linux-kernel - Re: [PATCH net-next v4] ptp: add Alibaba CIPU PTP clock driver

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <b4c54718-648b-44e6-baed-2a08f287c24a@linux.alibaba.com>
Date: Fri, 22 Aug 2025 10:07:32 +0800
From: Wen Gu <guwen@...ux.alibaba.com>
To: David Woodhouse <dwmw2@...radead.org>,
 Richard Cochran <richardcochran@...il.com>, Jakub Kicinski <kuba@...nel.org>
Cc: andrew+netdev@...n.ch, davem@...emloft.net, edumazet@...gle.com,
 pabeni@...hat.com, xuanzhuo@...ux.alibaba.com, dust.li@...ux.alibaba.com,
 netdev@...r.kernel.org, linux-kernel@...r.kernel.org,
 Thomas Gleixner <tglx@...utronix.de>
Subject: Re: [PATCH net-next v4] ptp: add Alibaba CIPU PTP clock driver



On 2025/8/22 04:12, David Woodhouse wrote:
> On Thu, 2025-08-21 at 13:05 -0700, Richard Cochran wrote:
>> On Fri, Aug 15, 2025 at 11:38:14AM -0700, Jakub Kicinski wrote:
>>
>>> Maybe it's just me, but in general I really wish someone stepped up
>>> and created a separate subsystem for all these cloud / vm clocks.
>>> They have nothing to do with PTP. In my mind PTP clocks are simple HW
>>> tickers on which we build all the time related stuff. While this driver
>>> reports the base year for the epoch and leap second status via sysfs.
>>
>> Yeah, that is my feeling as well.
> 
> Agreed. While vmclock is presented as a PTP clock for compatibility,
> it's more than that because it can do better than simply giving a
> single precise point in time. It actually tells the guest the precise
> relationship between the hardware counter and real time, much like the
> private data structure exposed from the kernel to vDSO gettimeofday.
> 
> We should work on having the kernel consume that *directly* and feed
> its CLOCK_REALTIME with it. Having hundreds of guests on the same host
> all recalculate the same thing based on a set of points in time,
> obtained through /dev/ptpX or otherwise, is just daft.

Hi David,

How does vmclock work in a bare‑metal scenario, given that there is no
guest–hypervisor architecture?

You mentioned "vmclock over PCI", do you mean passing a PCI device to the
bare‑metal? What is this PCI device, and which driver does it use?

Thanks.