Skip to contentSkip to conversation

crypto/qat: use intel-ipsec-mb for partial hash

PATCH In review Fan Zhang 6 participants 18 messages Apr 7, 2022 – May 19, 2022

Subsystems (2)
lore
Share & tools
lore

Fan Zhang

author
#0 lore raw

PATCH 1/1 crypto/qat: use intel-ipsec-mb for partial hash

Since openssl 3.0 now deprecates the low level API QAT required to perform partial hash operation when creating the session. This patch is to transfer such dependency from openssl to intel-ipsec-mb.

---
 drivers/common/qat/meson.build       |  10 +++
 drivers/crypto/qat/qat_sym_session.c | 101 +++++----------------------
 2 files changed, 28 insertions(+), 83 deletions(-)
drivers/common/qat/meson.build +10 -0
index b7027f3164..d35fc69d96 100644
@@ -35,6 +35,16 @@ if qat_crypto and not libcrypto.found()
             'missing dependency, libcrypto')
 endif
 
+
+IMB_required_ver = '1.0.0'
+libipsecmb = cc.find_library('IPSec_MB', required: false)
+if not lib.found()
+    build = false
+    reason = 'missing dependency, "libIPSec_MB"'
+else
+    ext_deps += libipsecmb
+endif
+
 # The driver should not build if both compression and crypto are disabled
 #FIXME common code depends on compression files so check only compress!
 if not qat_compress # and not qat_crypto
drivers/crypto/qat/qat_sym_session.c +18 -83

163 lines not on this page · raw message

Signature · 1 line
2.32.0
#1 lore raw

Re: [External] : [PATCH] crypto/qat: use intel-ipsec-mb for partial hash

Hi Fan,

Does it mean the intel-ipsec-mb would be prerequisite of applying QAT offloading for security application? It this is the case, as I know, the intel-ipsec-mb has no FIPS certification yet. Thus I am thinking this would impact existing QAT based security application, right?

Best Regards,
Changchun Zhang

From: Fan Zhang <[email protected]>
Date: Thursday, April 7, 2022 at 11:29 AM
To: [email protected] <[email protected]>
Cc: [email protected] <[email protected]>, [email protected] <[email protected]>, [email protected] <[email protected]>, Fan Zhang <[email protected]>
Subject: [External] : [PATCH] crypto/qat: use intel-ipsec-mb for partial hash
Since openssl 3.0 now deprecates the low level API QAT required to
perform partial hash operation when creating the session. This
patch is to transfer such dependency from openssl to intel-ipsec-mb.

---
 drivers/common/qat/meson.build       |  10 +++
 drivers/crypto/qat/qat_sym_session.c | 101 +++++----------------------
 2 files changed, 28 insertions(+), 83 deletions(-)
drivers/common/qat/meson.build +10 -0
index b7027f3164..d35fc69d96 100644
@@ -35,6 +35,16 @@ if qat_crypto and not libcrypto.found()
             'missing dependency, libcrypto')
 endif
 
+
+IMB_required_ver = '1.0.0'
+libipsecmb = cc.find_library('IPSec_MB', required: false)
+if not lib.found()
+    build = false
+    reason = 'missing dependency, "libIPSec_MB"'
+else
+    ext_deps += libipsecmb
+endif
+
 # The driver should not build if both compression and crypto are disabled
 #FIXME common code depends on compression files so check only compress!
 if not qat_compress # and not qat_crypto
drivers/crypto/qat/qat_sym_session.c +18 -83

163 lines not on this page · raw message

--
2.32.0

#2 lore raw

Re: [External] : [PATCH] crypto/qat: use intel-ipsec-mb for partial hash

Hello,

Can I know the status of this patch, and the possible impact on any existing applications because the partial hash is switched from OpenSSL to intel-ipsec-mb which is not under FIPS certification?

Thanks!
Changchun

From: Changchun Zhang <[email protected]>
Date: Thursday, April 14, 2022 at 2:46 PM
To: Fan Zhang <[email protected]>, [email protected] <[email protected]>
Cc: [email protected] <[email protected]>, [email protected] <[email protected]>, [email protected] <[email protected]>, Fan Zhang <[email protected]>
Subject: Re: [External] : [PATCH] crypto/qat: use intel-ipsec-mb for partial hash
Hi Fan,

Does it mean the intel-ipsec-mb would be prerequisite of applying QAT offloading for security application? It this is the case, as I know, the intel-ipsec-mb has no FIPS certification yet. Thus I am thinking this would impact existing QAT based security application, right?

Best Regards,
Changchun Zhang

From: Fan Zhang <[email protected]>
Date: Thursday, April 7, 2022 at 11:29 AM
To: [email protected] <[email protected]>
Cc: [email protected] <[email protected]>, [email protected] <[email protected]>, [email protected] <[email protected]>, Fan Zhang <[email protected]>
Subject: [External] : [PATCH] crypto/qat: use intel-ipsec-mb for partial hash
Since openssl 3.0 now deprecates the low level API QAT required to
perform partial hash operation when creating the session. This
patch is to transfer such dependency from openssl to intel-ipsec-mb.

---
 drivers/common/qat/meson.build       |  10 +++
 drivers/crypto/qat/qat_sym_session.c | 101 +++++----------------------
 2 files changed, 28 insertions(+), 83 deletions(-)
drivers/common/qat/meson.build +10 -0
index b7027f3164..d35fc69d96 100644
@@ -35,6 +35,16 @@ if qat_crypto and not libcrypto.found()
             'missing dependency, libcrypto')
 endif
 
+
+IMB_required_ver = '1.0.0'
+libipsecmb = cc.find_library('IPSec_MB', required: false)
+if not lib.found()
+    build = false
+    reason = 'missing dependency, "libIPSec_MB"'
+else
+    ext_deps += libipsecmb
+endif
+
 # The driver should not build if both compression and crypto are disabled
 #FIXME common code depends on compression files so check only compress!
 if not qat_compress # and not qat_crypto
drivers/crypto/qat/qat_sym_session.c +18 -83

163 lines not on this page · raw message

--
2.32.0

#3 lore raw

RE: [External] : [PATCH] crypto/qat: use intel-ipsec-mb for partial hash

Please see v2: http://patchwork.dpdk.org/project/dpdk/list/?series=22973

Regards

Kai

Does it mean the intel-ipsec-mb would be prerequisite of applying QAT offloading for security application? It this is the case, as I know, the intel-ipsec-mb has no FIPS certification yet. Thus I am thinking this would impact existing QAT based security application, right?

