calc_keyid_v2 is dependent to
> __read_keyid result.

I think I understood what you meant.

Thanks,

> 
> Thanks,
> 
> > 
> > thanks,
> > 
> > Mimi
> > 
> > > 
> > >  	st = "EVP_PKEY_CTX_new";
> > >  	if (!(ctx = EVP_PKEY_CTX_new(pkey, NULL)))
> > 

Re: [PATCH v6 1/3] ima-evm-utils: Allow manual setting keyid for signing

[Vitaly Chikunov]

Mimi,

On Wed, Jun 30, 2021 at 12:39:02PM -0400, Mimi Zohar wrote:
> On Tue, 2021-06-29 at 04:32 +0300, Vitaly Chikunov wrote:
> > On Mon, Jun 28, 2021 at 04:50:42PM -0400, Mimi Zohar wrote:
> > > 
> > > Thank you for the detailed explanation.
> > > 
> > > On Sat, 2021-06-26 at 03:42 +0300, Vitaly Chikunov wrote:
> > > 
> > > > > Requiring the optarg value to be prefixed with "0x" would
> > > > > simplify the strlen test.
> > > > > (The subsequent patch wouldn't need a contrived prefix.)
> > > > 
> > > > (I do not understand this remark at the moment.)
> > > > 
> > > > Base 16 will let user pass keyid just as a string, copy-pasting from
> > > > somewhere else.
> > > 
> > > strtoul() supports prefixing the ascii-hex string with "0x".  To
> > > differentiate between a keyid and pathname, why not require the keyid
> > > be prefixed with "0x", as opposed to requiring the pathname to be
> > > prefixed with '@', like "--keyid @/path/to/cert.pem".
> > 
> > I wanted to avoid (filename vs keyid) ambiguity of the argument to
> > `--keyid' - if user have file named "0x00112233" they would have hard
> > time passing it to `--keyid'. But, it's impossible to have keyid string
> > starting with "@". So, "@" perfectly distinguish type of `--keyid'
> > argument but "0x" isn't.
> > 
> > Also, in some software (zip, rar) "@" is common prefix meaning value
> > should be taken from the specified file. But, yes, "@" is not common
> > in Unix environments. Do you want me to create separate option like
> > `--keyid-from-file'?
> 
> It's highly unlikely that both the filename and pathname would be
> prefixed with "0x".  Defining a new option might be a good idea. 
> Possibly naming it --extract-cert-keyid,  ---cert-keyid, or --keyid-
> from-cert.  

It's not unlikely, because people may want to keep cert files named or
symlinked by keyid. (For example, local cert database keep cert files
symlinked like `f30dd6ad.0'.)

I will add an option, most understandable, perhaps, is `--keyid-from-cert'.

Also, I will remove base 16 from --keyid argument.

Thanks,

> 
> thanks,
> 
> Mimi

Re: [PATCH v6 1/3] ima-evm-utils: Allow manual setting keyid for signing

[Mimi Zohar]

