On Wed, 8 May 2019 22:30:00 +0200
Mattias Rönnblom <mattias.ronnb...@ericsson.com> wrote:

> On 2019-05-08 22:12, Stephen Hemminger wrote:
> > On Wed, 24 Apr 2019 14:33:19 +0200
> > Mattias Rönnblom <mattias.ronnb...@ericsson.com> wrote:
> >   
> >> diff --git a/lib/librte_eal/linux/eal/eal.c 
> >> b/lib/librte_eal/linux/eal/eal.c
> >> index 161399619..d6bf0e89e 100644
> >> --- a/lib/librte_eal/linux/eal/eal.c
> >> +++ b/lib/librte_eal/linux/eal/eal.c
> >> @@ -1083,8 +1083,6 @@ rte_eal_init(int argc, char **argv)
> >>   #endif
> >>    }
> >>   
> >> -  rte_srand(rte_rdtsc());  
> > 
> > The current model of seeding with TSC will have problems at scale.
> > If a whole rack of machines reboots at the same time because of power cycle,
> > it would not be surprising if several had the same TSC value.
> > 
> > A better solution would be to use the get_random() system call to get
> > seed, the kernel can get entropy from more sources.
> >   
> 
> Sounds reasonable.
> 
> What version of FreeBSD is DPDK supposed to support? getrandom(2) (and 
> its cousin getentropy(3)) seem not be available until FreeBSD 12.

Linux has had it since 3.17 but you often need a syscall wrapper.
Or just read /dev/urandom.

Reply via email to