diff mbox series

[v5] cryptodev: support multiple cipher data-units

Message ID 20210414202159.1075118-1-thomas@monjalon.net (mailing list archive)
State Accepted, archived
Delegated to: akhil goyal
Headers show
Series [v5] cryptodev: support multiple cipher data-units | expand

Checks

Context Check Description
ci/iol-testing fail Testing issues
ci/github-robot success github build: passed
ci/travis-robot success travis build: passed
ci/Intel-compilation success Compilation OK
ci/checkpatch success coding style OK

Commit Message

Thomas Monjalon April 14, 2021, 8:21 p.m. UTC 
From: Matan Azrad <matan@nvidia.com>

In cryptography, a block cipher is a deterministic algorithm operating
on fixed-length groups of bits, called blocks.

A block cipher consists of two paired algorithms, one for encryption
and the other for decryption. Both algorithms accept two inputs:
an input block of size n bits and a key of size k bits; and both yield
an n-bit output block. The decryption algorithm is defined to be the
inverse function of the encryption.

For AES standard the block size is 16 bytes.
For AES in XTS mode, the data to be encrypted\decrypted does not have to
be multiple of 16B size, the unit of data is called data-unit.
The data-unit size can be any size in range [16B, 2^24B], so, in this
case, a data stream is divided into N amount of equal data-units and
must be encrypted\decrypted in the same data-unit resolution.

For ABI compatibility reason, the size is limited to 64K (16-bit field).
The new field dataunit_len is inserted in a struct padding hole,
which is only 2 bytes long in 32-bit build.
It could be moved and extended later during an ABI-breakage window.

The current cryptodev API doesn't allow the user to select a specific
data-unit length supported by the devices.
In addition, there is no definition how the IV is detected per data-unit
when single operation includes more than one data-unit.

That causes applications to use single operation per data-unit even though
all the data is continuous in memory what reduces datapath performance.

Add a new feature flag to support multiple data-unit sizes, called
RTE_CRYPTODEV_FF_CIPHER_MULTIPLE_DATA_UNITS.
Add a new field in cipher capability, called dataunit_set,
where the devices can report the range of the supported data-unit sizes.
Add a new cipher transformation field, called dataunit_len, where the user
can select the data-unit length for all the operations.

All the new fields do not change the size of their structures,
by filling some struct padding holes.
They are added as exceptions in the ABI check file libabigail.abignore.

Using a bitmap to report the supported data-unit sizes capability allows
the devices to report a range simply as same as the user to read it
simply. also, thus sizes are usually common and probably will be shared
among different devices.

Signed-off-by: Matan Azrad <matan@nvidia.com>
Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
---
v1:
- Use data-unit term instead of block.
- Update cipher length description in OP.
- Improve descriptions on xform and capability.
- Improve commit log.

v2:
- Fix typo: MULITPLE->MULTIPLE.
- Remain only planned supported sizes for data-unit capability.

v3:
- Improve some comments.
- Fix ABI breakage.

v4:
- Remove RTE_CRYPTO_CIPHER_DATA_UNIT_LEN_1M_BYTES
- Fix dataunit_len placement in padding hole

v5:
- Improve comments
- Use RTE_BIT32

deprecation notice:
https://patches.dpdk.org/project/dpdk/patch/20210414201544.1063413-1-thomas@monjalon.net/
---
 devtools/libabigail.abignore               | 12 ++++++++-
 doc/guides/cryptodevs/features/default.ini |  1 +
 doc/guides/cryptodevs/overview.rst         |  3 +++
 doc/guides/rel_notes/release_21_05.rst     |  6 +++++
 lib/librte_cryptodev/rte_crypto_sym.h      | 29 ++++++++++++++++++++--
 lib/librte_cryptodev/rte_cryptodev.c       |  2 ++
 lib/librte_cryptodev/rte_cryptodev.h       | 16 ++++++++++++
 7 files changed, 66 insertions(+), 3 deletions(-)

