https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=236233
--- Comment #6 from Alex Richardson ---
For now setting vm.pmap.pti=0 in /boot/loader.conf brings the testing time back
down to 25.89s. Meltdown mitigation should not be too important for this server
so I will disable it for now.
I guess s
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=236233
--- Comment #5 from Konstantin Belousov ---
If you disable KPTI by a knob, does the speed loss recover ?
Can you try on bare metal ?
There is no PCID feature advertised by the virtualized CPU, so we flush TLB on
each syscall entry and exit
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=236233
--- Comment #4 from Alex Richardson ---
It is a virtual machine with an Intel CPU: CPU info from dmesg:
Hyper-V Version: 0.0.0 [SP0]
Features=0x870
PM Features=0x0 [C0]
Features3=0x0
XEN: Hypervisor version 4.7 detected.
CPU: Intel(R
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=236233
--- Comment #3 from Konstantin Belousov ---
This does not look too informative. Can you infer which syscalls take longer ?
Or is it some different timing for synchronization ? Or might be something
completely different.
Also, what is th
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=236233
Alex Richardson changed:
What|Removed |Added
Status|New |Open
Summary|Python m
