[PATCH v3 0/4] Use AF_ALG in checksum utilities

[Matteo Croce]

Let md5sum and all sha*sum utilities use Linux kernel cryptographic API via the
AF_ALG address family.

Speed gain depends on the CPU type:

Xeon E3-1265L V2:

    $ truncate -s 2GB 2g.bin
    $ time sha1sum 2g.bin
    752ef2367f479e79e4f0cded9c270c2890506ab0  2g.bin

    real    0m4.829s
    user    0m4.437s
    sys     0m0.391s
    $ time ./sha1sum-afalg 2g.bin
    752ef2367f479e79e4f0cded9c270c2890506ab0  2g.bin

    real    0m3.164s
    user    0m0.000s
    sys     0m3.162s

Xeon E3-1240 v6:

    $ time sha1sum disk.qcow2
    47be301f86c71c20eae4ccc5bab4c02e09e729a4  disk.qcow2

    real  0m28.390s
    user  0m13.352s
    sys  0m1.376s
    $ time ./sha1sum-afalg disk.qcow2
    47be301f86c71c20eae4ccc5bab4c02e09e729a4  disk.qcow2

    real  0m8.373s
    user  0m0.052s
    sys  0m8.308s

Marvell Armada 8040 - MACCHIATOBin Dual shot:

    $ dd if=/dev/zero bs=1G count=10 |time -p sha1sum
    a0b6e2ca4e28360a929943e8eb966f703a69dc44  2g.bin

    real  0m49.390s
    user  0m46.852s
    sys  0m2.076s
    $ dd if=/dev/zero bs=1G count=10 |time -p ./sha1sum-afalg
    a0b6e2ca4e28360a929943e8eb966f703a69dc44  2g.bin

    real  0m15.104s
    user  0m0.052s
    sys  0m15.008s

Correctness of the implementation was tested with the following script:

    algos='md5sum sha1sum sha224sum sha256sum sha384sum sha512sum'
    dd='dd status=none if=/dev/zero'

    for alg in $algos; do
        echo -n "Checking $alg..."
        for bs in 1 31 512 1K 4K 1M; do
            for count in 0 1 1234 4096; do
              hash1=$($dd bs=$bs count=$count |$alg)
              hash2=$($dd bs=$bs count=$count |src/$alg)
              [ "$hash1" = "$hash2" ] || \
                exec echo "\n$alg: hash differs with bs=$bs count=$count"
            done
            truncate -s $bs /tmp/filechk-$$
            hash1=$($alg /tmp/filechk-$$)
            hash2=$(src/$alg /tmp/filechk-$$)
            [ "$hash1" = "$hash2" ] || \
              exec echo "\n$alg: hash differs reading '/tmp/filechk-$$'!"
            rm -f /tmp/filechk-$$
            hash1=$($alg /proc/version)
            hash2=$(src/$alg /proc/version)
            [ "$hash1" = "$hash2" ] || \
              exec echo "\n$alg: hash differs reading from procfs!"
        done
        echo " ok"
    done

Changes since v2:
* added m4 script to check for af_alg in configury
* more strict error checking in af_alg.c
* don't allocate temporary buffer if using sendfile
* adhere to GNU coding style
* don't include more headers than necessary
* don't use sendfile for empty files as it breaks reading from procfs
* use sendfile for shared memory objects too

Changes since v1:
* add copyright nore in af_alg.c
* check for AF_ALG in sys/socket.h to avoid build failures when unavailable
* don't include unneeded header files in af_alg.h
* revert wrong Include directive in module description
* revert unneeded 'Hey Emacs!' blocks
* use correct GNU indentation
* prefer size_t over int to denote memory segments
* avoid possible overflow by checking arguments size
* return -EIO if sendfile() returns a short read/write count
* fix a file descriptor leak when bind() returns error

Matteo Croce (4):
  sha1sum: use AF_ALG when available
  sha256sum: use kernel crypto API
  sha512sum: use kernel crypto API
  md5sum: use kernel crypto API

 lib/af_alg.c          | 115 ++++++++++++++++++++++++++++++++++++++++++++++++++
 lib/af_alg.h          |  49 +++++++++++++++++++++
 lib/md5.c             |  18 +++++++-
 lib/sha1.c            |  17 +++++++-
 lib/sha256.c          |  32 +++++++++++++-
 lib/sha512.c          |  32 +++++++++++++-
 m4/linux-if-alg.m4    |  29 +++++++++++++
 modules/crypto/md5    |   6 ++-
 modules/crypto/sha1   |   6 ++-
 modules/crypto/sha256 |   6 ++-
 modules/crypto/sha512 |   6 ++-
 11 files changed, 306 insertions(+), 10 deletions(-)
 create mode 100644 lib/af_alg.c
 create mode 100644 lib/af_alg.h
 create mode 100644 m4/linux-if-alg.m4

-- 
2.14.3