Best Regards,
Changchun Zhang

#4 lore raw

PATCH v2 1/2 build: add in option for qat use intel ipsec-mb lib

In reply to Fan Zhang

Add in build option for QAT pmd use intel ipsec-mb lib instead openssl for precompute partial hash & aes.

---
 drivers/common/qat/meson.build | 12 ++++++++++++
 meson_options.txt              |  2 ++
 2 files changed, 14 insertions(+)
drivers/common/qat/meson.build +12 -0
index b7027f3164..d6eaff3e0e 100644
@@ -35,6 +35,18 @@ if qat_crypto and not libcrypto.found()
             'missing dependency, libcrypto')
 endif
 
+if get_option('qat_libipsecmb')
+    IMB_required_ver = '1.0.0'
+    libipsecmb = cc.find_library('IPSec_MB', required: false)
+    if not lib.found()
+        build = false
+        reason = 'missing dependency, "libIPSec_MB"'
+    else
+        ext_deps += libipsecmb
+        dpdk_conf.set('RTE_QAT_LIBIPSECMB', true)
+    endif
+endif
+
 # The driver should not build if both compression and crypto are disabled
 #FIXME common code depends on compression files so check only compress!
 if not qat_compress # and not qat_crypto
meson_options.txt +2 -0
index 7c220ad68d..e48e7dd966 100644
@@ -42,6 +42,8 @@ option('mbuf_refcnt_atomic', type: 'boolean', value: true, description:
        'Atomically access the mbuf refcnt.')
 option('platform', type: 'string', value: 'native', description:
        'Platform to build, either "native", "generic" or a SoC. Please refer to the Linux build guide for more information.')
+option('qat_libipsecmb', type: 'boolean', value: false, description:
+       'use ipsec mb lib instead openssl for QAT pmd')
 option('enable_trace_fp', type: 'boolean', value: false, description:
        'enable fast path trace points.')
 option('tests', type: 'boolean', value: true, description:
Signature · 1 line
2.17.1
#5 lore raw

PATCH v2 2/2 crypto/qat: use intel-ipsec-mb for partial hash & aes

Since openssl 3.0 now deprecates the low level API QAT required to perform partial hash & aes operation when creating the session. This patch is to transfer such dependency from openssl to intel-ipsec-mb.

---
 drivers/crypto/qat/qat_sym_session.c | 353 +++++++++++++++++++++++++--
 1 file changed, 334 insertions(+), 19 deletions(-)
drivers/crypto/qat/qat_sym_session.c +334 -19
index 9d6a19c0be..3f806a4b55 100644
@@ -7,6 +7,10 @@
 #include <openssl/md5.h>	/* Needed to calculate pre-compute values */
 #include <openssl/evp.h>	/* Needed for bpi runt block processing */
 
+#ifdef RTE_QAT_LIBIPSECMB
+#include <intel-ipsec-mb.h>
+#endif
+
 #include <rte_memcpy.h>
 #include <rte_common.h>
 #include <rte_spinlock.h>
@@ -22,6 +26,10 @@
 #include "qat_sym_session.h"
 #include "qat_sym.h"
 
+#if (OPENSSL_VERSION_NUMBER >= 0x30000000L)
+#include <openssl/provider.h>
+#endif
+
 /* SHA1 - 20 bytes - Initialiser state can be found in FIPS stds 180-2 */
 static const uint8_t sha1InitialState[] = {
 	0x67, 0x45, 0x23, 0x01, 0xef, 0xcd, 0xab, 0x89, 0x98, 0xba,
@@ -470,6 +478,21 @@ qat_sym_session_configure(struct rte_cryptodev *dev,
 		return -ENOMEM;
 	}
 
+#if (OPENSSL_VERSION_NUMBER >= 0x30000000L)
+	OSSL_PROVIDER * legacy;
+	OSSL_PROVIDER *deflt;
+
+	/* Load Multiple providers into the default (NULL) library context */
+	legacy = OSSL_PROVIDER_load(NULL, "legacy");
+	if (legacy == NULL)
+		return -EINVAL;
+
+	deflt = OSSL_PROVIDER_load(NULL, "default");
+	if (deflt == NULL) {
+		OSSL_PROVIDER_unload(legacy);
+		return  -EINVAL;
+	}
+#endif
 	ret = qat_sym_session_set_parameters(dev, xform, sess_private_data);
 	if (ret != 0) {
 		QAT_LOG(ERR,
@@ -483,6 +506,10 @@ qat_sym_session_configure(struct rte_cryptodev *dev,
 	set_sym_session_private_data(sess, dev->driver_id,
 		sess_private_data);
 
+# if (OPENSSL_VERSION_NUMBER >= 0x30000000L)
+	OSSL_PROVIDER_unload(legacy);
+	OSSL_PROVIDER_unload(deflt);
+# endif
 	return 0;
 }
 
@@ -1057,6 +1084,297 @@ static int qat_hash_get_block_size(enum icp_qat_hw_auth_algo qat_hash_alg)
 	return -EFAULT;
 }
 
