linux/Documentation/crypto/devel-algos.rst
Eric Biggers c626910f3f crypto: ahash - remove support for nonzero alignmask
Currently, the ahash API checks the alignment of all key and result
buffers against the algorithm's declared alignmask, and for any
unaligned buffers it falls back to manually aligned temporary buffers.

This is virtually useless, however.  First, since it does not apply to
the message, its effect is much more limited than e.g. is the case for
the alignmask for "skcipher".  Second, the key and result buffers are
given as virtual addresses and cannot (in general) be DMA'ed into, so
drivers end up having to copy to/from them in software anyway.  As a
result it's easy to use memcpy() or the unaligned access helpers.

The crypto_hash_walk_*() helper functions do use the alignmask to align
the message.  But with one exception those are only used for shash
algorithms being exposed via the ahash API, not for native ahashes, and
aligning the message is not required in this case, especially now that
alignmask support has been removed from shash.  The exception is the
n2_core driver, which doesn't set an alignmask.

In any case, no ahash algorithms actually set a nonzero alignmask
anymore.  Therefore, remove support for it from ahash.  The benefit is
that all the code to handle "misaligned" buffers in the ahash API goes
away, reducing the overhead of the ahash API.

This follows the same change that was made to shash.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-10-27 18:04:29 +08:00

