[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200305044713.GA173879@sc2-cpbu2-b0737.eng.vmware.com>
Date: Thu, 5 Mar 2020 04:47:28 +0000
From: Vivek Thampi <vithampi@...are.com>
To: David Miller <davem@...emloft.net>
CC: "richardcochran@...il.com" <richardcochran@...il.com>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
Pv-drivers <Pv-drivers@...are.com>,
"virtualization@...ts.linux-foundation.org"
<virtualization@...ts.linux-foundation.org>,
"x86@...nel.org" <x86@...nel.org>,
Thomas Hellstrom <thellstrom@...are.com>,
"jgross@...e.com" <jgross@...e.com>
Subject: Re: [PATCH RESEND] ptp: add VMware virtual PTP clock driver
Thanks for taking a look, David.
On Wed, Mar 04, 2020 at 02:04:10PM -0800, David Miller wrote:
> From: Vivek Thampi <vithampi@...are.com>
> Date: Fri, 28 Feb 2020 05:32:46 +0000
>
> > Add a PTP clock driver called ptp_vmw, for guests running on VMware ESXi
> > hypervisor. The driver attaches to a VMware virtual device called
> > "precision clock" that provides a mechanism for querying host system time.
> > Similar to existing virtual PTP clock drivers (e.g. ptp_kvm), ptp_vmw
> > utilizes the kernel's PTP hardware clock API to implement a clock device
> > that can be used as a reference in Chrony for synchronizing guest time with
> > host.
> >
> > The driver is only applicable to x86 guests running in VMware virtual
> > machines with precision clock virtual device present. It uses a VMware
> > specific hypercall mechanism to read time from the device.
> >
> > Reviewed-by: Thomas Hellstrom <thellstrom@...are.com>
> > Signed-off-by: Vivek Thampi <vithampi@...are.com>
> > ---
> > Based on feedback, resending patch to include a broader audience.
>
> If it's just providing a read of an accurate timesource, I think it's kinda
> pointless to provide a full PTP driver for it.
The point of a PTP driver for this timesource is to provide an interface
that can be consumed by Chrony (and possibly other applications) as a
reference clock for time sync. This is admittedly a very basic first
step providing just PTP_SYS_OFFSET functionality (which gets us to <1us
VM-host time sync). Down the line, I also intend to introduce
cross-timestamping for more precise offsets.
Using the PHC infrastructure seems like a good fit for this, but please
let me know if I'm missing something or misinterpreting your comment.
Thanks.
Powered by blists - more mailing lists