+#define HMAC_IPAD_VALUE	0x36
+#define HMAC_OPAD_VALUE	0x5c
+#define HASH_XCBC_PRECOMP_KEY_NUM 3
+
+static const uint8_t AES_CMAC_SEED[ICP_QAT_HW_AES_128_KEY_SZ];
+
+#ifdef RTE_QAT_LIBIPSECMB
+static int aes_ipsecmb_job(uint8_t *in, uint8_t *out, IMB_MGR *m,
+		const uint8_t *key, uint16_t auth_keylen)
+{
+	int err;
+	struct IMB_JOB *job;
+	DECLARE_ALIGNED(uint32_t expkey[4*15], 16);
+	DECLARE_ALIGNED(uint32_t dust[4*15], 16);
+
+	if (auth_keylen == ICP_QAT_HW_AES_128_KEY_SZ)
+		IMB_AES_KEYEXP_128(m, key, expkey, dust);
+	else if (auth_keylen == ICP_QAT_HW_AES_192_KEY_SZ)
+		IMB_AES_KEYEXP_192(m, key, expkey, dust);
+	else if (auth_keylen == ICP_QAT_HW_AES_256_KEY_SZ)
+		IMB_AES_KEYEXP_256(m, key, expkey, dust);
+	else
+		return -EFAULT;
+
+	job = IMB_GET_NEXT_JOB(m);
+
+	job->src = in;
+	job->dst = out;
+	job->enc_keys = expkey;
+	job->key_len_in_bytes = auth_keylen;
+	job->msg_len_to_cipher_in_bytes = 16;
+	job->iv_len_in_bytes = 0;
+	job->cipher_direction = IMB_DIR_ENCRYPT;
+	job->cipher_mode = IMB_CIPHER_ECB;
+	job->hash_alg = IMB_AUTH_NULL;
+
+	while (IMB_FLUSH_JOB(m) != NULL)
+		;
+
+	job = IMB_SUBMIT_JOB(m);
+	if (job) {
+		if (job->status == IMB_STATUS_COMPLETED)
+			return 0;
+	}
+
+	err = imb_get_errno(m);
+	if (err)
+		QAT_LOG(ERR, "Error: %s!\n", imb_get_strerror(err));
+
+	return -EFAULT;
+}
+
+static int
+partial_hash_compute(enum icp_qat_hw_auth_algo hash_alg,
+		uint8_t *data_in, uint8_t *data_out)
+{
+	int digest_size;
+	uint8_t digest[qat_hash_get_digest_size(
+			ICP_QAT_HW_AUTH_ALGO_DELIMITER)];
+	uint32_t *hash_state_out_be32;
+	uint64_t *hash_state_out_be64;
+	int i;
+	IMB_MGR *m;
+
+	m = alloc_mb_mgr(0);
+	if (m == NULL)
+		return -ENOMEM;
+	init_mb_mgr_auto(m, NULL);
+
+	/* Initialize to avoid gcc warning */
+	memset(digest, 0, sizeof(digest));
+
+	digest_size = qat_hash_get_digest_size(hash_alg);
+	if (digest_size <= 0)
+		return -EFAULT;
+
+	hash_state_out_be32 = (uint32_t *)data_out;
+	hash_state_out_be64 = (uint64_t *)data_out;
+
+	switch (hash_alg) {
+	case ICP_QAT_HW_AUTH_ALGO_SHA1:
+		IMB_SHA1_ONE_BLOCK(m, data_in, digest);
+		for (i = 0; i < digest_size >> 2; i++, hash_state_out_be32++)
+			*hash_state_out_be32 =
+				rte_bswap32(*(((uint32_t *)digest)+i));
+		break;
+	case ICP_QAT_HW_AUTH_ALGO_SHA224:
+		IMB_SHA224_ONE_BLOCK(m, data_in, digest);
+		for (i = 0; i < digest_size >> 2; i++, hash_state_out_be32++)
+			*hash_state_out_be32 =
+				rte_bswap32(*(((uint32_t *)digest)+i));
+		break;
+	case ICP_QAT_HW_AUTH_ALGO_SHA256:
+		IMB_SHA256_ONE_BLOCK(m, data_in, digest);
+		for (i = 0; i < digest_size >> 2; i++, hash_state_out_be32++)
+			*hash_state_out_be32 =
+				rte_bswap32(*(((uint32_t *)digest)+i));
+		break;
+	case ICP_QAT_HW_AUTH_ALGO_SHA384:
+		IMB_SHA384_ONE_BLOCK(m, data_in, digest);
+		for (i = 0; i < digest_size >> 3; i++, hash_state_out_be64++)
+			*hash_state_out_be64 =
+				rte_bswap64(*(((uint64_t *)digest)+i));
+		break;
+	case ICP_QAT_HW_AUTH_ALGO_SHA512:
+		IMB_SHA512_ONE_BLOCK(m, data_in, digest);
+		for (i = 0; i < digest_size >> 3; i++, hash_state_out_be64++)
+			*hash_state_out_be64 =
+				rte_bswap64(*(((uint64_t *)digest)+i));
+		break;
+	case ICP_QAT_HW_AUTH_ALGO_MD5:
+		IMB_MD5_ONE_BLOCK(m, data_in, data_out);
+		break;
+	default:
+		QAT_LOG(ERR, "invalid hash alg %u", hash_alg);
+		return -EFAULT;
+	}
+	free_mb_mgr(m);
+	return 0;
+}
+
+static int qat_sym_do_precomputes(enum icp_qat_hw_auth_algo hash_alg,
+				const uint8_t *auth_key,
+				uint16_t auth_keylen,
+				uint8_t *p_state_buf,
+				uint16_t *p_state_len,
+				uint8_t aes_cmac)
+{
+	int block_size;
+	uint8_t ipad[qat_hash_get_block_size(ICP_QAT_HW_AUTH_ALGO_DELIMITER)];
+	uint8_t opad[qat_hash_get_block_size(ICP_QAT_HW_AUTH_ALGO_DELIMITER)];
+	int i;
+
+	IMB_MGR *m;
+	m = alloc_mb_mgr(0);
+	if (m == NULL)
+		return -ENOMEM;
+
+	init_mb_mgr_auto(m, NULL);
+
+	if (hash_alg == ICP_QAT_HW_AUTH_ALGO_AES_XCBC_MAC) {
+
+		/* CMAC */
+		if (aes_cmac) {
+			uint8_t *in = NULL;
+			uint8_t k0[ICP_QAT_HW_AES_128_KEY_SZ];
+			uint8_t *k1, *k2;
+
+			auth_keylen = ICP_QAT_HW_AES_128_KEY_SZ;
+
+			in = rte_zmalloc("AES CMAC K1",
+					 ICP_QAT_HW_AES_128_KEY_SZ, 16);
+
+			if (in == NULL) {
+				QAT_LOG(ERR, "Failed to alloc memory");
+				return -ENOMEM;
+			}
+
+			rte_memcpy(in, AES_CMAC_SEED,
+				   ICP_QAT_HW_AES_128_KEY_SZ);
+			rte_memcpy(p_state_buf, auth_key, auth_keylen);
+
+			DECLARE_ALIGNED(uint32_t expkey[4*15], 16);
+			DECLARE_ALIGNED(uint32_t dust[4*15], 16);
+			IMB_AES_KEYEXP_128(m, p_state_buf, expkey, dust);
+			k1 = p_state_buf + ICP_QAT_HW_AES_XCBC_MAC_STATE1_SZ;
+			k2 = k1 + ICP_QAT_HW_AES_XCBC_MAC_STATE1_SZ;
+
+			IMB_AES_CMAC_SUBKEY_GEN_128(m, expkey, k1, k2);
+			memset(k0, 0, ICP_QAT_HW_AES_128_KEY_SZ);
+			*p_state_len = ICP_QAT_HW_AES_XCBC_MAC_STATE2_SZ;
+			rte_free(in);
+			free_mb_mgr(m);
+			return 0;
+		}
+
+		static uint8_t qat_aes_xcbc_key_seed[
+				ICP_QAT_HW_AES_XCBC_MAC_STATE2_SZ] = {
+			0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01,
+			0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01,
+			0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02,
+			0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02,
+			0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x03,
+			0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x03,
+		};
+
+		uint8_t *in = NULL;
+		uint8_t *out = p_state_buf;
+		int x;
+
+		in = rte_zmalloc("working mem for key",
+				ICP_QAT_HW_AES_XCBC_MAC_STATE2_SZ, 16);
+		if (in == NULL) {
+			QAT_LOG(ERR, "Failed to alloc memory");
+			return -ENOMEM;
+		}
+
+		rte_memcpy(in, qat_aes_xcbc_key_seed,
+				ICP_QAT_HW_AES_XCBC_MAC_STATE2_SZ);
+		for (x = 0; x < HASH_XCBC_PRECOMP_KEY_NUM; x++) {
+			if (aes_ipsecmb_job(in, out, m, auth_key, auth_keylen)) {
+				rte_free(in -
+				  (x * ICP_QAT_HW_AES_XCBC_MAC_KEY_SZ));
+				memset(out -
+				   (x * ICP_QAT_HW_AES_XCBC_MAC_KEY_SZ),
+				  0, ICP_QAT_HW_AES_XCBC_MAC_STATE2_SZ);
+				return -EFAULT;
+			}
+
+			in += ICP_QAT_HW_AES_XCBC_MAC_KEY_SZ;
+			out += ICP_QAT_HW_AES_XCBC_MAC_KEY_SZ;
+		}
+		*p_state_len = ICP_QAT_HW_AES_XCBC_MAC_STATE2_SZ;
+		rte_free(in - x*ICP_QAT_HW_AES_XCBC_MAC_KEY_SZ);
+		free_mb_mgr(m);
+		return 0;
+
+	} else if ((hash_alg == ICP_QAT_HW_AUTH_ALGO_GALOIS_128) ||
+		(hash_alg == ICP_QAT_HW_AUTH_ALGO_GALOIS_64)) {
+		uint8_t *in = NULL;
+		uint8_t *out = p_state_buf;
+
+		memset(p_state_buf, 0, ICP_QAT_HW_GALOIS_H_SZ +
+				ICP_QAT_HW_GALOIS_LEN_A_SZ +
+				ICP_QAT_HW_GALOIS_E_CTR0_SZ);
+		in = rte_zmalloc("working mem for key",
+				ICP_QAT_HW_GALOIS_H_SZ, 16);
+		if (in == NULL) {
+			QAT_LOG(ERR, "Failed to alloc memory");
+			return -ENOMEM;
+		}
+
+		memset(in, 0, ICP_QAT_HW_GALOIS_H_SZ);
+		if (aes_ipsecmb_job(in, out, m, auth_key, auth_keylen))
+			return -EFAULT;
+
+		*p_state_len = ICP_QAT_HW_GALOIS_H_SZ +
+				ICP_QAT_HW_GALOIS_LEN_A_SZ +
+				ICP_QAT_HW_GALOIS_E_CTR0_SZ;
+		rte_free(in);
+		free_mb_mgr(m);
+		return 0;
+	}
+
+	block_size = qat_hash_get_block_size(hash_alg);
+	if (block_size < 0)
+		return block_size;
+	/* init ipad and opad from key and xor with fixed values */
+	memset(ipad, 0, block_size);
+	memset(opad, 0, block_size);
+
+	if (auth_keylen > (unsigned int)block_size) {
+		QAT_LOG(ERR, "invalid keylen %u", auth_keylen);
+		return -EFAULT;
+	}
+	rte_memcpy(ipad, auth_key, auth_keylen);
+	rte_memcpy(opad, auth_key, auth_keylen);
+
+	for (i = 0; i < block_size; i++) {
+		uint8_t *ipad_ptr = ipad + i;
+		uint8_t *opad_ptr = opad + i;
+		*ipad_ptr ^= HMAC_IPAD_VALUE;
+		*opad_ptr ^= HMAC_OPAD_VALUE;
+	}
+
+	/* do partial hash of ipad and copy to state1 */
+	if (partial_hash_compute(hash_alg, ipad, p_state_buf)) {
+		memset(ipad, 0, block_size);
+		memset(opad, 0, block_size);
+		QAT_LOG(ERR, "ipad precompute failed");
+		return -EFAULT;
+	}
+
+	/*
+	 * State len is a multiple of 8, so may be larger than the digest.
+	 * Put the partial hash of opad state_len bytes after state1
+	 */
+	*p_state_len = qat_hash_get_state1_size(hash_alg);
+	if (partial_hash_compute(hash_alg, opad, p_state_buf + *p_state_len)) {
+		memset(ipad, 0, block_size);
+		memset(opad, 0, block_size);
+		QAT_LOG(ERR, "opad precompute failed");
+		return -EFAULT;
+	}
+
+	/*  don't leave data lying around */
+	memset(ipad, 0, block_size);
+	memset(opad, 0, block_size);
+	return 0;
+}
+#else
 static int partial_hash_sha1(uint8_t *data_in, uint8_t *data_out)
 {
 	SHA_CTX ctx;
@@ -1124,6 +1442,20 @@ static int partial_hash_md5(uint8_t *data_in, uint8_t *data_out)
 	return 0;
 }
 
