AF_ALG zero-size hash fails

[Russell King - ARM Linux <linux@armlinux.org.uk>]

Hi,

While testing AF_ALG with openssl af-alg-rr, I've found that:

OPENSSL_CONF=/shared/crypto/openssl-imx.cnf openssl dgst -sha1 </dev/null

fails with a zero hash result:

socket(PF_ALG, SOCK_SEQPACKET, 0)       = 3
close(3)                                = 0
socket(PF_ALG, SOCK_SEQPACKET, 0)       = 3
bind(3, {sa_family=AF_ALG, sa_data="hash\0\0\0\0\0\0\0\0\0\0"}, 88) = 0
accept(3, 0, NULL)                      = 4
fstat64(0, {st_mode=S_IFCHR|0666, st_rdev=makedev(1, 3), ...}) = 0
ioctl(0, SNDCTL_TMR_TIMEBASE or SNDRV_TIMER_IOCTL_NEXT_DEVICE or TCGETS, 0xbed50d5c) = -1 ENOTTY (Inappropriate ioctl for device)
mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb6fe4000
read(0, "", 8192)                       = 0
read(4, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0", 20) = 20
close(4)                                = 0
close(3)                                = 0

tested with the Freescale CAAM driver with SHA1 and MD5 hashes, and
the ARM SHA1 shash implementation.  Should there at least be a single
write to the socket (of zero size) in this case, or should the kernel
return the correct hash on the first read without a preceding
write/send?

I don't think this is a regression afaik - it's been around for some
time (at least from when I first started to poke at the ARM crypto
stuff around v4.2 or v4.3 time.)

Thanks.

-- 
RMK's Patch system: http://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.