upgrade problem / cryptodev

[nwestfal]

Trying to upgrade my pogo V2/V4s:

[root@alarm ~]# pacman -Syu
:: Synchronizing package databases...
core 183.0 KiB 5.96M/s 00:00 [################################################################] 100%
extra 2.0 MiB 7.82M/s 00:00 [################################################################] 100%
community 3.7 MiB 7.77M/s 00:00 [################################################################] 100%
alarm is up to date
aur is up to date
:: Starting full system upgrade...
resolving dependencies...
looking for conflicting packages...
error: failed to prepare transaction (could not satisfy dependencies)
:: openssl-cryptodev: installing linux-kirkwood-dt (4.10.4-1) breaks dependency 'cryptodev_friendly'
[root@alarm ~]#

[pepedog]

https://github.com/archlinuxarm/PKGBUILDs/blob/master/alarm/cryptodev-dkms/PKGBUILD

[nwestfal]

thanks

installed cryptodev-dkms. It was also complaining about missing headers, so I also installed linux-kirkwood-dt-headers.

[pepedog]

Is it "solved"?

[nwestfal]

Yes, thank you!

But I guess cryptodev is broken.

[wwenigma]

Any update guys?

[Sergeanter]

Yes. It stopped working completely . I had to uninstall cryptodev-dkms and openssl-cryptodev. SSH was not able to connect using hardware accelerated ciphers aes-128-cbc and aes-256-cbc . I am planning to roll back to kernel 4.8.1-1 this weekend because of cryptodev and another bug with ext4 present in current kernel.

[kriztioan]

@Sergeanter No, at this point the problem is not so much that cryptodev isn't working; your problem is that it is working with kernel 4.10.12-1-ARCH. The SSH-daemon will correctly try to open /dev/crypto, however, the sandboxing prevents it doing so. Changing 'UsePrivilegeSeparation' from 'sandbox' to 'yes' in sshd.conf resolves the issue. It is also wise to add a non-accelerated cipher at the end of the Cipher-list. You can check /proc/interrupts for a count on f1030000.crypto; if its greater than zero it works. Also, connecting with 'ssh -vv' will show you the negotiation pathways.

[Sergeanter]

@kriztioan
Thank you so much for the hint regarding the sandbox.
Changing 'UsePrivilegeSeparation' from 'sandbox' to 'yes' in /etc/ssh/sshd_config did fix sshd connection with accelerated ciphers.
Also, I confirm that acceleration is working properly.
Baseline benchmark with openssl 1.0.2.k-1
$this->bbcode_second_pass_code('', 'openssl speed -evp aes-256-cbc
Doing aes-256-cbc for 3s on 16 size blocks: 1605427 aes-256-cbc's in 3.00s
Doing aes-256-cbc for 3s on 64 size blocks: 485871 aes-256-cbc's in 2.99s
Doing aes-256-cbc for 3s on 256 size blocks: 128527 aes-256-cbc's in 3.00s
Doing aes-256-cbc for 3s on 1024 size blocks: 32593 aes-256-cbc's in 2.98s
Doing aes-256-cbc for 3s on 8192 size blocks: 4089 aes-256-cbc's in 3.00s
OpenSSL 1.0.2k 26 Jan 2017
built on: reproducible build, date unspecified
options:bn(64,32) rc4(ptr,char) des(idx,cisc,16,long) aes(partial) idea(int) blowfish(ptr)
compiler: gcc -I. -I.. -I../include -fPIC -DOPENSSL_PIC -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -Wa,--noexecstack -D_FORTIFY_SOURCE=2 -march=armv5te -O2 -pipe -fstack-protector --param=ssp-buffer-size=4 -Wl,-O1,--sort-common,--as-needed,-z,relro -O3 -Wall -DOPENSSL_BN_ASM_MONT -DOPENSSL_BN_ASM_GF2m -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DAES_ASM -DBSAES_ASM -DGHASH_ASM
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
aes-256-cbc 8562.28k 10399.91k 10967.64k 11199.74k 11165.70k
')
Benchmark with openssl-cryptodev-1.0.2.h-1
$this->bbcode_second_pass_code('', '
openssl speed -evp aes-256-cbc
Doing aes-256-cbc for 3s on 16 size blocks: 48691 aes-256-cbc's in 0.10s
Doing aes-256-cbc for 3s on 64 size blocks: 48619 aes-256-cbc's in 0.13s
Doing aes-256-cbc for 3s on 256 size blocks: 43764 aes-256-cbc's in 0.16s
Doing aes-256-cbc for 3s on 1024 size blocks: 30701 aes-256-cbc's in 0.07s
Doing aes-256-cbc for 3s on 8192 size blocks: 7577 aes-256-cbc's in 0.06s
OpenSSL 1.0.2h 3 May 2016
built on: reproducible build, date unspecified
options:bn(64,32) rc4(ptr,char) des(idx,cisc,16,long) aes(partial) idea(int) blowfish(ptr)
compiler: gcc -I. -I.. -I../include -fPIC -DOPENSSL_PIC -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DHAVE_CRYPTODEV -DHASH_MAX_LEN=64 -Wa,--noexecstack -D_FORTIFY_SOURCE=2 -march=armv5te -O2 -pipe -fstack-protector --param=ssp-buffer-size=4 -Wl,-O1,--sort-common,--as-needed,-z,relro -O3 -Wall -DOPENSSL_BN_ASM_MONT -DOPENSSL_BN_ASM_GF2m -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DAES_ASM -DBSAES_ASM -DGHASH_ASM
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
aes-256-cbc 7790.56k 23935.51k 70022.40k 449111.77k 1034513.07k

')

[kriztioan]

@Sergeanter Of course, today's update to openssh (v7.5) deprecated UsePrivilegeSeparation - forcing it to use the sandbox. Therefore, OpenSSH will kill the forced process when it is trying to open /dev/crypto... I was thinking of rolling back to v7.4, however, I didn't have the pkg in the cache and doing a full makepkg seems a lot of work... In any case, openssl still works work with its recent update.