+static void aes_cmac_key_derive(uint8_t *base, uint8_t *derived)
+{
+	int i;
+
+	derived[0] = base[0] << 1;
+	for (i = 1; i < ICP_QAT_HW_AES_BLK_SZ ; i++) {
+		derived[i] = base[i] << 1;
+		derived[i - 1] |= base[i] >> 7;
+	}
+
+	if (base[0] & 0x80)
+		derived[ICP_QAT_HW_AES_BLK_SZ - 1] ^= QAT_AES_CMAC_CONST_RB;
+}
+
 static int
 partial_hash_compute(enum icp_qat_hw_auth_algo hash_alg,
 		uint8_t *data_in, uint8_t *data_out)
@@ -1192,25 +1524,6 @@ partial_hash_compute(enum icp_qat_hw_auth_algo hash_alg,
 
 	return 0;
 }
-#define HMAC_IPAD_VALUE	0x36
-#define HMAC_OPAD_VALUE	0x5c
-#define HASH_XCBC_PRECOMP_KEY_NUM 3
-
-static const uint8_t AES_CMAC_SEED[ICP_QAT_HW_AES_128_KEY_SZ];
-
-static void aes_cmac_key_derive(uint8_t *base, uint8_t *derived)
-{
-	int i;
-
-	derived[0] = base[0] << 1;
-	for (i = 1; i < ICP_QAT_HW_AES_BLK_SZ ; i++) {
-		derived[i] = base[i] << 1;
-		derived[i - 1] |= base[i] >> 7;
-	}
-
-	if (base[0] & 0x80)
-		derived[ICP_QAT_HW_AES_BLK_SZ - 1] ^= QAT_AES_CMAC_CONST_RB;
-}
 
 static int qat_sym_do_precomputes(enum icp_qat_hw_auth_algo hash_alg,
 				const uint8_t *auth_key,
@@ -1279,6 +1592,7 @@ static int qat_sym_do_precomputes(enum icp_qat_hw_auth_algo hash_alg,
 			uint8_t *in = NULL;
 			uint8_t *out = p_state_buf;
 			int x;
+
 			AES_KEY enc_key;
 
 			in = rte_zmalloc("working mem for key",
@@ -1385,6 +1699,7 @@ static int qat_sym_do_precomputes(enum icp_qat_hw_auth_algo hash_alg,
 	memset(opad, 0, block_size);
 	return 0;
 }
+#endif
 
 static void
 qat_sym_session_init_common_hdr(struct qat_sym_session *session)
Signature · 1 line
2.17.1
author Acked-by on v2 2/2
#6 lore raw