Hi Vitaly,

On Wed, 2021-06-30 at 22:44 +0300, Vitaly Chikunov wrote:
> On Wed, Jun 30, 2021 at 12:39:02PM -0400, Mimi Zohar wrote:
> > On Tue, 2021-06-29 at 04:32 +0300, Vitaly Chikunov wrote:
> > > On Mon, Jun 28, 2021 at 04:50:42PM -0400, Mimi Zohar wrote:
> > > > 
> > > > Thank you for the detailed explanation.
> > > > 
> > > > On Sat, 2021-06-26 at 03:42 +0300, Vitaly Chikunov wrote:
> > > > 
> > > > > > Requiring the optarg value to be prefixed with "0x" would
> > > > > > simplify the strlen test.
> > > > > > (The subsequent patch wouldn't need a contrived prefix.)
> > > > > 
> > > > > (I do not understand this remark at the moment.)
> > > > > 
> > > > > Base 16 will let user pass keyid just as a string, copy-pasting from
> > > > > somewhere else.
> > > > 
> > > > strtoul() supports prefixing the ascii-hex string with "0x".  To
> > > > differentiate between a keyid and pathname, why not require the keyid
> > > > be prefixed with "0x", as opposed to requiring the pathname to be
> > > > prefixed with '@', like "--keyid @/path/to/cert.pem".
> > > 
> > > I wanted to avoid (filename vs keyid) ambiguity of the argument to
> > > `--keyid' - if user have file named "0x00112233" they would have hard
> > > time passing it to `--keyid'. But, it's impossible to have keyid string
> > > starting with "@". So, "@" perfectly distinguish type of `--keyid'
> > > argument but "0x" isn't.
> > > 
> > > Also, in some software (zip, rar) "@" is common prefix meaning value
> > > should be taken from the specified file. But, yes, "@" is not common
> > > in Unix environments. Do you want me to create separate option like
> > > `--keyid-from-file'?
> > 
> > It's highly unlikely that both the filename and pathname would be
> > prefixed with "0x".  Defining a new option might be a good idea. 
> > Possibly naming it --extract-cert-keyid,  ---cert-keyid, or --keyid-
> > from-cert.  
> 
> It's not unlikely, because people may want to keep cert files named or
> symlinked by keyid. (For example, local cert database keep cert files
> symlinked like `f30dd6ad.0'.)
> 
> I will add an option, most understandable, perhaps, is `--keyid-from-cert'.
> 
> Also, I will remove base 16 from --keyid argument.

I must be missing something.  "openssl x509", "keyctl show
%keyring:<keyring name>", and "getfattr -m security.ima -e hex --dump"  
all display the skid in hex.  Why would anyone provide the skid as
something other than in hex?

thanks,

Mimi 

Re: [PATCH v6 1/3] ima-evm-utils: Allow manual setting keyid for signing

[Vitaly Chikunov]

Mimi,

On Wed, Jun 30, 2021 at 04:47:53PM -0400, Mimi Zohar wrote:
> On Wed, 2021-06-30 at 22:44 +0300, Vitaly Chikunov wrote:
> > On Wed, Jun 30, 2021 at 12:39:02PM -0400, Mimi Zohar wrote:
> > > On Tue, 2021-06-29 at 04:32 +0300, Vitaly Chikunov wrote:
> > > > On Mon, Jun 28, 2021 at 04:50:42PM -0400, Mimi Zohar wrote:
> > > > > 
> > > > > Thank you for the detailed explanation.
> > > > > 
> > > > > On Sat, 2021-06-26 at 03:42 +0300, Vitaly Chikunov wrote:
> > > > > 
> > > > > > > Requiring the optarg value to be prefixed with "0x" would
> > > > > > > simplify the strlen test.
> > > > > > > (The subsequent patch wouldn't need a contrived prefix.)
> > > > > > 
> > > > > > (I do not understand this remark at the moment.)
> > > > > > 
> > > > > > Base 16 will let user pass keyid just as a string, copy-pasting from
> > > > > > somewhere else.
> > > > > 
> > > > > strtoul() supports prefixing the ascii-hex string with "0x".  To
> > > > > differentiate between a keyid and pathname, why not require the keyid
> > > > > be prefixed with "0x", as opposed to requiring the pathname to be
> > > > > prefixed with '@', like "--keyid @/path/to/cert.pem".
> > > > 
> > > > I wanted to avoid (filename vs keyid) ambiguity of the argument to
> > > > `--keyid' - if user have file named "0x00112233" they would have hard
> > > > time passing it to `--keyid'. But, it's impossible to have keyid string
> > > > starting with "@". So, "@" perfectly distinguish type of `--keyid'
> > > > argument but "0x" isn't.
> > > > 
> > > > Also, in some software (zip, rar) "@" is common prefix meaning value
> > > > should be taken from the specified file. But, yes, "@" is not common
> > > > in Unix environments. Do you want me to create separate option like
> > > > `--keyid-from-file'?
> > > 
> > > It's highly unlikely that both the filename and pathname would be
> > > prefixed with "0x".  Defining a new option might be a good idea. 
> > > Possibly naming it --extract-cert-keyid,  ---cert-keyid, or --keyid-
> > > from-cert.  
> > 
> > It's not unlikely, because people may want to keep cert files named or
> > symlinked by keyid. (For example, local cert database keep cert files
> > symlinked like `f30dd6ad.0'.)
> > 
> > I will add an option, most understandable, perhaps, is `--keyid-from-cert'.
> > 
> > Also, I will remove base 16 from --keyid argument.
> 
> I must be missing something.  "openssl x509", "keyctl show
> %keyring:<keyring name>", and "getfattr -m security.ima -e hex --dump"  
> all display the skid in hex.  Why would anyone provide the skid as
> something other than in hex?