Comments

Akhil Goyal April 15, 2021, 8:35 a.m. UTC | #1 
> From: Matan Azrad <matan@nvidia.com>
> 
> In cryptography, a block cipher is a deterministic algorithm operating
> on fixed-length groups of bits, called blocks.
> 
> A block cipher consists of two paired algorithms, one for encryption
> and the other for decryption. Both algorithms accept two inputs:
> an input block of size n bits and a key of size k bits; and both yield
> an n-bit output block. The decryption algorithm is defined to be the
> inverse function of the encryption.
> 
> For AES standard the block size is 16 bytes.
> For AES in XTS mode, the data to be encrypted\decrypted does not have to
> be multiple of 16B size, the unit of data is called data-unit.
> The data-unit size can be any size in range [16B, 2^24B], so, in this
> case, a data stream is divided into N amount of equal data-units and
> must be encrypted\decrypted in the same data-unit resolution.
> 
> For ABI compatibility reason, the size is limited to 64K (16-bit field).
> The new field dataunit_len is inserted in a struct padding hole,
> which is only 2 bytes long in 32-bit build.
> It could be moved and extended later during an ABI-breakage window.
> 
> The current cryptodev API doesn't allow the user to select a specific
> data-unit length supported by the devices.
> In addition, there is no definition how the IV is detected per data-unit
> when single operation includes more than one data-unit.
> 
> That causes applications to use single operation per data-unit even though
> all the data is continuous in memory what reduces datapath performance.
> 
> Add a new feature flag to support multiple data-unit sizes, called
> RTE_CRYPTODEV_FF_CIPHER_MULTIPLE_DATA_UNITS.
> Add a new field in cipher capability, called dataunit_set,
> where the devices can report the range of the supported data-unit sizes.
> Add a new cipher transformation field, called dataunit_len, where the user
> can select the data-unit length for all the operations.
> 
> All the new fields do not change the size of their structures,
> by filling some struct padding holes.
> They are added as exceptions in the ABI check file libabigail.abignore.
> 
> Using a bitmap to report the supported data-unit sizes capability allows
> the devices to report a range simply as same as the user to read it
> simply. also, thus sizes are usually common and probably will be shared
> among different devices.
> 
> Signed-off-by: Matan Azrad <matan@nvidia.com>
> Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> ---
Acked-by: Akhil Goyal <gakhil@marvell.com>
Akhil Goyal April 15, 2021, 7:01 p.m. UTC | #2 
> > From: Matan Azrad <matan@nvidia.com>
> >
> > In cryptography, a block cipher is a deterministic algorithm operating
> > on fixed-length groups of bits, called blocks.
> >
> > A block cipher consists of two paired algorithms, one for encryption
> > and the other for decryption. Both algorithms accept two inputs:
> > an input block of size n bits and a key of size k bits; and both yield
> > an n-bit output block. The decryption algorithm is defined to be the
> > inverse function of the encryption.
> >
> > For AES standard the block size is 16 bytes.
> > For AES in XTS mode, the data to be encrypted\decrypted does not have to
> > be multiple of 16B size, the unit of data is called data-unit.
> > The data-unit size can be any size in range [16B, 2^24B], so, in this
> > case, a data stream is divided into N amount of equal data-units and
> > must be encrypted\decrypted in the same data-unit resolution.
> >
> > For ABI compatibility reason, the size is limited to 64K (16-bit field).
> > The new field dataunit_len is inserted in a struct padding hole,
> > which is only 2 bytes long in 32-bit build.
> > It could be moved and extended later during an ABI-breakage window.
> >
> > The current cryptodev API doesn't allow the user to select a specific
> > data-unit length supported by the devices.
> > In addition, there is no definition how the IV is detected per data-unit
> > when single operation includes more than one data-unit.
> >
> > That causes applications to use single operation per data-unit even though
> > all the data is continuous in memory what reduces datapath performance.
> >
> > Add a new feature flag to support multiple data-unit sizes, called
> > RTE_CRYPTODEV_FF_CIPHER_MULTIPLE_DATA_UNITS.
> > Add a new field in cipher capability, called dataunit_set,
> > where the devices can report the range of the supported data-unit sizes.
> > Add a new cipher transformation field, called dataunit_len, where the user
> > can select the data-unit length for all the operations.
> >
> > All the new fields do not change the size of their structures,
> > by filling some struct padding holes.
> > They are added as exceptions in the ABI check file libabigail.abignore.
> >
> > Using a bitmap to report the supported data-unit sizes capability allows
> > the devices to report a range simply as same as the user to read it
> > simply. also, thus sizes are usually common and probably will be shared
> > among different devices.
> >
> > Signed-off-by: Matan Azrad <matan@nvidia.com>
> > Signed-off-by: Thomas Monjalon <thomas@monjalon.net>
> > ---
> Acked-by: Akhil Goyal <gakhil@marvell.com>

Applied to dpdk-next-crypto

This patch is causing ABI breakage at my end, but since CI is passing, I am applying this patch.
I believe my libabigail version is older than what CI is using.

@thomas : Please pull crypto tree to main and send the dependent patches again so that CI can run on them.