RE: [dpdk-dev v2 2/2] crypto/qat: use intel-ipsec-mb for partial hash & aes

Quoted 8 lines

-----Original Message-----
From: Ji, Kai <[email protected]>
Sent: Tuesday, May 17, 2022 3:17 PM
To: [email protected]
Cc: Ji, Kai <[email protected]>; Zhang, Roy Fan <[email protected]>
Subject: [dpdk-dev v2 2/2] crypto/qat: use intel-ipsec-mb for partial hash & aes

Quoted 7 lines

Since openssl 3.0 now deprecates the low level API QAT required to perform partial hash & aes operation when creating the session. This patch is to transfer such dependency from openssl to intel-ipsec-mb.

---

Please don't forget to CC maintainer next time :-)

author Acked-by on v2 1/2
#7 lore raw

RE: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

In reply to Kai Ji
Quoted 7 lines

-----Original Message-----
From: Kai Ji <[email protected]>
Sent: Tuesday, May 17, 2022 3:17 PM
To: [email protected]
Cc: Ji, Kai <[email protected]>
Subject: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

Quoted 5 lines

Add in build option for QAT pmd use intel ipsec-mb lib instead openssl for precompute partial hash & aes.

---

Please don't forget to CC maintainer next time :-)

#8 lore raw

Re: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

In reply to Kai Ji
Quoted 19 lines · Kai Ji

Add in build option for QAT pmd use intel ipsec-mb lib instead openssl for precompute partial hash & aes.

---
 drivers/common/qat/meson.build | 12 ++++++++++++
 meson_options.txt              |  2 ++
 2 files changed, 14 insertions(+)
drivers/common/qat/meson.build +2 -0
index b7027f3164..d6eaff3e0e 100644
@@ -35,6 +35,18 @@ if qat_crypto and not libcrypto.found()
             'missing dependency, libcrypto')
 endif
 
+if get_option('qat_libipsecmb')
+    IMB_required_ver = '1.0.0'

drivers/common/qat/meson.build if qat_crypto and not libcrypto.found() · line 49

Quoted 10 lines
drivers/common/qat/meson.build +10 -0
+    libipsecmb = cc.find_library('IPSec_MB', required: false)
+    if not lib.found()
+        build = false
+        reason = 'missing dependency, "libIPSec_MB"'
+    else
+        ext_deps += libipsecmb
+        dpdk_conf.set('RTE_QAT_LIBIPSECMB', true)
+    endif
+endif
+

In general it's not a good idea for individual drivers to add top-level meson.build options. If every driver does this for all its options we will have an unmanageable set of hundreds of options. Unfortunately, though, I don't see a really good way to replace this with something other than a build option - I assume it's not just good enough to check which of the dependencies is found/not-found? Is there some way that this can be a generic option, that may be applicable to other drivers? Could the option be instead to prioritize FIPs compatibility?

/Bruce

#9 lore raw

Re: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

Quoted 31 lines · Bruce Richardson
Quoted 19 lines · Kai Ji

Add in build option for QAT pmd use intel ipsec-mb lib instead openssl for precompute partial hash & aes.

---
  drivers/common/qat/meson.build | 12 ++++++++++++
  meson_options.txt              |  2 ++
  2 files changed, 14 insertions(+)
drivers/common/qat/meson.build +2 -0
index b7027f3164..d6eaff3e0e 100644
@@ -35,6 +35,18 @@ if qat_crypto and not libcrypto.found()
              'missing dependency, libcrypto')
  endif
 
+if get_option('qat_libipsecmb')
+    IMB_required_ver = '1.0.0'

drivers/common/qat/meson.build if qat_crypto and not libcrypto.found() · line 49

Quoted 10 lines
drivers/common/qat/meson.build +10 -0
+    libipsecmb = cc.find_library('IPSec_MB', required: false)
+    if not lib.found()
+        build = false
+        reason = 'missing dependency, "libIPSec_MB"'
+    else
+        ext_deps += libipsecmb
+        dpdk_conf.set('RTE_QAT_LIBIPSECMB', true)
+    endif
+endif
+
Quoted 9 lines

In general it's not a good idea for individual drivers to add top-level meson.build options. If every driver does this for all its options we will have an unmanageable set of hundreds of options. Unfortunately, though, I don't see a really good way to replace this with something other than a build option - I assume it's not just good enough to check which of the dependencies is found/not-found? Is there some way that this can be a generic option, that may be applicable to other drivers? Could the option be instead to prioritize FIPs compatibility?

+1 to not have top level meson option for a driver.

Will it work to give priority to intel-ipsec-mb lib, like when ipsec-mb library exists use it, else use openssl?

#10 lore raw

Re: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

Quoted 41 lines · Ferruh Yigit
Quoted 30 lines · Bruce Richardson
Quoted 18 lines · Kai Ji

Add in build option for QAT pmd use intel ipsec-mb lib instead openssl for precompute partial hash & aes.

---
  drivers/common/qat/meson.build | 12 ++++++++++++
  meson_options.txt              |  2 ++
  2 files changed, 14 insertions(+)
drivers/common/qat/meson.build +2 -0
index b7027f3164..d6eaff3e0e 100644
@@ -35,6 +35,18 @@ if qat_crypto and not libcrypto.found()
              'missing dependency, libcrypto')
  endif
+if get_option('qat_libipsecmb')
+    IMB_required_ver = '1.0.0'

drivers/common/qat/meson.build if qat_crypto and not libcrypto.found() · line 48

Quoted 10 lines
drivers/common/qat/meson.build +10 -0
+    libipsecmb = cc.find_library('IPSec_MB', required: false)
+    if not lib.found()
+        build = false
+        reason = 'missing dependency, "libIPSec_MB"'
+    else
+        ext_deps += libipsecmb
+        dpdk_conf.set('RTE_QAT_LIBIPSECMB', true)
+    endif
+endif
+
Quoted 9 lines

In general it's not a good idea for individual drivers to add top-level meson.build options. If every driver does this for all its options we will have an unmanageable set of hundreds of options. Unfortunately, though, I don't see a really good way to replace this with something other than a build option - I assume it's not just good enough to check which of the dependencies is found/not-found? Is there some way that this can be a generic option, that may be applicable to other drivers? Could the option be instead to prioritize FIPs compatibility?

