[Bug 216759] [kern] Memory speed with small blocks (1K) up to 35 times slower than host system (under QEMU emulation, but not only)

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=216759

andrew at azar-a.net changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #181048|                            |maintainer-approval+
              Flags|                            |

--- Comment #18 from andrew at azar-a.net ---
Created attachment 181048
  --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=181048&action=edit
tsc.c patch to allow KVM hypervisor with host cpu to pass through the good TSC

This patch is not "correct" because it just whitelists Generic virtualization
(VM_GUEST_VM), however I don't see how it can be done correctly at this point
of overall coding quality in tsc.c regarding virtualization.

I guess a CPU_VENDOR_VIRTUAL should be considered as an option, so we can leave
host pass-through CPU as a real host CPU and allow virtual CPUs like
QEMU_X86_64 or whatever to be treated separately where all ifs and switches are
tested for stability of CPU TSC (that seems to be the Linux approach).

-- 
You are receiving this mail because:
You are the assignee for the bug.