Regards,
Akhil
David Marchand April 15, 2021, 7:31 p.m. UTC | #3 
On Thu, Apr 15, 2021 at 9:01 PM Akhil Goyal <gakhil@marvell.com> wrote:
> This patch is causing ABI breakage at my end, but since CI is passing, I am applying this patch.
> I believe my libabigail version is older than what CI is using.

I discussed this earlier with Thomas.
This is likely the use of anonymous struct/union that was not
understood by libabigail until recently.

The public CI (Travis and GHA) uses libabigail 1.8.
diff mbox series

Patch

diff --git a/devtools/libabigail.abignore b/devtools/libabigail.abignore
index 6c0b38984e..bce940f2df 100644
--- a/devtools/libabigail.abignore
+++ b/devtools/libabigail.abignore
@@ -19,4 +19,14 @@ 
 ; Ignore fields inserted in cacheline boundary of rte_cryptodev
 [suppress_type]
         name = rte_cryptodev
-        has_data_member_inserted_between = {offset_after(attached), end}
\ No newline at end of file
+        has_data_member_inserted_between = {offset_after(attached), end}
+
+; Ignore fields inserted in union boundary of rte_cryptodev_symmetric_capability
+[suppress_type]
+        name = rte_cryptodev_symmetric_capability
+        has_data_member_inserted_between = {offset_after(cipher.iv_size), end}
+
+; Ignore fields inserted in middle padding of rte_crypto_cipher_xform
+[suppress_type]
+        name = rte_crypto_cipher_xform
+        has_data_member_inserted_between = {offset_after(key), offset_of(iv)}
diff --git a/doc/guides/cryptodevs/features/default.ini b/doc/guides/cryptodevs/features/default.ini
index 17b177fc45..978bb30cc1 100644
--- a/doc/guides/cryptodevs/features/default.ini
+++ b/doc/guides/cryptodevs/features/default.ini
@@ -31,6 +31,7 @@  CPU crypto             =
 Symmetric sessionless  =
 Non-Byte aligned data  =
 Sym raw data path API  =
+Cipher multiple data units =
 
 ;
 ; Supported crypto algorithms of a default crypto driver.
diff --git a/doc/guides/cryptodevs/overview.rst b/doc/guides/cryptodevs/overview.rst
index e2a1e08ec1..e24e3e1993 100644
--- a/doc/guides/cryptodevs/overview.rst
+++ b/doc/guides/cryptodevs/overview.rst
@@ -46,6 +46,9 @@  Supported Feature Flags
    - "Digest encrypted" feature flag means PMD support hash-cipher cases,
      where generated digest is appended to and encrypted with the data.
 
+   - "CIPHER_MULTIPLE_DATA_UNITS" feature flag means PMD support operations
+      on multiple data-units message.
+
 
 Supported Cipher Algorithms
 ---------------------------
diff --git a/doc/guides/rel_notes/release_21_05.rst b/doc/guides/rel_notes/release_21_05.rst
index 2e3bf5c95a..57b2744d1d 100644
--- a/doc/guides/rel_notes/release_21_05.rst
+++ b/doc/guides/rel_notes/release_21_05.rst
@@ -150,6 +150,12 @@  New Features
 
   * Added support for preferred busy polling.
 
+* **Added support of multiple data-units in cryptodev API.**
+
+  The cryptodev library has been enhanced to allow operations on multiple
+  data-units for AES-XTS algorithm, the data-unit length should be set in the
+  transformation. A capability for it was added too.
+
 * **Updated Mellanox RegEx PMD.**
 
   * Added support for multi-segments mbuf.
diff --git a/lib/librte_cryptodev/rte_crypto_sym.h b/lib/librte_cryptodev/rte_crypto_sym.h
index 9d572ec057..4e365b1eab 100644
--- a/lib/librte_cryptodev/rte_crypto_sym.h
+++ b/lib/librte_cryptodev/rte_crypto_sym.h
@@ -195,6 +195,9 @@  struct rte_crypto_cipher_xform {
 	enum rte_crypto_cipher_algorithm algo;
 	/**< Cipher algorithm */
 