Quoted 4 lines

+1 to not have top level meson option for a driver.

Will it work to give priority to intel-ipsec-mb lib, like when ipsec-mb library exists use it, else use openssl?

Looking at the patch a bit more, is there a reason why this cannot be made a run-time option when both libraries are available?

#11 lore raw

RE: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

Hi Bruce,

We want user to have the ability to choice which lib they want to use, so the original design was to make the build option temporary, and it will get removed after ipsec-mb FIP certified.

Will look into the runtime option to avoid top level build option modification.

Regards

Kai

Quoted 2 lines

Looking at the patch a bit more, is there a reason why this cannot be made a run- time option when both libraries are available?

#12 lore raw

RE: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

In reply to Bruce Richardson
Quoted 53 lines

-----Original Message-----
From: Richardson, Bruce <[email protected]>
Sent: Wednesday, May 18, 2022 11:33 AM
To: Ferruh Yigit <[email protected]>
Cc: Ji, Kai <[email protected]>; [email protected]; Zhang, Roy Fan <[email protected]>
Subject: Re: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

Quoted 5 lines · Ferruh Yigit
Quoted 4 lines · Bruce Richardson
Quoted 3 lines · Kai Ji

Add in build option for QAT pmd use intel ipsec-mb lib instead openssl for precompute partial hash & aes.

Quoted 7 lines
Quoted 7 lines

Review

Quoted 7 lines
---
  drivers/common/qat/meson.build | 12 ++++++++++++
  meson_options.txt              |  2 ++
  2 files changed, 14 insertions(+)
(quoted diff) +0 -0

b/drivers/common/qat/meson.build

Quoted 14 lines
Quoted 14 lines
Quoted 14 lines

index b7027f3164..d6eaff3e0e 100644

drivers/common/qat/meson.build +8 -0
@@ -35,6 +35,18 @@ if qat_crypto and not libcrypto.found()
              'missing dependency, libcrypto')
  endif
+if get_option('qat_libipsecmb')
+    IMB_required_ver = '1.0.0'
+    libipsecmb = cc.find_library('IPSec_MB', required: false)
+    if not lib.found()
+        build = false
+        reason = 'missing dependency, "libIPSec_MB"'
+    else
+        ext_deps += libipsecmb
Quoted 8 lines
Quoted 8 lines

drivers/common/qat/meson.build if qat_crypto and not libcrypto.found() · line 49

Quoted 4 lines
drivers/common/qat/meson.build +4 -0
+        dpdk_conf.set('RTE_QAT_LIBIPSECMB', true)
+    endif
+endif
+

In general it's not a good idea for individual drivers to add top-level meson.build options. If every driver does this for all its options we will have an unmanageable set of hundreds of options. Unfortunately,

though, I

Quoted 7 lines
Quoted 6 lines

don't see a really good way to replace this with something other than a build option - I assume it's not just good enough to check which of the dependencies is found/not-found? Is there some way that this can be a generic option, that may be applicable to other drivers? Could the option be instead to prioritize FIPs compatibility?

Quoted 7 lines
Quoted 4 lines

+1 to not have top level meson option for a driver.

Will it work to give priority to intel-ipsec-mb lib, like when ipsec-mb library exists use it, else use openssl?

Looking at the patch a bit more, is there a reason why this cannot be made a run-time option when both libraries are available?

Hi Bruce,

Thanks for the review.
Are you suggesting adding a EAL option to the QAT PMD?

Regards,
Fan

#13 lore raw

Re: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

Quoted 58 lines · Roy Fan
Quoted 58 lines

-----Original Message-----
From: Richardson, Bruce <[email protected]>
Sent: Wednesday, May 18, 2022 11:33 AM
To: Ferruh Yigit <[email protected]>
Cc: Ji, Kai <[email protected]>; [email protected]; Zhang, Roy Fan <[email protected]>
Subject: Re: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

Quoted 5 lines · Ferruh Yigit
Quoted 4 lines · Bruce Richardson
Quoted 3 lines · Kai Ji

Add in build option for QAT pmd use intel ipsec-mb lib instead openssl for precompute partial hash & aes.

Quoted 7 lines
Quoted 7 lines

Review

Quoted 7 lines
---
  drivers/common/qat/meson.build | 12 ++++++++++++
  meson_options.txt              |  2 ++
  2 files changed, 14 insertions(+)
(quoted diff) +0 -0

b/drivers/common/qat/meson.build

Quoted 14 lines
Quoted 14 lines
Quoted 14 lines

index b7027f3164..d6eaff3e0e 100644

drivers/common/qat/meson.build +8 -0
@@ -35,6 +35,18 @@ if qat_crypto and not libcrypto.found()
              'missing dependency, libcrypto')
  endif
+if get_option('qat_libipsecmb')
+    IMB_required_ver = '1.0.0'
+    libipsecmb = cc.find_library('IPSec_MB', required: false)
+    if not lib.found()
+        build = false
+        reason = 'missing dependency, "libIPSec_MB"'
+    else
+        ext_deps += libipsecmb
Quoted 8 lines
Quoted 8 lines

drivers/common/qat/meson.build if qat_crypto and not libcrypto.found() · line 49

Quoted 4 lines
drivers/common/qat/meson.build +4 -0
+        dpdk_conf.set('RTE_QAT_LIBIPSECMB', true)
+    endif
+endif
+

In general it's not a good idea for individual drivers to add top-level meson.build options. If every driver does this for all its options we will have an unmanageable set of hundreds of options. Unfortunately,

though, I

Quoted 11 lines
Quoted 6 lines

don't see a really good way to replace this with something other than a build option - I assume it's not just good enough to check which of the dependencies is found/not-found? Is there some way that this can be a generic option, that may be applicable to other drivers? Could the option be instead to prioritize FIPs compatibility?

+1 to not have top level meson option for a driver.

Will it work to give priority to intel-ipsec-mb lib, like when ipsec-mb library exists use it, else use openssl?

Quoted 8 lines
Quoted 2 lines

Looking at the patch a bit more, is there a reason why this cannot be made a run-time option when both libraries are available?

Hi Bruce,

Thanks for the review.
Are you suggesting adding a EAL option to the QAT PMD?