239 lines
8.7 KiB
ReStructuredText

Developing Cipher Algorithms
============================
Registering And Unregistering Transformation
--------------------------------------------
There are three distinct types of registration functions in the Crypto
API. One is used to register a generic cryptographic transformation,
while the other two are specific to HASH transformations and
COMPRESSion. We will discuss the latter two in a separate chapter, here
we will only look at the generic ones.
Before discussing the register functions, the data structure to be
filled with each, struct crypto_alg, must be considered -- see below
for a description of this data structure.
The generic registration functions can be found in
include/linux/crypto.h and their definition can be seen below. The
former function registers a single transformation, while the latter
works on an array of transformation descriptions. The latter is useful
when registering transformations in bulk, for example when a driver
implements multiple transformations.
::
int crypto_register_alg(struct crypto_alg *alg);
int crypto_register_algs(struct crypto_alg *algs, int count);
The counterparts to those functions are listed below.
::
void crypto_unregister_alg(struct crypto_alg *alg);
void crypto_unregister_algs(struct crypto_alg *algs, int count);
The registration functions return 0 on success, or a negative errno
value on failure. crypto_register_algs() succeeds only if it
successfully registered all the given algorithms; if it fails partway
through, then any changes are rolled back.
The unregistration functions always succeed, so they don't have a
return value. Don't try to unregister algorithms that aren't
currently registered.
Single-Block Symmetric Ciphers [CIPHER]
---------------------------------------
Example of transformations: aes, serpent, ...
This section describes the simplest of all transformation
implementations, that being the CIPHER type used for symmetric ciphers.
The CIPHER type is used for transformations which operate on exactly one
block at a time and there are no dependencies between blocks at all.
Registration specifics
~~~~~~~~~~~~~~~~~~~~~~
The registration of [CIPHER] algorithm is specific in that struct
crypto_alg field .cra_type is empty. The .cra_u.cipher has to be
filled in with proper callbacks to implement this transformation.
See struct cipher_alg below.
Cipher Definition With struct cipher_alg
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Struct cipher_alg defines a single block cipher.
Here are schematics of how these functions are called when operated from
other part of the kernel. Note that the .cia_setkey() call might happen
before or after any of these schematics happen, but must not happen
during any of these are in-flight.
::
KEY ---. PLAINTEXT ---.
v v
.cia_setkey() -> .cia_encrypt()
|
'-----> CIPHERTEXT
Please note that a pattern where .cia_setkey() is called multiple times
is also valid:
::
KEY1 --. PLAINTEXT1 --. KEY2 --. PLAINTEXT2 --.
v v v v
.cia_setkey() -> .cia_encrypt() -> .cia_setkey() -> .cia_encrypt()
| |
'---> CIPHERTEXT1 '---> CIPHERTEXT2
Multi-Block Ciphers
-------------------
Example of transformations: cbc(aes), chacha20, ...
This section describes the multi-block cipher transformation
implementations. The multi-block ciphers are used for transformations
which operate on scatterlists of data supplied to the transformation
functions. They output the result into a scatterlist of data as well.
Registration Specifics
~~~~~~~~~~~~~~~~~~~~~~
The registration of multi-block cipher algorithms is one of the most
standard procedures throughout the crypto API.
Note, if a cipher implementation requires a proper alignment of data,
the caller should use the functions of crypto_skcipher_alignmask() to
identify a memory alignment mask. The kernel crypto API is able to
process requests that are unaligned. This implies, however, additional
overhead as the kernel crypto API needs to perform the realignment of
the data which may imply moving of data.
Cipher Definition With struct skcipher_alg
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Struct skcipher_alg defines a multi-block cipher, or more generally, a
length-preserving symmetric cipher algorithm.
Scatterlist handling
~~~~~~~~~~~~~~~~~~~~
Some drivers will want to use the Generic ScatterWalk in case the
hardware needs to be fed separate chunks of the scatterlist which
contains the plaintext and will contain the ciphertext. Please refer
to the ScatterWalk interface offered by the Linux kernel scatter /
gather list implementation.
Hashing [HASH]
--------------
Example of transformations: crc32, md5, sha1, sha256,...
Registering And Unregistering The Transformation
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
There are multiple ways to register a HASH transformation, depending on
whether the transformation is synchronous [SHASH] or asynchronous
[AHASH] and the amount of HASH transformations we are registering. You
can find the prototypes defined in include/crypto/internal/hash.h:
::
int crypto_register_ahash(struct ahash_alg *alg);
int crypto_register_shash(struct shash_alg *alg);
int crypto_register_shashes(struct shash_alg *algs, int count);
The respective counterparts for unregistering the HASH transformation
are as follows:
::
void crypto_unregister_ahash(struct ahash_alg *alg);
void crypto_unregister_shash(struct shash_alg *alg);
void crypto_unregister_shashes(struct shash_alg *algs, int count);
Cipher Definition With struct shash_alg and ahash_alg
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Here are schematics of how these functions are called when operated from
other part of the kernel. Note that the .setkey() call might happen
before or after any of these schematics happen, but must not happen
during any of these are in-flight. Please note that calling .init()
followed immediately by .final() is also a perfectly valid
transformation.
::
I) DATA -----------.
v
.init() -> .update() -> .final() ! .update() might not be called
^ | | at all in this scenario.
'----' '---> HASH
II) DATA -----------.-----------.
v v
.init() -> .update() -> .finup() ! .update() may not be called
^ | | at all in this scenario.
'----' '---> HASH
III) DATA -----------.
v
.digest() ! The entire process is handled
| by the .digest() call.
'---------------> HASH
Here is a schematic of how the .export()/.import() functions are called
when used from another part of the kernel.
::
KEY--. DATA--.
v v ! .update() may not be called
.setkey() -> .init() -> .update() -> .export() at all in this scenario.
^ | |
'-----' '--> PARTIAL_HASH
----------- other transformations happen here -----------
PARTIAL_HASH--. DATA1--.
v v
.import -> .update() -> .final() ! .update() may not be called
^ | | at all in this scenario.
'----' '--> HASH1
PARTIAL_HASH--. DATA2-.
v v
.import -> .finup()
|
'---------------> HASH2
Note that it is perfectly legal to "abandon" a request object:
- call .init() and then (as many times) .update()
- _not_ call any of .final(), .finup() or .export() at any point in future
In other words implementations should mind the resource allocation and clean-up.
No resources related to request objects should remain allocated after a call
to .init() or .update(), since there might be no chance to free them.
Specifics Of Asynchronous HASH Transformation
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Some of the drivers will want to use the Generic ScatterWalk in case the
implementation needs to be fed separate chunks of the scatterlist which
contains the input data.