+	RTE_STD_C11
+	union { /* temporary anonymous union for ABI compatibility */
+
 	struct {
 		const uint8_t *data;	/**< pointer to key data */
 		uint16_t length;	/**< key length in bytes */
@@ -222,6 +225,27 @@  struct rte_crypto_cipher_xform {
 	 *  - Each key can be either 128 bits (16 bytes) or 256 bits (32 bytes).
 	 *  - Both keys must have the same size.
 	 **/
+
+	RTE_STD_C11
+	struct { /* temporary anonymous struct for ABI compatibility */
+		const uint8_t *_key_data; /* reserved for key.data union */
+		uint16_t _key_length;     /* reserved for key.length union */
+		/* next field can fill the padding hole */
+
+	uint16_t dataunit_len;
+	/**< When RTE_CRYPTODEV_FF_CIPHER_MULTIPLE_DATA_UNITS is enabled,
+	 * this is the data-unit length of the algorithm,
+	 * otherwise or when the value is 0, use the operation length.
+	 * The value should be in the range defined by the dataunit_set field
+	 * in the cipher capability.
+	 *
+	 * - For AES-XTS it is the size of data-unit, from IEEE Std 1619-2007.
+	 * For-each data-unit in the operation, the tweak (IV) value is
+	 * assigned consecutively starting from the operation assigned IV.
+	 */
+
+	}; }; /* temporary struct nested in union for ABI compatibility */
+
 	struct {
 		uint16_t offset;
 		/**< Starting point for Initialisation Vector or Counter,
@@ -701,9 +725,10 @@  struct rte_crypto_sym_op {
 					 /**< The message length, in bytes, of the
 					  * source buffer on which the cryptographic
 					  * operation will be computed.
+					  * This is also the same as the result length.
 					  * This must be a multiple of the block size
-					  * if a block cipher is being used. This is
-					  * also the same as the result length.
+					  * or a multiple of data-unit length
+					  * as described in xform.
 					  *
 					  * @note
 					  * For SNOW 3G @ RTE_CRYPTO_AUTH_SNOW3G_UEA2,
diff --git a/lib/librte_cryptodev/rte_cryptodev.c b/lib/librte_cryptodev/rte_cryptodev.c
index 40f55a3cd0..e02e001325 100644
--- a/lib/librte_cryptodev/rte_cryptodev.c
+++ b/lib/librte_cryptodev/rte_cryptodev.c
@@ -617,6 +617,8 @@  rte_cryptodev_get_feature_name(uint64_t flag)
 		return "SYM_SESSIONLESS";
 	case RTE_CRYPTODEV_FF_NON_BYTE_ALIGNED_DATA:
 		return "NON_BYTE_ALIGNED_DATA";
+	case RTE_CRYPTODEV_FF_CIPHER_MULTIPLE_DATA_UNITS:
+		return "CIPHER_MULTIPLE_DATA_UNITS";
 	default:
 		return NULL;
 	}
diff --git a/lib/librte_cryptodev/rte_cryptodev.h b/lib/librte_cryptodev/rte_cryptodev.h
index ae34f33f69..7e80f4be26 100644
--- a/lib/librte_cryptodev/rte_cryptodev.h
+++ b/lib/librte_cryptodev/rte_cryptodev.h
@@ -95,6 +95,14 @@  struct rte_crypto_param_range {
 	 */
 };
 
+/**
+ * Data-unit supported lengths of cipher algorithms.
+ * A bit can represent any set of data-unit sizes
+ * (single size, multiple size, range, etc).
+ */
+#define RTE_CRYPTO_CIPHER_DATA_UNIT_LEN_512_BYTES             RTE_BIT32(0)
+#define RTE_CRYPTO_CIPHER_DATA_UNIT_LEN_4096_BYTES            RTE_BIT32(1)
+
 /**
  * Symmetric Crypto Capability
  */
@@ -127,6 +135,12 @@  struct rte_cryptodev_symmetric_capability {
 			/**< cipher key size range */
 			struct rte_crypto_param_range iv_size;
 			/**< Initialisation vector data size range */
+			uint32_t dataunit_set;
+			/**<
+			 * Supported data-unit lengths:
+			 * RTE_CRYPTO_CIPHER_DATA_UNIT_LEN_* bits
+			 * or 0 for lengths defined in the algorithm standard.
+			 */
 		} cipher;
 		/**< Symmetric Cipher transform capabilities */
 		struct {
@@ -461,6 +475,8 @@  rte_cryptodev_asym_get_xform_enum(enum rte_crypto_asym_xform_type *xform_enum,
 /**< Support operations on data which is not byte aligned */
 #define RTE_CRYPTODEV_FF_SYM_RAW_DP			(1ULL << 24)
 /**< Support accelerator specific symmetric raw data-path APIs */
+#define RTE_CRYPTODEV_FF_CIPHER_MULTIPLE_DATA_UNITS	(1ULL << 25)
+/**< Support operations on multiple data-units message */
 
 /**
  * Get the name of a crypto device feature flag