|
if PACKAGE_libopenssl-afalg_sync
|
|
comment "Build Options"
|
|
|
|
config AFALG_DIGESTS
|
|
bool "Build support for digest acceleration"
|
|
help
|
|
Digests are fast in software, and accessing AF_ALG adds latency, so
|
|
you'll need a large request (16KB) just to match software speed.
|
|
This increases memory usage, and has problems when process fork
|
|
with open digest contexts (openssh will not work because of it).
|
|
|
|
config AFALG_FALLBACK
|
|
bool "Enable software fallback feature"
|
|
default y
|
|
help
|
|
Use software to fulfill small requests. Using AF_ALG adds latency,
|
|
which makes it slow to perform small requests. Enabling this
|
|
option overcomes this problem, at the cost of increased memory
|
|
and CPU usage. This is a new, experimental feature; if you
|
|
encounter any problem, this is the first option to disable.
|
|
The fallback will fail if you enable this engine alongside
|
|
devcrypto, so you'll not be able to install both at the same
|
|
time if this option is enabled.
|
|
|
|
config AFALG_UPDATE_CTR_IV
|
|
bool "Don't rely on kernel to update CTR IV"
|
|
default y
|
|
help
|
|
Don't count on the kernel driver to update the CTR-mode counter
|
|
(IV). At least one driver does not update the IV as a workaround
|
|
for DMA issues. With this option turned on, the engine will keep
|
|
track of the counter, and the IV will be sent with every update.
|
|
If fallback is enabled, then the counter needs to be updated by
|
|
the engine anyway, and sent with the request everytime there's a
|
|
switch from software to hardware, so this won't bring much gain in
|
|
that case.
|
|
|
|
config AFALG_ZERO_COPY
|
|
bool "Use Zero-Copy Mode"
|
|
help
|
|
Uses a Zero-Copy interface. Even though it is supposed to improve
|
|
performance, actual measurements indicate otherwise.
|
|
endif
|