aboutsummaryrefslogtreecommitdiffstats
path: root/arch/s390/include/asm/ipl.h
diff options
context:
space:
mode:
authorHarald Freudenberger <freude@linux.vnet.ibm.com>2017-10-27 15:53:49 +0200
committerHeiko Carstens <heiko.carstens@de.ibm.com>2017-11-08 09:47:51 +0100
commitf44fa88745eda1530083b361e300e1ca4e15a6c5 (patch)
treee71acb270f3b118f2a3d70b5e8d69f4c2395fd6c /arch/s390/include/asm/ipl.h
parents390/pci: do not require AIS facility (diff)
downloadlinux-dev-f44fa88745eda1530083b361e300e1ca4e15a6c5.tar.xz
linux-dev-f44fa88745eda1530083b361e300e1ca4e15a6c5.zip
s390/archrandom: Reconsider s390 arch random implementation
The reworked version of the random device driver now calls the arch_get_random_* functions on a very high frequency. It does about 100.000 calls to arch_get_random_long for providing 10 MB via /dev/urandom. Each invocation was fetching entropy from the hardware random generator which has a rate limit of about 4 MB/s. As the trng invocation waits until enough entropy is gathered, the random device driver is slowed down dramatically. The s390 true random generator is not designed for such a high rate. The TRNG is more designed to be used together with the arch_get_random_seed_* functions. This is similar to the way how powerpc has implemented their arch random functionality. This patch removes the invocations of the s390 TRNG for arch_get_random_long() and arch_get_random_int() but leaving the invocations for arch_get_random_seed_long() and arch_get_random_seed_int(). So the s390 arch random implementation now contributes high quality entropy to the kernel random device for reseeding. Signed-off-by: Harald Freudenberger <freude@linux.vnet.ibm.com> Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Diffstat (limited to 'arch/s390/include/asm/ipl.h')
0 files changed, 0 insertions, 0 deletions