I was thinking you can add a driver option, to allow switching from the default if both openssl and ipsec-mb were available to be compiled in.

#14 lore raw

RE: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

Hi Bruce,

Quoted 76 lines

-----Original Message-----
From: Richardson, Bruce <[email protected]>
Sent: Thursday, May 19, 2022 11:23 AM
To: Zhang, Roy Fan <[email protected]>
Cc: Ferruh Yigit <[email protected]>; Ji, Kai <[email protected]>; [email protected]
Subject: Re: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

Quoted 7 lines · Roy Fan
Quoted 7 lines

-----Original Message-----
From: Richardson, Bruce <[email protected]>
Sent: Wednesday, May 18, 2022 11:33 AM
To: Ferruh Yigit <[email protected]>
Cc: Ji, Kai <[email protected]>; [email protected]; Zhang, Roy Fan <[email protected]>
Subject: Re: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-

mb

Quoted 38 lines
Quoted 38 lines

lib

Quoted 5 lines · Ferruh Yigit
Quoted 4 lines · Bruce Richardson
Quoted 3 lines · Kai Ji

Add in build option for QAT pmd use intel ipsec-mb lib instead openssl for precompute partial hash & aes.

Quoted 7 lines
Quoted 7 lines

Review

Quoted 7 lines
---
  drivers/common/qat/meson.build | 12 ++++++++++++
  meson_options.txt              |  2 ++
  2 files changed, 14 insertions(+)
(quoted diff) +0 -0

b/drivers/common/qat/meson.build

Quoted 22 lines
Quoted 22 lines
Quoted 8 lines

index b7027f3164..d6eaff3e0e 100644

drivers/common/qat/meson.build +2 -0
@@ -35,6 +35,18 @@ if qat_crypto and not libcrypto.found()
              'missing dependency, libcrypto')
  endif
+if get_option('qat_libipsecmb')
+    IMB_required_ver = '1.0.0'

drivers/common/qat/meson.build if qat_crypto and not libcrypto.found() · line 48

Quoted 10 lines
drivers/common/qat/meson.build +10 -0
+    libipsecmb = cc.find_library('IPSec_MB', required: false)
+    if not lib.found()
+        build = false
+        reason = 'missing dependency, "libIPSec_MB"'
+    else
+        ext_deps += libipsecmb
+        dpdk_conf.set('RTE_QAT_LIBIPSECMB', true)
+    endif
+endif
+

In general it's not a good idea for individual drivers to add top-level meson.build options. If every driver does this for all its options we will have an unmanageable set of hundreds of options. Unfortunately,

Quoted 2 lines
Quoted 2 lines

though, I

Quoted 1 line
Quoted 1 line

don't see a really good way to replace this with something other than

a

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

build option - I assume it's not just good enough to check which of the dependencies is found/not-found? Is there some way that this can be

a

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

generic option, that may be applicable to other drivers? Could the

option

Quoted 9 lines
Quoted 9 lines
Quoted 7 lines
Quoted 2 lines

be instead to prioritize FIPs compatibility?

+1 to not have top level meson option for a driver.

Will it work to give priority to intel-ipsec-mb lib, like when ipsec-mb library exists use it, else use openssl?

Looking at the patch a bit more, is there a reason why this cannot be

made

Quoted 2 lines
Quoted 1 line

a run-time option when both libraries are available?

Quoted 7 lines
Quoted 5 lines

Hi Bruce,

Thanks for the review.
Are you suggesting adding a EAL option to the QAT PMD?

I was thinking you can add a driver option, to allow switching from the default if both openssl and ipsec-mb were available to be compiled in.

We may have to use macros to filter out openssl deprecated API calls otherwise the warning may still be there. Also what kind of driver option are we talking about - any reference you can point to us? Much appreciated!

Regards,
Fan

#15 lore raw

Re: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

Quoted 83 lines · Roy Fan

Hi Bruce,

Quoted 81 lines

-----Original Message-----
From: Richardson, Bruce <[email protected]>
Sent: Thursday, May 19, 2022 11:23 AM
To: Zhang, Roy Fan <[email protected]>
Cc: Ferruh Yigit <[email protected]>; Ji, Kai <[email protected]>; [email protected]
Subject: Re: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

Quoted 7 lines · Roy Fan
Quoted 7 lines

-----Original Message-----
From: Richardson, Bruce <[email protected]>
Sent: Wednesday, May 18, 2022 11:33 AM
To: Ferruh Yigit <[email protected]>
Cc: Ji, Kai <[email protected]>; [email protected]; Zhang, Roy Fan <[email protected]>
Subject: Re: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-

mb

Quoted 38 lines
Quoted 38 lines

lib

Quoted 5 lines · Ferruh Yigit
Quoted 4 lines · Bruce Richardson
Quoted 3 lines · Kai Ji

Add in build option for QAT pmd use intel ipsec-mb lib instead openssl for precompute partial hash & aes.

Quoted 7 lines
Quoted 7 lines

Review

Quoted 7 lines
---
  drivers/common/qat/meson.build | 12 ++++++++++++
  meson_options.txt              |  2 ++
  2 files changed, 14 insertions(+)
(quoted diff) +0 -0

b/drivers/common/qat/meson.build

Quoted 22 lines
Quoted 22 lines
Quoted 8 lines

index b7027f3164..d6eaff3e0e 100644

drivers/common/qat/meson.build +2 -0
@@ -35,6 +35,18 @@ if qat_crypto and not libcrypto.found()
              'missing dependency, libcrypto')
  endif
+if get_option('qat_libipsecmb')
+    IMB_required_ver = '1.0.0'

drivers/common/qat/meson.build if qat_crypto and not libcrypto.found() · line 48

Quoted 10 lines
drivers/common/qat/meson.build +10 -0
+    libipsecmb = cc.find_library('IPSec_MB', required: false)
+    if not lib.found()
+        build = false
+        reason = 'missing dependency, "libIPSec_MB"'
+    else
+        ext_deps += libipsecmb
+        dpdk_conf.set('RTE_QAT_LIBIPSECMB', true)
+    endif
+endif
+

In general it's not a good idea for individual drivers to add top-level meson.build options. If every driver does this for all its options we will have an unmanageable set of hundreds of options. Unfortunately,

Quoted 2 lines
Quoted 2 lines

though, I

Quoted 1 line
Quoted 1 line

don't see a really good way to replace this with something other than

a

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

build option - I assume it's not just good enough to check which of the dependencies is found/not-found? Is there some way that this can be

a

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

generic option, that may be applicable to other drivers? Could the

option

Quoted 9 lines
Quoted 9 lines
Quoted 7 lines
Quoted 2 lines

be instead to prioritize FIPs compatibility?