Ah, I thought you want '0x' prefix mandatory. Will not remove base 16
then. Also, /proc/keys shows keyid without '0x' prefix.

Thanks,


> 
> thanks,
> 
> Mimi 
> 

Re: [PATCH v6 1/3] ima-evm-utils: Allow manual setting keyid for signing

[Mimi Zohar]

On Thu, 2021-07-01 at 00:10 +0300, Vitaly Chikunov wrote:
> Mimi,
> 
> On Wed, Jun 30, 2021 at 04:47:53PM -0400, Mimi Zohar wrote:
> > On Wed, 2021-06-30 at 22:44 +0300, Vitaly Chikunov wrote:
> > > On Wed, Jun 30, 2021 at 12:39:02PM -0400, Mimi Zohar wrote:
> > > > On Tue, 2021-06-29 at 04:32 +0300, Vitaly Chikunov wrote:
> > > > > On Mon, Jun 28, 2021 at 04:50:42PM -0400, Mimi Zohar wrote:
> > > > > > 
> > > > > > Thank you for the detailed explanation.
> > > > > > 
> > > > > > On Sat, 2021-06-26 at 03:42 +0300, Vitaly Chikunov wrote:
> > > > > > 
> > > > > > > > Requiring the optarg value to be prefixed with "0x" would
> > > > > > > > simplify the strlen test.
> > > > > > > > (The subsequent patch wouldn't need a contrived prefix.)
> > > > > > > 
> > > > > > > (I do not understand this remark at the moment.)
> > > > > > > 
> > > > > > > Base 16 will let user pass keyid just as a string, copy-pasting from
> > > > > > > somewhere else.
> > > > > > 
> > > > > > strtoul() supports prefixing the ascii-hex string with "0x".  To
> > > > > > differentiate between a keyid and pathname, why not require the keyid
> > > > > > be prefixed with "0x", as opposed to requiring the pathname to be
> > > > > > prefixed with '@', like "--keyid @/path/to/cert.pem".
> > > > > 
> > > > > I wanted to avoid (filename vs keyid) ambiguity of the argument to
> > > > > `--keyid' - if user have file named "0x00112233" they would have hard
> > > > > time passing it to `--keyid'. But, it's impossible to have keyid string
> > > > > starting with "@". So, "@" perfectly distinguish type of `--keyid'
> > > > > argument but "0x" isn't.
> > > > > 
> > > > > Also, in some software (zip, rar) "@" is common prefix meaning value
> > > > > should be taken from the specified file. But, yes, "@" is not common
> > > > > in Unix environments. Do you want me to create separate option like
> > > > > `--keyid-from-file'?
> > > > 
> > > > It's highly unlikely that both the filename and pathname would be
> > > > prefixed with "0x".  Defining a new option might be a good idea. 
> > > > Possibly naming it --extract-cert-keyid,  ---cert-keyid, or --keyid-
> > > > from-cert.  
> > > 
> > > It's not unlikely, because people may want to keep cert files named or
> > > symlinked by keyid. (For example, local cert database keep cert files
> > > symlinked like `f30dd6ad.0'.)
> > > 
> > > I will add an option, most understandable, perhaps, is `--keyid-from-cert'.
> > > 
> > > Also, I will remove base 16 from --keyid argument.
> > 
> > I must be missing something.  "openssl x509", "keyctl show
> > %keyring:<keyring name>", and "getfattr -m security.ima -e hex --dump"  
> > all display the skid in hex.  Why would anyone provide the skid as
> > something other than in hex?
> 
> Ah, I thought you want '0x' prefix mandatory. Will not remove base 16
> then. Also, /proc/keys shows keyid without '0x' prefix.

Why?  One doesn't preclude the other.   "keyid = strtoul(optarg, &eptr,
16);" recognizes an "0x" prefixed string as being ascii-hex.

Mimi