[Bug 256781] EC2 Nitro: TSC-low timecounter lags

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=256781

--- Comment #44 from Colin Percival <cperciva@FreeBSD.org> ---
I vaguely recall that another VM system implemented 0x40000010 on the basis of
"it's an unofficial standard now".  Maybe qemu?  Credit goes to Alex Graf for
making me aware of this anyway.  A few years ago when I was hacking on
FreeBSD/Firecracker I wrote up a patch to make Firecracker expose that leaf too
-- the only reason it didn't go into the tree was because they were in the
middle of reworking how they exposed CPU types to guests so it wasn't a good
time for the patch to land.  It always annoyed me that Lambda was paying the
cost of a Linux TSC calibration loop on every cold start, but I don't work for
Amazon and there was only so long I could keep waving the "look guys you're
leaving money on the table here" flag.

And yes, we make use of the Xen leaf as well on those systems (not that we
particularly care about this in EC2).

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