+1 to not have top level meson option for a driver.

Will it work to give priority to intel-ipsec-mb lib, like when ipsec-mb library exists use it, else use openssl?

Looking at the patch a bit more, is there a reason why this cannot be

made

Quoted 7 lines
Quoted 1 line

a run-time option when both libraries are available?

Hi Bruce,

Thanks for the review.
Are you suggesting adding a EAL option to the QAT PMD?

Quoted 7 lines
Quoted 2 lines

I was thinking you can add a driver option, to allow switching from the default if both openssl and ipsec-mb were available to be compiled in.

We may have to use macros to filter out openssl deprecated API calls otherwise the warning may still be there. Also what kind of driver option are we talking about - any reference you can point to us? Much appreciated!

Many drivers support additional parameters, but the vdev's are the most common case. For example, AF_XDP:

http://git.dpdk.org/dpdk/tree/drivers/net/af_xdp/rte_eth_af_xdp.c#n2056

or for a physical PCI devices, i40e:

http://git.dpdk.org/dpdk/tree/drivers/net/i40e/i40e_ethdev.c#n12133

/Bruce

#16 lore raw

RE: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

In reply to Ferruh Yigit
Quoted 55 lines

-----Original Message-----
From: Ferruh Yigit <[email protected]>
Sent: Wednesday, May 18, 2022 11:26 AM
To: Richardson, Bruce <[email protected]>; Ji, Kai <[email protected]>
Cc: [email protected]; Zhang, Roy Fan <[email protected]>
Subject: Re: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

Quoted 11 lines · Bruce Richardson

Review

Quoted 10 lines · Kai Ji

Add in build option for QAT pmd use intel ipsec-mb lib instead openssl for precompute partial hash & aes.

---
  drivers/common/qat/meson.build | 12 ++++++++++++
  meson_options.txt              |  2 ++
  2 files changed, 14 insertions(+)
(quoted diff) +0 -0

b/drivers/common/qat/meson.build

Quoted 7 lines
Quoted 7 lines

index b7027f3164..d6eaff3e0e 100644

drivers/common/qat/meson.build +0 -0
@@ -35,6 +35,18 @@ if qat_crypto and not libcrypto.found()
              'missing dependency, libcrypto')
  endif
 
Quoted 16 lines

drivers/common/qat/meson.build if qat_crypto and not libcrypto.found() · line 49

Quoted 12 lines
drivers/common/qat/meson.build +12 -0
+if get_option('qat_libipsecmb')
+    IMB_required_ver = '1.0.0'
+    libipsecmb = cc.find_library('IPSec_MB', required: false)
+    if not lib.found()
+        build = false
+        reason = 'missing dependency, "libIPSec_MB"'
+    else
+        ext_deps += libipsecmb
+        dpdk_conf.set('RTE_QAT_LIBIPSECMB', true)
+    endif
+endif
+

In general it's not a good idea for individual drivers to add top-level meson.build options. If every driver does this for all its options we will have an unmanageable set of hundreds of options. Unfortunately, though,

I

Quoted 6 lines

don't see a really good way to replace this with something other than a build option - I assume it's not just good enough to check which of the dependencies is found/not-found? Is there some way that this can be a generic option, that may be applicable to other drivers? Could the option be instead to prioritize FIPs compatibility?

Quoted 4 lines

+1 to not have top level meson option for a driver.

Will it work to give priority to intel-ipsec-mb lib, like when ipsec-mb library exists use it, else use openssl?

Hi Ferruh,

If both ipsec-mb and openssl3.0 presents in the system.
- Choose openssl 3.0 will result in deprecated API warnings.
- Choose ipsec-mb will ruin the FIPs compliance of QAT PMD.
That's why we think it shall be up to user to select which way to do small cipher/hash op.

Regards,
fan

#17 lore raw

Re: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

Quoted 59 lines · Roy Fan
Quoted 54 lines

-----Original Message-----
From: Ferruh Yigit <[email protected]>
Sent: Wednesday, May 18, 2022 11:26 AM
To: Richardson, Bruce <[email protected]>; Ji, Kai <[email protected]>
Cc: [email protected]; Zhang, Roy Fan <[email protected]>
Subject: Re: [dpdk-dev v2 1/2] build: add in option for qat use intel ipsec-mb lib

Quoted 11 lines · Bruce Richardson

Review

Quoted 10 lines · Kai Ji

Add in build option for QAT pmd use intel ipsec-mb lib instead openssl for precompute partial hash & aes.

---
   drivers/common/qat/meson.build | 12 ++++++++++++
   meson_options.txt              |  2 ++
   2 files changed, 14 insertions(+)
(quoted diff) +0 -0

b/drivers/common/qat/meson.build

Quoted 7 lines
Quoted 7 lines

index b7027f3164..d6eaff3e0e 100644

drivers/common/qat/meson.build +0 -0
@@ -35,6 +35,18 @@ if qat_crypto and not libcrypto.found()
               'missing dependency, libcrypto')
   endif
 
Quoted 16 lines

drivers/common/qat/meson.build if qat_crypto and not libcrypto.found() · line 49

Quoted 12 lines
drivers/common/qat/meson.build +12 -0
+if get_option('qat_libipsecmb')
+    IMB_required_ver = '1.0.0'
+    libipsecmb = cc.find_library('IPSec_MB', required: false)
+    if not lib.found()
+        build = false
+        reason = 'missing dependency, "libIPSec_MB"'
+    else
+        ext_deps += libipsecmb
+        dpdk_conf.set('RTE_QAT_LIBIPSECMB', true)
+    endif
+endif
+

In general it's not a good idea for individual drivers to add top-level meson.build options. If every driver does this for all its options we will have an unmanageable set of hundreds of options. Unfortunately, though,

I

Quoted 6 lines

don't see a really good way to replace this with something other than a build option - I assume it's not just good enough to check which of the dependencies is found/not-found? Is there some way that this can be a generic option, that may be applicable to other drivers? Could the option be instead to prioritize FIPs compatibility?

Quoted 4 lines

+1 to not have top level meson option for a driver.

Will it work to give priority to intel-ipsec-mb lib, like when ipsec-mb library exists use it, else use openssl?

Quoted 7 lines

Hi Ferruh,

If both ipsec-mb and openssl3.0 presents in the system.
- Choose openssl 3.0 will result in deprecated API warnings.
- Choose ipsec-mb will ruin the FIPs compliance of QAT PMD.
That's why we think it shall be up to user to select which way to do small cipher/hash op.

Hi Fan, got it. So the runtime devarg option Bruce mentioned can be a solution, if it works.